Sustainability and Innovation Coordinating Editor Jens Horbach University of Applied Sciences Anhalt, Bernburg, Germany Series Editors Eberhard Feess RWTH Aachen, Germany Jens Hemmelskamp University of Heidelberg, Germany Joseph Huber University of Halle-Wittenberg, Germany René Kemp University of Maastricht, The Netherlands Marco Lehmann-Waffenschmidt Dresden University of Technology, Germany Arthur P. J. Mol Wageningen Agricultural University, The Netherlands Fred Steward Brunel University, London, United Kingdom
Sustainability and Innovation Published Volumes: Jens Horbach (Ed.) Indicator Systems for Sustainable Innovation 2005. ISBN 978-3-7908-1553-5 Bernd Wagner, Stefan Enzler (Eds.) Material Flow Management 2006. ISBN 978-3-7908-1591-7 A. Ahrens, A. Braun, A.v. Gleich, K. Heitmann, L. Lißner Hazardous Chemicals in Products and Processes 2006. ISBN 978-3-7908-1642-6 Ulrike Grote, Arnab K. Basu, Nancy H. Chau (Eds.) New Frontiers in Environmental and Social Labeling 2007. ISBN 978-3-7908-1755-3 Marco Lehmann-Waffenschmidt (Ed.) Innovations Towards Sustainability 2007. ISBN 978-3-7908-1649-5
Tobias Wittmann
Agent-Based Models of Energy Investment Decisions
Physica-Verlag A Springer Company
Dr.-Ing. Tobias Wittmann Technische Universität Berlin Institute for Energy Engineering Marchstr.18 10587 Berlin
[email protected]
ISBN 978-3-7908-2003-4
e-ISBN 978-3-7908-2004-1
DOI 10.1007/978-3-7908-2004-1 Sustainability and Innovations ISSN 1860-1030 Library of Congress Control Number: 2008922630 © 2008 Physica-Verlag Heidelberg This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Physica-Verlag. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Production: le-tex Jelonek, Schmidt & Vöckler GbR, Leipzig Cover design: WMX Design GmbH, Heidelberg Printed on acid-free paper 987654321 springer.com
Dedication
This book was written as a PhD thesis at the Technical University of Berlin. Four years of research are compiled in this work. Having discussed various topics and thoughts throughout these years with people around the world, some of them have especially contributed to my work. First, I would like to thank my supervisors Thomas Bruckner and George Tsatsaronis, who have invited me to work with them at TU Berlin. Both have guided me through my research and I owe them much. Second, I would like to thank Christoph Engel from the Max Planck Institute for Research on Collective Goods in Bonn. The discussions with him always moved my work forward and helped me to understand the coherences. Finally, Frank Behrendt chaired my viva-voce. Further, I have spent an inspiring time with my colleagues from the research group Energy Engineering and Protection of the Environment. I would especially like to name Robbie Morrison and thank him for the helpful discussions and the time spend jointly, both in Berlin and Wellington. In addition, my work benefited from the input of my students. I would like to thank Zaida Milena Contreras, Donato Imbrici and Julius Richter. Most of this work was supported by a scholarship from the Foundation of German Businesses. Providing the money necessary and opening a creative and inspiring environment are just some of the several benefits I have received. Last but not least, I would like to thank Kate and my family for supporting me throughout this work and giving me confidence.
Tobias Wittmann
Berlin, September 2007
Abstract
st
At the start of the 21 century societies face the challenge of securing an efficient and environmentally sound supply of energy for present and future generations. Sector deregulation, the emergence of novel distributed technologies, firms focusing on these new options and competing in selected markets, and the requirements to reduce energy related greenhouse gas emissions might change the structure of energy systems significantly. Densely populated urban areas, which allow for the operation of sophisticated energy infrastructures are the most suitable to see essential changes in their energy infrastructure. This book develops a new model to study the development of urban energy systems. It combines a technical, highly resolved energy system model with an agent-based approach. The technical, highly resolved energy model is used to simulate the operation of technologies. Different agents are developed to capture the investment decisions of actors. Two classes of actors are distinguished: private and commercial actors. The decisions of private actors are modeled using a bounded rational decision model which can be parameterized by socio-demographic surveys. The decisions of commercial actors are approached with a rational choice model, but taking into account different perspectives of firms with regard to future market developments. A proof of concept implementation demonstrates the potential of the developed approach. Diffusion curves for conversion technologies and efficiency upgrades in the residential sector were obtained and the overall energy savings were calculated. Further, the impact of firms’ competition on diffusion curves could be estimated and different business models were tested.
Contents
1 Drivers of Change and Energy Models................................................. 1 1.1 Introduction ...................................................................................... 1 1.2 Drivers of Change............................................................................. 2 1.2.1 Introductory Remarks ................................................................ 2 1.2.2 Market Deregulation.................................................................. 2 1.2.3 Technological Change ............................................................... 3 1.2.4 Energy Firm Conduct ................................................................ 5 1.2.5 Climate Policy ........................................................................... 8 1.3 Energy Models – a Review of the State of the Art ........................... 9 1.4 Motivation and Research Questions ............................................... 11 2 Model Design ......................................................................................... 13 2.1 Introduction .................................................................................... 13 2.2 Users and Intentions ....................................................................... 13 2.3 Geographical and Socio-economic Scope ...................................... 14 2.3.1 Introductory Remarks .............................................................. 14 2.3.2 Geographical Scope................................................................. 15 2.3.3 Socio-economic Scope ............................................................ 15 2.4 The Layer Concept ......................................................................... 16 2.4.1 Basic Concept .......................................................................... 16 2.4.2 Modeling Timeframes ............................................................. 17 2.4.3 Technical Layer ....................................................................... 19 2.4.4 Agent Layer ............................................................................. 23 2.4.5 Energy Markets ....................................................................... 26 2.4.6 Financial Incentives and Regulations ...................................... 28 2.5 Discussion....................................................................................... 28 3 Private Actor Model ............................................................................. 31 3.1 Introduction .................................................................................... 31 3.2 Private Energy Investment Decisions ............................................. 32 3.2.1 Introductory Remarks .............................................................. 32 3.2.2 Neoclassical Perspective.......................................................... 33 3.2.3 Behavioral Perspective ............................................................ 34
X
Contents
3.3 Bounded Rational Decision Models ............................................... 36 3.3.1 Introductory Remarks .............................................................. 36 3.3.2 Goals........................................................................................ 36 3.3.3 Search Rules ............................................................................ 37 3.3.4 Analysis Tools ......................................................................... 38 3.3.5 Decision Strategies .................................................................. 40 3.4 Modeling Private Energy Investment Decisions ............................ 41 3.4.1 Introductory Remarks .............................................................. 41 3.4.2 Aggregation of Technology and Infrastructure Information ... 42 3.4.3 Aggregation of Socio-economic Information.......................... 44 3.5 Results ............................................................................................ 51 3.5.1 Introductory Remarks .............................................................. 51 3.5.2 General Decision Matrix ......................................................... 51 3.5.3 Single Decision Outcomes....................................................... 55 3.5.4 Aggregated Decision Outcomes .............................................. 56 3.6 Discussion....................................................................................... 64 4 Commercial Actor Model..................................................................... 67 4.1 Introduction .................................................................................... 67 4.2 Commercial Energy Investment Decisions .................................... 68 4.2.1 Introductory Remarks .............................................................. 68 4.2.2 Theoretical Background .......................................................... 68 4.2.3 Empirical Evidence ................................................................. 71 4.3 Aggregation of the Firm ................................................................. 73 4.3.1 Introductory Remarks .............................................................. 73 4.3.2 Aggregation of Options ........................................................... 73 4.3.3 Aggregation of Business Units ................................................ 75 4.3.4 Definition of Strategies and Perspectives ................................ 77 4.4 Decisions Model ............................................................................. 79 4.4.1 Basic Concepts ........................................................................ 79 4.4.2 Operational Decisions.............................................................. 79 4.4.3 Low-stake Structural Decisions............................................... 80 4.4.4 High-stake Structural Decisions .............................................. 82 4.5 Application and Results.................................................................. 82 4.5.1 Impact on the Decisions of Private Agents.............................. 84 4.5.2 Impact of Competition............................................................. 88 4.6 Discussion....................................................................................... 92 5 Conclusions............................................................................................ 95 5.1 Introduction .................................................................................... 95 5.2 Discussion of the Model Design..................................................... 95 5.3 Outlook ........................................................................................... 97
Contents
XI
Appendix................................................................................................... 99 A.1 Private Actor Model ...................................................................... 99 A.1.1 Supply Superstructure and Networks ..................................... 99 A.1.2 Agent Specific Diffusion Curves.......................................... 102 A.1.3 Results from Weighted Adding Strategy .............................. 103 A.2 Commercial Actor Model ............................................................ 105 References............................................................................................... 107
List of Abbreviations, Symbols and Indices
Abbreviations CO2 CFO deeco EFOM EU FERC GEMS LEX MARKAL OECD SAT SO2 TIMES UNFCCC WADD
carbon dioxide chief financial officer energy model: dynamic energy, emissions, and cost optimization Energy Flow Optimization Model European Union United States Federal Energy Regulatory Commission energy model: German Electricity Market Simulation decision strategy: lexicographic strategy energy model: Market Allocation Organisation for Economic Cooperation and Development decision strategy: satisficing strategy sulfur dioxide energy model: The Integrated Markal Efom System United Nations Framework Convention on Climate Change decision strategy: weighted adding
Symbols a A al AL AT b BU c C CP cs DS e E g G
year number of accesses to network infrastructure aspiration level set of aspiration levels set of analysis tools business unit set of business units contract cash-flow available investment capital share of available investment capital set of decision strategies energy carrier set of energy carriers goals set of decision goals
XIV
List of Abbreviations, Symbols and Indices
i I JND n npv O p ǻp RD
sˆ S SR SD t T u Į ȕ Ȗ IJ Ȧ
interest rate investment cost just noticeable difference number of clients net present value set of options price change in price set of reference domains strategy vector set of strategies set of search rules set of search domains time time horizon utility expected responses towards advertising for a certain contract expected sensitivity towards price changes for a certain contract expected sensitivity towards network extensions pay-back period weight factor
Indices b c conventional e el f g ref R&D t th
business unit contract conventional capital energy carrier electrical firm goal reference research and development capital time thermal
1 Drivers of Change and Energy Models
1.1 Introduction Primary energy demand has been continuously growing over the last century and is expected to grow further.1 A secure supply of energy is a condition for stability and growth of any economy (Ayres et al. 2003). Energy, as it occurs in nature, is rarely suited to provide energy services; most forms of energy need to be transformed first. Further, energy is not always found or cannot be transformed cost-effectively close to demand. As a result, a large international industry that extracts, transports, transforms, and supplies energy has developed. st At the start of the 21 century societies face the challenge of securing an efficient and environmentally sound supply of energy for present and future generations. The high dependency on fossil resources and their decreasing reserves, the prospect of climate change and local pollution, but also the development of sound technologies and the deregulation of energy markets have created a demanding environment for researchers, governments, and firms dealing with energy. Classic issues like financial cost, environmental protection, and supply security are nowadays accompanied by institutional issues like effective regulation and network access, the invention and diffusion of new technologies, and the emergence of decentralized structures. The development of a consistent public energy policy framework, of successful long-term company strategies, and of research and development priorities requires that the various complexities involved are suitably addressed. Results obtained from energy models may provide useful insights to decision makers. Energy models address different questions and have various scopes, ranging from game theoretical analysis of single market competition (Höffler and Wittmann 2007) to international, intertemporal models of interconnected technical systems and markets (Hamacher et al. 2001). Sophisticated energy models have to account for the relevant actors, technologies, markets, and drivers of change in the area they address.
1
The International Energy Agency anticipates the demand for primary energy to grow at 1.7% per annum until 2030. The growth rate over the last three decades has been 2.0% (IEA 2004).
2
1 Drivers of Change and Energy Models
A survey of the major drivers of change in the energy industry suggests that urban areas are the most likely to see essential changes in their energy infrastructure. Nonetheless, sophisticated energy models addressing the future development of urban areas are still insufficient. This work develops an energy model to be used to investigate how energy systems in industrialized countries might develop over the next 20–50 years. It offers an agent-based, spatially highly resolved model to estimate the dynamic structure of future energy systems and markets in cities. It can account for the most relevant effects arising from sector deregulation, the invention and diffusion of distributed technologies, firms conduct, and policy interventions addressing climate change. The remainder of this chapter is structured as follows. First, an overview over the main drivers of change in the energy industry – deregulation, technological change, firms conduct, and climate policy – is given, followed by a survey of the state of the art of related energy models. Finally, the motivations and research questions for this work are discussed.
1.2 Drivers of Change 1.2.1 Introductory Remarks Scientific progress, technological change, economic growth, and globalization are generally viewed as the main drivers of change. Further, public policy can stimulate or hinder these developments (Freeman and Soete 1997). Today’s energy industry is undergoing fundamental institutional, commercial, and technological developments. These changes can basically be attributed to four main drivers: market deregulation (Pfaffenberger and Sioshansi 2006), technological change (Grubler et al. 1999), energy firms (Christensen 2000), and climate policy (IPCC 2001). The importance and impact of each of these areas are discussed and summarized below. 1.2.2 Market Deregulation The deregulation of energy markets started about 30 years ago, beginning in North America and spreading towards the European Union. The United States Federal Energy Regulatory Commission (FERC) issued Order 436 in 1985 that ensured open access to all interstate natural gas pipelines for local distribution companies, gas producers, marketers and large volume customers. Access to pipeline capacities was attributed on a first-comefirst-served basis. With the deregulation of the gas pipeline transportation services completed, FERC moved to deregulate the bundled ancillary gas
1.2 Drivers of Change
3
services such as storage and extraction, in association with interstate delivery. In 1992, Order 636 was introduced to provide for the unbundling of those ancillary services as well as to prohibit the interstate pipeline companies to own gas for resale. This natural gas deregulation provided a template for FERC with regard to the development of competitive wholesale electricity markets and open access to transmission capacity. The Electricity Title of the Energy Policy Act became law in 1992 and resulted in major changes in the market for electricity generation and retail access. This granted FERC authority to allow the transmission of power from new independent wholesale generators, and started the move toward a wholesale electricity market. The concept of integrated resource planning was promoted by state regulators to enhance the move toward retail competition and to provide an initiative to industry players to restructure their gas and electric utilities in order to promote wholesale and retail competition and customer choice. In contrast to the United States, the European Union (EU) started to deregulate electricity, not gas, markets in the 1990’s by adopting directive 96/92/EC by the European Parliament and the European Council in 1996. This directive specified common rules for the internal market for electricity and was revised and replaced by directive 2003/55/EC in 2003. Further, regulation 1228/2003 adopted in 2003 sets conditions for network access regarding cross-border exchanges of electricity. Directive 2005/89/EC, which addresses measures to safeguard the security of electricity supply and infrastructure investment within the EU, was adopted in 2006. Likewise, the gas market was deregulated in 1998 by the adoption of directive 98/30/EC, which specified common rules for the internal market of natural gas. This was revised and replaced by directive 2003/55/EC in 2003. Regulation 1775/2005, adopted in 2005, sets out the conditions for access to the natural gas transmission networks within the EU. Thesis I – Deregulation: Sector deregulation has changed the institutional settings of the energy industry considerably by unbundling the vertically integrated energy companies, assuring non-discriminatory third-party network access, fostering resale and retail competition, enabling consumers to choose their supplier, and facilitating cross-border energy trading. Thus, the energy industry is not in an economic equilibrium. 1.2.3 Technological Change Energy services such as movement of goods and people, a comfortable indoor temperature, or task lightning are provided by a range of different technologies. Each technology relies on some sort of energy input since
4
1 Drivers of Change and Energy Models
energy is necessarily conserved. Therefore, technological change in the energy industry can affect the whole value chain, from fuel extraction, transportation, transformation, and delivery to usage. This section concentrates on recent developments in technologies which transform fuels into heat and electric power and which either have or might gain considerable market shares. In order to understand the importance of technological change, it is useful to distinguish between central and distributed energy technologies. Central generation units have capacities well above 10MWel, are connected to the transmission network and require considerable capital investments. Most central generation units are thermal power plants fueled by lignite, hard coal, uranium, gas or oil. Distributed energy technologies are small to medium-scale technologies, usually located within distribution networks. They do not necessarily require well developed transmission networks and can be realized with smaller investments. Examples of distributed technologies include: micro-gas turbines, reciprocating engine cogeneration, solar cells, solar thermal collectors, micro-hydro generation, stationary fuel cells, and wind generators. While central generation capacities are designed to provide bulk energy as efficiently as possible and at lowest prices, distributed technologies offer a different value proposition to consumers (Bruckner et al. 2005). They are small, usually flexible to operate and may be installed spatially close to demand. Further, some provide cogeneration, the simultaneous supply of heat and power 2. They take advantage of the fact that heat cannot be transported over long distances because of losses, and thus district heating grids have a necessarily limited coverage. Other distributed technologies rely on renewable energy input, and are not affected by shortages in or price increases for fossil fuel supply and carbon emissions pricing. Further, novel and mature technologies can be distinguished. Innovative technologies such as CO2 sequestration, solar cells, fuel cells, and microcogeneration are under development or offered at high prices and thus only under some circumstances cost-effective. Due to experience effects, arising from labor efficiency, standardization, specialization, technical method improvements, a better use of equipment, and product redesign, manufacturing costs are expected to fall (Junginger et al. 2005, Riahi et al. 2004, Zilker et al. 1997). In contrast, efficiency improvements and cost reductions for mature technologies are neither expected to be as high nor to occur as rapidly.
2
Heat can also be extracted with high efficiency from central generation facilities. This option is cost-effective if such stations are located close to demand.
1.2 Drivers of Change
5
Thesis II – Technological Change: New distributed technologies have a range of unique features which cannot be easily provided by central technologies so far. The unit costs of novel technologies are continuously decreasing. Novel distributed technologies are likely to diffuse quickly into energy systems, particularly where the business environment is favorable. 1.2.4 Energy Firm Conduct In his textbook “The Innovators’ Dilemma” Christensen (2000) developed a framework which explains the failure of established and well managed companies to stay atop their industries when they confront certain types of market and technological change. Christensen attributes these failures to an inability to cope with the emergence of disruptive technologies in contrast to well developed skills in dealing with sustaining technologies. Sustaining technologies focus on improved product performance. Sustaining technology innovations can be discontinuous and radical in nature or incremental in character. But all sustaining technologies have in common that “they improve the performance of established products, along the dimension of performance that mainstream customers in major markets have historically valued” (Christensen 2000, page xviii). Typically, most innovations in an industry are sustaining in nature. Occasionally, disruptive technologies emerge. Disruptive technologies are characterized by worse product performance, that is, they underperform established technologies in major markets. However, they bring along a very different value proposition in comparison to what has been available previously. Those new features may be of interest to a few niche and often new customers. Further, Christensen states that disruptive technologies are typically cheaper, simpler, smaller, and frequently more convenient to use. Some examples of sustaining and disruptive technologies are given in Table 1.1. Table 1.1. Examples of sustaining and disruptive technologies Sustaining technology Silver halide-based photographic film Landline telephony Electric utility companies Microsoft Windows operating system and application software written in C++ Graduate schools of management
Disruptive technology Digital photography Mobile telephony Distributed power generation Internet protocols and Java applet model
Corporate universities and in-house management training programs Adopted from Christensen (2000, page xxix).
6
1 Drivers of Change and Energy Models
The recent market launches and ongoing diffusion of distributed technologies such as micro-gas turbines, micro-cogeneration units, Sterling engines, solar cells, solar thermal collectors, wind generators, pellet boilers, and fuels cells can be framed as the emergence of disruptive technologies. They contrast improvements in sustaining technology settings such as the central generation of electricity in combination with conventional building heating systems. Distributed technologies have in common that they are comparatively small, and can be operated close to demand. Further, distributed technologies are easy to operate, they mostly run independently and maintenance contracts can be signed with specialized firms at appropriate cost. Equipped with such advantages, distributed technologies are attractive to a range of new customers. The required capital is rather small and can be accumulated by small and medium sized firms; even residential building owners can constitute suitable purchasers. By investing in distributed technologies, investors mostly replace the demand for a high cost energy carrier by the demand for technology and a lower cost energy carrier. This is especially true for technologies which transform renewable resources like wind, water, and solar energy into heat or electricity. Thus, they transform non-cost energy into a marketable good. Instead of price concerns about their supply chain, investors and operators face the risk of high cost technology investments. Additionally, distributed technologies enable operators to become energy suppliers. Entrepreneurs have profited from the emergence of distributed energy technologies and market deregulation. New start-up businesses have been formed and are successfully competing in the energy market. Likewise, some established firms have taken the opportunity to enter into these new markets as well. Among the rapidly growing firms worldwide are companies manufacturing solar cells. They report growth rates of above 30%, and demand is expected to increase. The world leading manufacturers are Sharp, Kyocera, and BP Solar, but also a number of new entrants, such as Solarworld and Qcells (both German), could gain a foothold. Likewise, the wind power industry has seen constant growth rates. In contrast to the photovoltaic industry, Siemens and GE Energy, two large electrical equipment suppliers, recently entered the wind market purchasing established manufacturers 3. But there is still a range of dedicated wind-turbine builders leading the market. The top manufacturers are Vestas (Denmark), Gamesa (Spain), Enercon (Germany), GE Energy (USA), Siemens (Denmark), and Suzlon (India). The average plant size has increased from 30kW in 1980 to 1– 3MW on-shore and 1.5–5MW off-shore. Consequentially, blade diameters have risen to 125m. Costs have declined by 12–18% with each doubling of 3
GE Energy bought Enron Wind in 2003 and Siemens bought Bonus in 2004.
1.2 Drivers of Change
7
the global capacity, thus since 1990 costs have been cut by half (Worldwatch Institute 2005). Solar cells and wind turbines have been installed by commercial and private investors in various regions of the earth. Supported by different market introduction programs, renewable energy technologies enable operators to make profitable investments and, in some cases, sell energy to consumers and thereby become energy suppliers4. Further, micro-cogeneration units were developed for private and commercial investors who have to satisfy their heat demand. These systems, which are operated in homes or small commercial buildings, are mostly driven by heat demand, delivering electricity as a by-product. Heat storages can be integrated to flatten demand peaks or to increase the electricity generation. Reciprocating engines with capacities around 5kWel and 12kWth with electrical efficiencies ranging from 25–30% and thermal efficiencies around 60% are commercially available. Further, Stirling engines with a capacity of 1.2kWel and 8kWth are close to market entry and fuel cells are under development 5. Micro-cogeneration also enables private and commercial consumers to become energy suppliers. Further, there might be attractive contracting options for energy firms to enter the generation market and to serve consumers via long-term heat supply contracts (Pehnt et al. 2006). Thesis III – Energy Firm Conduct: Distributed technologies are disruptive technologies with the potential to fundamentally change firms, markets and energy systems. Therefore, firms’ status quo, their perspectives on the future market needs, and their ability to cope with disruptive technologies are likely to play an important role in the future trajectory of energy systems.
4
Worldwatch Institute (2005) estimates the energy cost for solar cells to be 0.16– 0.32€/kWh and for wind turbines to be 0.03–0.05€/kWh depending on local conditions. 5 For example Senertec (www.senertec.com), Ecopower (www.ecopower.de), and Climate Energy (www.climate-energy.com) offer micro-cogeneration units based on reciprocating engines. WhisperGen (www.whispergen.com) offers a Stirling engine micro-cogeneration unit. Vaillant (www.vaillant.com) and Sulzer Hexis (www.hexis.com) are commercializing fuel cell micro-cogeneration.
8
1 Drivers of Change and Energy Models
1.2.5 Climate Policy As part of the Conference on Environment and Development held in Rio de Janeiro in 1992, the United Nations Framework Convention on Climate Change (UNFCCC) was signed with the aim of reducing emissions of greenhouse gases in order to avoid dangerous anthropogenic interference with the climate system 6. Since this first official recognition of climate change much has happened. One of the most important developments from the original UNFCCC is the Kyoto Protocol negotiated at the third Conference of the Parties in December 1997 in Japan. The signing nations agreed that a joint effort of the industrialized countries (as specified in Annex I of the UNFCCC) will be undertaken to reduce greenhouse gas emissions by 5.2% over the period 2008–2012 compared to 1990 levels. The Kyoto Protocol let to the development of a range of policy measures aiming to reduce greenhouse gas emissions worldwide. But even countries which did not sign or ratify the Kyoto Protocol, like Australia and the United States, have undertaken actions to combat climate change 7. Emissions trading is similarly regarded to be an efficient means to reduce greenhouse gas emissions. The introduction of emissions rights, which are to be consecutively reduced in each trading interval, and a trading platform, encourages emissions reduction where cheapest. The European Union has successfully established such a scheme amongst member states8. Emissions arising from road travel, private sector room heating, and agriculture, which are not currently included in emissions trading, may be addressed through, for instance, energy taxes and energy efficiency programs. Further, governments are seeking to increase the utilization of renewable energies and cogeneration by supporting investments or rewarding the feed-in of electricity. In addition to the emissions reductions realized immediately, the support of wind generators, solar cells, solar thermal collectors, biomass generation or cogeneration should also stimulate technological development and might lead to significant cost reductions in manufacturing and thereby enable long-term cost-effective emissions reduction.
6
See http://www.unfccc.int for details on the UNFCCC. Australia, India, Japan, China, South Korea, and the United States engaged in the Asia-Pacific Partnership on Clean Development and Climate in 2005, formed to establish a cooperation on the development and transfer of technology which potentially results in the reduction of greenhouse gas emissions. 8 The European Union Emissions Trading Scheme is based on Directive 2003/87/EC and is described at http://ec.europa.eu/environment/climat/emission.htm. 7
1.3 Energy Models – a Review of the State of the Art
9
Thesis IV – Climate Policy: Efforts to stabilize and reduce greenhouse gas emissions will be continuing. Potential and actual policy instruments could price the utilization of non-renewable fuels and efficient and renewable technologies would benefit from investment support, feed-in tariffs or supportive tax measures.
1.3 Energy Models – a Review of the State of the Art Energy models in most cases attempt to account for the above summarized drivers of change in different ways. This section discusses three bottom-up modeling approaches which are related to this work and which have recently been applied to study energy systems: technical high-resolution energy system models, intertemporal optimization models, and agent-based simulations. Technical high-resolution energy system models are suited to the study of energy systems operations and to identify synergies and counteractions between technologies. Further, they might be used to estimate the effect of exogenous price changes, investment, decommissioning, taxes, and support schemes on the operation and performance of energy systems. Technologies are modeled using input-output relations, taking ambient conditions and energy intensities like heat flow attributes into account. Demand profiles are usually specified by a one hour resolution for a representative year. The optimal share of technologies to supply a given demand is either determined using optimization routines or calculated by applying control heuristics. Technical high-resolution energy system models attempt to simulate the real performance of an energy system as close as necessary. Therefore, they may be used to support plant scheduling, to support market bidding and to control the real time dispatch. This approach is realized in models like deeco (Bruckner et al. 2003, Bruckner et al. 1997, Groscurth et al. 1995), TopEnergy (Augenstein et al. 2005), and BoFit (Scheidt et al. 2004, Stock and Mertsch 1997). Technical high-resolution energy system models are especially suited to account for the context sensitive performance of distributed technologies (Bruckner et al. 2005). If large energy systems are to be investigated or the long-term evolutions of such systems are to be studied, data and computing time may be considerable. In contrast, intertemporal energy system optimization models are supposed to study the evolution of energy systems over a given time period with regard to different price scenarios, policy frameworks, and technical innovations. In this case a superstructure representing all possible technologies and interconnections of the system under investigation is articulated. Technologies are modeled by average-valued input-output relations.
10
1 Drivers of Change and Energy Models
In most cases, the operation of energy technologies is only simulated for representative days using typical load profiles. The structural development of the entire system is then determined using an optimization routine, taking the average operation, investment, and decommissioning of technologies into account. Intertemporal energy system optimization models are suited to the study of the long-term development of large energy systems with respect to different socio-economical scenarios. This approach is realized in models like MARKAL (Seebregts et al. 2002, Fishbone and Abilock 1981), EFOM (Van der Voort et al. 1985), MESSAGE (Messner and Strubegger 1995) and TIMES (Remme et al. 2002). In order to contain computing time and data needs, the structures of energy systems are simplified, single technologies are aggregated to technology types, and representatively aggregated demand profiles are used. Therefore, intertemporal energy system optimization models are not well placed to account for distributed technologies. Further, it is assumed that one single rational decision maker with perfect foresight administers the entire system – heterogeneity of actors is not accounted for. The third approach introduces autonomous agents, which interact through defined interfaces. Agent-based models have been used to study the emergence of social phenomena in general (Epstein and Axtell 1996), to assist urban traffic planning (Casti 1997), to estimate water usage patterns and demand profiles (Ernst et al. 2004), and to understand technology diffusion and resource use in the agricultural sector (Berger 2001). Agentbased models differ from the approaches discussed above in that the evolution of a system is determined by repeated interaction of heterogeneous agents. Moreover, agent decision making procedures do not necessarily involve rational choice, but can be based on heuristics instead. Through the introduction of agents these models account for the heterogeneity of actors, they may include different decision algorithms and market interactions, and they may account for distributed technologies and policy frameworks. This broad scope requires models to focus on specific domains in order to contain the demand for data. The agent-based simulation approach has been successfully applied to the energy sector as well. Existing models have addressed the bidding behavior of actors within energy exchanges (Hu 2004, North et al. 2002) or focused on the consolidation within the energy sector (Bower et al. 2001). Another approach uses agents to investigate the long term development of national energy systems (Grozev 2004, Veselka et al. 2002).
1.4 Motivation and Research Questions
11
1.4 Motivation and Research Questions Several key subjects have been introduced in this chapter: sector deregulation, the emergence of novel distributed technologies, firms focusing on these new options and competing in selected markets, and the requirements to reduce energy related greenhouse gas emissions. These four drivers of change might change the structure of energy systems significantly. Moreover as argued by some authors, one might expect a shift from central toward distributed generation structures. As a result, this may lead to a new paradigm in the energy industry (Silberman 2001). Therefore, models addressing the future evolution of energy systems should address the following questions:
• • •
Which distributed technologies percolate into energy systems, what determines the diffusion rate, and how does the demand for and mix of energy change? How does this diffusion alter the ownership structure of generation technologies and what is the likely impact on central generation? To which degree do the status quo of infrastructure and ownership, corporate strategies, and public policy shape the future structure of an energy system and its related emissions, demands, and prices?
Technical high-resolution models, intertemporal optimization models and agent-based models are suited to yield insight into different aspects of the questions outlined above. Nonetheless, combined approaches are still missing. This work proposes a more integrated modeling framework, which accounts for the drivers of change in a novel way.
• • •
This new model will focus on densely populated urban areas, which are the most suitable for the diffusion of distributed technologies. The limited spatial scope enables one to use high-resolution modeling techniques which account adequately for the context sensitive performance of distributed technologies and for the effect of policy measures. Operation, investment, and decommissioning decisions will be undertaken by heterogeneous agents, who supply and demand energy. Competition among firms is modelled as a battle of perspectives.
The subsequent chapters are structured as follows. Chap. 2 will describe the overall model design. A technical layer and an agent layer are introduced. Further, two classes of actors – private and commercial actors – who exhibit distinct energy related behaviors are described. Their agent models are combined with a highly resolved technical energy system optimization model, which simulates the operation of the energy system and
12
1 Drivers of Change and Energy Models
makes it possible to allocate energy, cash, and emission flows to the different agents. Chap. 3 and 4 discuss the decision modeling of private and commercial actors, respectively. Different modeling approaches for each actor class and for operation and investment decisions are used relying on heuristics, rational choice, and bounded rational models. Chap. 5 discusses the results and gives an outlook.
2 Model Design
2.1 Introduction A model is a reduction. It extracts only those elements of a more complex reality which are necessary to reveal the underlying relationships. Stachowiak (1973) characterizes models: (i) to represent an original, (ii) to simplify that original in such way that only the relevant parts are shown, and (iii) to be pragmatic in the sense that simplifications are made, regarding the intentions a user has applying the model in a given time period. This chapter motivates the design of the model presented in this work by characterizing potential users and their intentions. Further, it draws a distinction between the original to be modeled and its environment. All assumptions necessary to transform the original into a model are stated. The model can be used to estimate the future structure of urban energy systems. The overall long-term evolution is obtained by the repeated simulation of investment decisions of building owners and energy firms into energy technologies and energy efficiency upgrades. The chapter is subdivided as follows: Sect. 2.2 shortly characterizes potential future users and their intentions. Sect. 2.3 specifies the geographical and socio-economic boundaries chosen, which separate the original from its environment. Sect. 2.4 introduces the technical layer, which represents the technical energy system, the agent layer, which captures actors’ decisions, and markets, which coordinate actors’ interactions and policies. Some closing remarks are made in Sect. 2.5.
2.2 Users and Intentions The model described here serves as a tool to be applied by potential users to investigate the future development of energy systems in industrialized countries facing ongoing deregulation, technological change, technology diffusion, and public policy interventions addressing, for instance, climate change. The model seeks to replicate the important behaviors and relations that exist and develop within the technical energy system and between the relevant actors.
14
2 Model Design
Local, regional and national governments seeking to increase the efficiency of energy markets, to secure energy supply, to foster technological innovation, to stimulate local and national growth, to increase employment, and to decrease local and global emissions form one group which can inform their decisions by using results from the model users (Andersen 2001, Jank 2000, IPCC 2001). This group implements and modifies regulations, incentives or support schemes to reach short and long-term goals. Thus, questions relating to policy design, interdependencies between different incentives, regulations, and interventions and their impact on the economy are of vital interest. A second group of users from a commercial background may wish to improve the competitive standing of a firm or a technology in energy markets. This group often evaluates, alters, and develops business strategies to gain sustained competitive advantages over rival firms (Midttun 2001). These analyses typically examine the competitive environment of the firm, evaluate the situation of existing and future suppliers and buyers, search for potential new entrants, and assess possible substitutes for products or services (Porter 2004). Thus, questions regarding strategy design, the market potential of new technologies, and the current value and long-term performance of a company are of particular interest.
2.3 Geographical and Socio-economic Scope 2.3.1 Introductory Remarks As outlined in Chap. 1, the sector deregulation, the invention and diffusion of distributed technologies, the conduct of firms, and policy measures addressing climate change may especially affect the highly interconnected urban energy systems found in developed countries. Thus, the model is designed to investigate the evolution of municipal energy systems in industrialized countries. These energy systems have been established to provide industry, commerce and households located in urban areas 9 with fuels and energy services. The following subsections specify the relevant geographical and socio-economic boundaries.
9
Dependent on national norms, urban areas are defined as areas in which the population density exceeds 200 inhabitants/km² or where houses are not further than 200m apart. The total population should be above 50,000.
2.3 Geographical and Socio-economic Scope
15
2.3.2 Geographical Scope A high population density and the resultant large demand for energy services in urban areas permit the operation of sophisticated energy infrastructures such as electricity networks, district heating grids and gas grids to distribute energy among consumers in a cost-effective manner. Likewise, fuels such as heating oil, coal or wood pellets can easily be delivered and prices only slightly depend on the transport. As a consequence, highly interconnected infrastructures develop within urban areas (Graham and Marvin 2001). Energy which is distributed and delivered to consumers located within an urban area is usually extracted and often transformed outside that area. In industrialized countries, a large number of urban areas are connected by transmission infrastructure and jointly supplied by large wholesale markets. Therefore, a distinction may be drawn between the transmission system which interconnects urban areas and the distribution system in the urban area itself 10. A distribution system is characterized by a highly interconnected energy infrastructure, a short distance between stocks and demands, and a high demand density. In contrast, a transmission system has a low demand density and large average distances between stocks and demand11. Assumption I – Geographical Scope: An urban distribution system can be distinguished from the upstream transmission system. The distribution system serves as the original, the transmission system as its environment. Any actions undertaken in the investigated distribution system do not affect intensive properties of the upstream transmission system.
2.3.3 Socio-economic Scope A variety of actors are involved in the operation and development of distribution systems. Those actors do not only directly and indirectly affect 10
The terms ‘transmission system’ and ‘distribution system’, as used in this work, do not only refer to networks such as electricity and gas grids, but are used to cover the whole energy infrastructure including technologies, grids, and transport. 11 The demand density for electricity for the regional East German energy supplier e.dis Energie Nord AG was 328MWh/km²/a in 2001. e.dis supplied 3,257,000 customers. In contrast, the Neckarwerke Stuttgart AG faced a demand density of 10,938MWh/km²/a in 2001. They supplied 2,192,000 customers (ARE 2002).
16
2 Model Design
such systems; they also influence each other. For instance, the revenues of an energy trader depend on the contract selections of consumers; local governments seeking to increase the sustainability of the energy supply offer monetary incentives or educational trainings, which are regularly evaluated and subsequently either prolonged or suspended; firms represent their common interests through lobby associations toward governments aiming at influencing political decisions; finally regulations are altered as a consequence of abuse, failures, new insights or elections. Actors’ interactions are governed through different institutions such as the political system, the judicial system, the media, and markets. Among those institutions, markets for energy are particularly designed to exchange energy and are most relevant for the day-to-day interactions of actors. Actors who do not participate in markets influence the technical energy system and prices indirectly through e.g. policy, regulations, or advertisements. Assumption II – Socio-Economic Scope: The most important interactions influencing the future evolution of distribution systems are governed through markets. Therefore, it is only accounted for actors who are directly participating in energy markets and who demand or supply energy in the distribution system.
2.4 The Layer Concept 2.4.1 Basic Concept Actors can operate their technologies applying different unit commitment protocols and can supply or demand energy. They may change the structure of a distribution system by investing in or decommissioning plants or infrastructure. They might assess technical and financial performance data of each technology to better inform their decisions. The interactions between actors and the energy system occur only through a small number of interfaces. Generally, one can distinguish between the operation of the technical energy system and the actors’ decision making processes, which are not highly interconnected.
2.4 The Layer Concept
17
Fig. 2.1. Technical layer and agent layer
Assumption III – Layers: A technical energy system and a socio-economic actor system can be distinguished. The technical system can be modeled using an energy system model; the decisions of actors can be included using agent-based models. The energy system model is used to inform actors’ decisions and to calculate actors’ cash-flows. The remainder of this section is structured as follows: Sect. 2.4.2 defines timeframes and their alternation, followed by a sketch of the technical (Sect. 2.4.3) and the agent layer (Sect. 2.4.4). Sect. 2.4.5 outlines how the commercial interactions of actors are captured, and finally Sect. 2.4.6 indicates how government policies are integrated into the model. 2.4.2 Modeling Timeframes To coordinate the interaction between the technical and the agent layer a definition of timeframes is necessary. The model is built on three different hierarchic timeframes, the operational timeframe, the structural timeframe and the scenario timeframe. Each of these frames has a different discrete time resolution. The operational timeframe is supposed to simulate the operation of the distribution system. Energy firms apply sophisticated software tools to support their operational decisions. Demands, prices, congestions, and weather conditions are estimated; the dispatch is scheduled and, if necessary, adjusted. Spot markets for electricity enable traders to sell and buy energy in intervals as short as one hour 12 and gas is expected to be traded 12
For example spot contracts at the European Energy Exchange and at the Nord Pool are traded in one hour intervals.
18
2 Model Design
in one hour intervals in the near future. Further, demand profiles can change considerably over one day, over weeks, and across seasons due to individual behaviors, production schedules, and weather conditions13. Finally, the context dependent performance of distributed technologies regarding demand profiles, temperatures, and market prices requires a temporally highly resolved modeling. The structural timeframe is intended to simulate the structural changes which occur in a given distribution system from a single actors’ point of view. Investment in energy technologies is undertaken infrequently. The average lifetime of small boilers, solar panels, solar thermal collectors, wind generators, and cogeneration engines is typically 10–20 years; power stations are designed to operate over 40 years. The construction of energy technologies ranges from 3 months to over 10 years depending on the technology. The energy demand growth has been small and without any discontinuities over the last 30 years in all OECD countries.14 All such investments are undertaken in the structural timeframe. The scenario timeframe is intended to set the maximum period of time of a simulation. It further alternates the operational and structural timeframe. The length of the scenario timeframe depends on the intention of the user and also on the required accuracy of the simulation results. If the timeframe is short, imprecision mostly arises from the aggregation of technologies, infrastructures and the inaccuracy regarding the status quo representation. If the timeframe is longer, imprecision can mostly be attributed to the assumed exogenous development of energy prices, tax rates, support schemes, and technological development, and to the simulation model itself. Further, the scenario timeframe provides time-series for the prices of any energy which is imported into the distribution system.
Fig. 2.2. Hierarchy of timeframes 13
The maximal (minimal) demand for electricity in Germany on the third Wednesday in January 1998 has been 71GW (52GW), in April 66GW (46GW), in July 62GW (36GW), and in October 69GW (45GW) according to Kramer (2002). The related demand curves differ significantly. 14 See http://www.iea.org following the link Energy Information Centre, Country Search.
2.4 The Layer Concept
19
Assumption IV – Modeling Timeframes: A one hour resolution covering each day of the year is suitable as the operational timeframe to simulate the operation of a distribution system appropriately. A one year resolution provides a sufficient structural timeframe to model the investment decisions into energy technologies and infrastructure in a distribution energy system. The scenario timeframe alternately activates the operational timeframe and the structural timeframe. It should cover at least 10 years and should not be extended beyond 50 years.
2.4.3 Technical Layer The technical layer comprises the energy conversion units, storages, and infrastructure of the distribution system under investigation. A process/flow graph is used to represent the network of components which source, store, transport, and transform fuels and/or supply energy services. All components and storages are connected by the required networks. Energy can be imported into and exported from the distribution system. All technical components are aggregated wherever possible in order to reduce the amount of data needed. The technical network is further subdivided into control domains, which reflect the ownership and operational characteristics of the system. Each control domain can be operated using a unit commitment protocol, which determines the dispatch of every technology. Aggregation of Infrastructure and Technologies
To facilitate numerical modeling, the amount of data used to specify the status quo and the investment options is contained by a suitable aggregation of infrastructure, technology, and efficiency options. Urban areas differ significantly with respect to population density, types of buildings and their utilization, distances between buildings, and the energy infrastructure in place. As a consequence, demand density, the number of infrastructure connections per square kilometer and associated costs depend on the location. Roth et al. (1980) developed a typology of neighborhoods, which defines distinct sets of parameters to characterize and distinguish different sections of urban areas. It refers to the predominant type of building stock, distances between buildings, the availability of different infrastructures (e.g. gas grid, district heating grid, electricity grid) and associated connection costs in a specific section.
20
2 Model Design
Fig. 2.3. Aggregation of infrastructure15
The annual heat energy demand of the buildings in any one of the sections mostly depends on the size and type of the building and the year it was built. Over different construction periods, different materials were predominant. Further, the regulation of energy efficiency has evolved over time. Typologies of buildings with distinct annual heat energy demands based on a classification of type (e.g. single family house, semi-detached building, small residential building, large residential building, multi-story building) and construction period (e.g. before 1900, 1901–1918, 1919– 1948, etc.) have been derived for different countries (US DOE 2001, Hake et al. 1999). The relevant data of the present state of a building can be obtained by combining the original building type and construction period, with the energy efficiency measures undertaken so far, and the heating system in place. Energy systems evolve when investment or decommission decisions are undertaken. Different technologies or efficiency measures are normally available. Investors can choose among different types of technologies (e.g. conventional boilers, condensing boilers, heat pumps, cogeneration units, gas turbines, power plants) and they can select different manufacturers. Efficiency options can range from simple maintenance to a complete retrofit of a building. Despite the diversity of products distinct clusters of technology options and efficiency measures can be identified based on statistical data and market surveys (Wenzel et al. 1997).
15
The figure shows how parts of a city can be clustered into neighborhood types based on a satellite photograph. The picture was taken from Google™ Earth (http://earth.google.com/), an internet based software tool offering satellite photographs for almost every city of the world. The clustering should be accompanied by field surveys and interviews with the local infrastructure operators.
2.4 The Layer Concept
21
Assumption V – Aggregation of Infrastructure and Technology: An urban area can be divided into a small number of sections which can be assigned to a representative neighborhood prototype. Each building in a section of an urban area can be assigned to a representative building prototype. The annual energy demand of a section can be obtained by adding up the annual energy demand of each representative building in its present state. Technologies and efficiency measures can be aggregated to investment options which represent the main differences among products in the product range.
Control Domains
Control domains16 cluster network components and connections which are within the responsibility of a single actor and are controlled by the same unit commitment protocol. Thus, each control domain is associated with the agent who operates it. Connected control domains are interfaced by gateways which pass across energy demand, intensity, and price information. Each gateway is associated with one or more legal contract, covering connection, market participation and supply. The control domains form a control domain graph which is operated sequentially. Each upstream control domain has to supply the downstream control domains with energy. Hence, the control domain graph is directed and acyclic in terms of demand transfer.
Fig. 2.4. Example of a simplified control domain graph (cd: control domain)
16
I am thankful to Robbie Morrison for helpful discussions and for suggesting this term to me.
22
2 Model Design
Assumption VI: The energy demand and the technology dispatch within downstream control domains is not affected by the energy demand and technology dispatch in upstream control domains in one time interval. Control domains can therefore be operated sequentially. Unit Commitment Protocols
The technology dispatch in a control domain is determined by the application of a unit commitment protocol. Each control domain is treated independently; its operation is constrained by the components’ capacities, the energy demand within the control domain, and the supply obligations for downstream control domains. Different unit commitment protocols are possible, ranging from simple heuristics over more sophisticated rules like merit order to the application of optimization tools. Simple heuristics may be used when the consumption behavior of households is to be modeled. Heuristics include maintaining a fixed indoor temperature during days and nights, a day-night dependent demand for hot water and electricity services, and daylight dependent use of illumination. The operation of energy conversion units and storage can be determined using a merit order rule which dispatches conversion technologies with respect to their typical marginal costs. More sophistication is obtained when optimization solvers are applied. The technical layer integrates routines which can be used to simulate the unit commitment protocols of different agents. In this work, the dynamic energy, emission and cost optimization model deeco17 (Bruckner et. al. 2003) is used to operate control domains. Energy demand profiles are specified by time-series which can be obtained by applying heuristics, simulation models or real-world historical measurements. With deeco, the cost minimal share of each component of a control domain which meets energy demands is determined by a linear optimization solver. The resolution of deeco is flexible and can range down to 15 minutes. Deeco supports a large number of technologies such as boilers, heat pumps, cogeneration plants, solar thermal collectors, storages, photovoltaic and wind generators, heating grids, and steam and gas turbines. The impact of environment
17
The software tool deeco was developed as a PhD project at the University of Würzburg in Germany by Bruckner (1997). It serves as an analysis tool for different research projects and has provided decision support to energy companies. Presently, the software is maintained and further developed at the Technical University of Berlin, Germany. Other energy system models providing similar features to deeco could be used as well.
2.4 The Layer Concept
23
conditions like temperature, wind speed, and insulation on the performance of technologies may also be included18. 2.4.4 Agent Layer The agent layer comprises models of all actors who demand or supply energy in the distribution system. Each agent is associated with one control domain, located in the technical layer. Agents can select the unit commitment protocol for their control domain and thereby operate it, they have access to the operational data of its components, and they may change its structure by investment or decommissioning decisions. Further, the agent layer includes models for actors who trade energy across the boundary of the distribution system under investigation. Those agents do not necessarily have their own control domain. If a firm possesses more than one control domain which are either not interconnected or operated with different unit commitment protocols, the associated actors are grouped to form a legal entity, to allow common accounting. Actors in a distribution system are heterogeneous and may include consumers, traders, utilities and independent producers. Actors operate their control domain and invest in technology and infrastructure in different ways based on their preferences, knowledge, resources and habits. The payments by private households for energy services are only a small part of their annual consumption expenditure.19 As a result, they do not regularly optimize their portfolio and evaluate energy investments options. In contrast, local utilities or independent producers sell energy as their core business, develop long-term strategies, regularly evaluate investment options, and try to minimize the costs of supplying their customers. Assumption VII – Decisions and Actors: A distinction is drawn between operational and structural decisions. Operational decisions concern the unit commitment protocol and are undertaken in the operational timeframe, structural decisions may alter the structure of a control domain and are undertaken in the structural timeframe. Two classes of actors – private and commercial – are distinguished to build agent-based operational and structural decision models. The decisions of both classes differ considerably with regard to knowledge, financial and technical resources, access to information and preferences. 18 19
A list of deeco features can be obtained from http://iet.tu-berlin.de/deeco. Eurostat (http://epp.eurostat.ec.europa.eu/) estimates the 2004 share of the consumption expenditure of private households for housing, water, electricity, gas, and other fuels to be 21.3% in the EU 25. Transport accounts with 13.5% followed by food and non-alcoholic beverages with 12.7%.
24
2 Model Design
Private actors
Private actors consume energy services and operate small conversion units such as boilers, solar thermal collectors, solar cells, and microcogeneration units to supply their own demand. Additionally, they buy energy from retail markets and sign contracts for longer time periods20. Any operational and structural decision undertaken by a private actor can only marginally affect the entire distribution system. Private actors include households, private building owners, real estate management companies, and small commercial energy demanders. Private actors make operational decisions quite frequently e.g. switching on lights, cooking, washing, watching TV, using hot water, and selecting an indoor room temperature. Mostly such choices are made intuitively; they do not involve careful considerations. Such behaviors developed over a long time and may not be changed easily (Lutzenhiser 1993). The structural decisions of private actors include investment in energy efficiency measures and small energy conversion technologies as well as supply contract selection. To make such decisions, the performance data of different technology and efficiency measure options has to be gathered and the future energy demand and associated costs need to be estimated. Most of the information required cannot easily be found in the environment. Further, private actors only rarely face such situations, because the average lifetime of energy conversion units such as boilers is around 15 years, and renovation cycles of buildings typically range from 25–50 years. In addition, such decisions involve high capital investment which may exceed the budget of private households; therefore loans are often taken out. Finally, the many private actors in a distribution system are heterogeneous. Actors have different budgets and preferences, they differ with respect to their knowledge and their ability to gather and process information. Nevertheless empirical sociological research has revealed that a society can be divided into groups which can be easily distinguished from one another (Bourdieu 1984).
20
Only 3.7% of private consumers switched electricity supply contracts between 1998 and 2001 in Germany, according to Bauknecht (2003). The highest switching rates are reported from Sweden, Norway, and the UK, where 29%, 24% and 13% of consumers switched respectively, between 2000 and 2005, according to Power UK (2005).
2.4 The Layer Concept
25
Assumption VIII – Private Actors: Private actors use heuristics to make operational decisions. Heuristics used for the private actor decision model cannot be altered in the scenario timeframe. The information which is needed to make optimal structural decisions is not necessarily provided by the environment to private actors. Further, private actors differ with respect to their abilities to explore the information provided. Therefore, bounded rational decision models are used to simulate private actor structural decision making processes. Distinct clusters of private actors can be derived empirically, so that one operational and one structural agent decision model can be used to simulate the average decision outcome of all private actors belonging to the same cluster.
Commercial actors
Commercial actors operate and/or invest into energy systems as one of their core businesses. They trade energy across the borders of the distribution system, run different energy conversion units, offer supply contracts to other actors, and have a certain number of clients. Further, they possess an adequate knowledge of the market, develop and adjust their strategies and have access to a sufficient budget. Commercial actors include utilities, gas suppliers, heat suppliers, independent energy producers, and large commercial energy consumers. The operational decisions, such as unit commitment and market bid-set formulation, of commercial actors are often supported by sophisticated software tools 21. Such software holds the technical data of all demands, conversion units, storage, and infrastructure in a control domain, and estimates future demand profiles and market prices, taking ambient conditions and consumer behaviors into account. Some structural decisions such as the investment in small plants like cogeneration units, boilers, and other distributed technologies, small infrastructure extensions, and plant decommissioning are regularly made by commercial actors. Those decisions require only limited investment and do not fundamentally alter the asset profile of the firm in question. Other structural decisions such as contract price offers have a direct and perceptible influence on other market participants. Therefore, strategic considerations play an important role. In contrast, investments in large power stations, and the development of new infrastructure are rarely carried out and have a major impact on the future evolution of the firm. 21
A range of software tools is commercially available. See e.g. http://www.siemens.at/dems/index_en.htm and http://www.procom.de/en/products/bofit for detailed information.
26
2 Model Design
Firms can be conceived as a bundle of resources, which can be classified in three categories: physical resources, human resources, and organizational resources. Physical resources are technologies, infrastructures, plants and equipment, human resources are training, judgment, intelligence, relationships, and insights of individual managers and workers, and organizational resources are the reporting structure, the formal and informal planning, controlling and coordination systems of a firm. Barney (1991) explains the emergence of sustained competitive advantages among firms by the heterogeneity and immobility of some crucial resources. Those resources must be valuable, rare among firms, and imperfectly imitable. Further, resources are only crucial if there are no strategically equivalent substitutes that are as valuable and neither rare nor imperfectly imitable. Moreover, firms seeking to maintain and increase their competitive advantage need to develop their resources in a way that future markets can be exploited. Assumption IX – Commercial Actors: Operational decisions of commercial actors can be simulated using optimization methods in order to find minimal cost solutions. Structural decisions of commercial actors are divided into two classes: low-stake and high-stake. Low-stake decisions are modeled using heuristics based on real-world observations and a rational choice approach. High-stake decisions will be addressed either by treating each decision option as a new scenario or by including human subjects in the run-time decision loop. Key resources can be clustered and attributed to different business units of a firm. A strategy can be modeled by assigning different forms of capital structure (e.g. debt, equity, venture capital, rate of return) to the business unit perceived to be essential for developing a sustained competitive advantage. Structural decision options are evaluated with regard to the set of strategies and either selected or rejected.
2.4.5 Energy Markets Energy which is consumed but not extracted within the distribution system needs to be purchased on energy wholesale markets and imported into the system. Likewise, energy which is extracted or generated from within the distribution system but not consumed needs to be exported from the system and sold on energy wholesale markets. Only commercial agents participate in wholesale markets. Different energy wholesale markets for electricity, oil, coal, gas, and so on exist. Contracts can be signed for a whole number of operational timeframes. Commercial agents can decide to rely on the hourly offers of energy spot markets as well. Usually, the aggre-
2.4 The Layer Concept
27
gated demand of all commercial agents in one distribution system on an energy wholesale market is much smaller than the total demand on that market, which itself provides for the supply of the large number of other distribution systems. In this scenario, energy wholesale markets provide the price information time-series for the entire scenario timeframe exogenously. Two time-series are fed into the energy wholesale markets: hourly price information and the long-term contract price information. If the impact of different price scenarios on national and international markets on the distribution system is to be investigated, the model needs to be executed several times. Wholesale market time-series can be generated by a national or transnational energy system model 22. Retail markets are expected to mediate the commercial interactions between agents in the agent layer. Energy which is exchanged between actors in the distribution system is traded over retail markets. Commercial agents mostly act as sellers; private agents mostly act as buyers. Different retail markets for electricity, oil, coal, gas, heat, wood pellets, etc. exist. Sellers post their contract offers on markets, buyers can choose among the available offers. Contracts can be signed for various numbers of operational timeframes, each agent needs to hold a contract if she is reliant on energy imports into her control domain. If a transaction between agents is agreed, a contract is signed. Each contract is naturally associated with the gateway which interfaces the control domains of the contractual partners in the technical layer. It specifies the conditions of energy exchange regarding prices, capacity and quality. Agents operate their control domain taking into account the actual contract conditions. Contract formation over retail markets involves two steps. First, all suppliers post their offers for an operational timeframe at the same time. Thereafter, buyers can select among offers. Assumption X – Energy Markets: Two markets, the wholesale market and the retail market, can be distinguished. Energy which is imported into the distribution system is purchased on wholesale markets; energy which is exchanged between agents is sold on retail markets. The aggregated demand of all commercial actors in the distribution system under investigation on a wholesale market does not alter the price. Wholesale markets are thus inelastic, prices are exogenously supplied. Once a supplier has posted an offer on a retail market it is firm for at least one timeframe. The offers of all sellers for a specific operational timeframe are public information. The number of contracts which can be signed is not limited. All suppliers post their offers simultaneously.
22
Models such as MARKAL, TIMES and GEMS may be used.
28
2 Model Design
2.4.6 Financial Incentives and Regulations The development of energy systems not only depends on commodity prices. Local and national governments aiming to reach their policy goals implement a range of measures targeting actors’ decision making processes. The model accounts for some aspects of two different measures governments can take to influence the future development of distribution energy systems: financial schemes and market and network regulations. Financial incentives and regulations are always firm during one scenario and cannot be influenced by actors’ decisions. A range of financial schemes can influence actors in energy markets. Taxes increase the price of energy, support schemes provide financial incentives to investors to choose innovative technologies, subsidized interest rates make credits easier, feed-in tariffs reward the generation of electivity from renewable resources, and emissions trading schemes charge for emissions like CO2 or SO2. Regulatory energy acts define rules for network and market accesses. For certain networks, network operators are obliged to connect users to their networks free of charge. In contrast, suppliers have to fulfill a range of criteria in order to be allowed to feed energy into networks. Likewise, the access to markets is restricted. Generators aiming to sell energy have to obtain a license to be able to make contracts with consumers. Both financial schemes and regulations are provided exogenously for each scenario timeframe. If the impact of different policy measures is to be investigated, the model needs to be executed several times. Assumption XI – Financial Incentives and Regulations: Financial schemes are modeled by fixed or possibly progressively increasing or decreasing support schemes for energy, capital or technologies. These changes are exogenously supplied and are unable to be altered endogenously. Regulations are modeled by restricting network or market access to specific types of agent. Regulations are exogenously supplied and cannot be altered endogenously.
2.5 Discussion The assumptions introduced so far enable us to build a computational model exploring the future evolution of urban energy systems in deregulated market environments. The model combines approaches from sociology, economics, and engineering science within an integrated framework
2.5 Discussion
29
that enables the actual structure of energy supply systems, their current operation, and their future development to be modeled. It places agents that utilize local profit maximization routines together with agents that exhibit bounded rationality into a complex setting which itself is characterized by a range of interdependencies arising from the technical system and from market interactions. The decisions of agents alter the operation and structure of the networks connecting them and thereby create a complex and adaptive system. The proposed model design adds a new dimension to energy system modeling. In contrast to the planner-orientated structural optimization approaches which usually rely on a single actor’s rational choice problem with prescribed energy demand scenarios, the decision making process and the interactions between energy providers and consumers are modeled explicitly. This allows for the exploration of the impact of the socioeconomic structure of an urban area on technology diffusion, market size, competition and environmental performance. Further, the expected changes in urban energy systems can be investigated using a high spatial resolution. Technical, infrastructural, economic and socio-economic regional differences within cities as well as differences between cities can be modeled. In addition, the overall model provides detailed decision models to the agent-based simulation domain. Most agent-based decision models applied so far use heuristics which are based on trial-and-error rules or rational choice approaches. Although the model accounts for a range of actors and their interactions, its application has some limitations. Firstly, it is restricted to energy systems within urban areas. The pricing of energy on wholesale and spot markets as well as the emissions in the superordinated systems cannot be simulated and must be supplied exogenously. Further, the set of actors is limited to those who participate in urban energy markets. Interactions between governments, lobby organizations, researchers, and generators and consumers are not included. In addition, not all actor decisions are modeled endogenously; the so called high-stake decisions of commercial actors need to be supplied exogenously as scenarios or human input. Finally, the model requires a considerable amount of structural, technological, and socio-economical data to be parameterized. Nevertheless, the expected outcomes and new insights justify the development efforts undertaken.
3 Private Actor Model
3.1 Introduction A good understanding as to how the private demand for energy carriers (electricity, oil, gas, district heating, wood pellets, etc.) and technologies will change over time is essential for stakeholders. For instance, residential heat demand depends on the insulation standard of the building and the consumption behavior of the occupants. Heat can be supplied by a range of conventional (e.g. gas and oil boilers) and new technologies (e.g. microcogeneration, pellet boilers, solar thermal installations). Electricity demand likewise depends on the technologies available and utilization patterns. The thermal performance of buildings, the conversion technologies available, and consumption profiles may therefore have a major influence on market size, competition levels on supply markets, prices, consumer relations, overall CO2 emissions, and supply security. Research on technology diffusion especially focuses on the question when and how fast things happen. Diffusion problems mostly involve many people making decisions, often in an interdependent manner. Further, no basic reference points, which could be used as a metric to measure the passage of time, are available for such processes. Therefore, most technology diffusion models focus on the stylized fact that the time path of usage usually follows an S-shaped curve: diffusion rates first rise and then fall over time, leading to a period of slow take-up, followed by a relatively rapid adoption and finally to a late period of a slow approach to saturation (Geroski 2000). The two most popular explanations of S-curves are epidemic models of information diffusion, and probit models arguing that differences in adoption time reflect differences in goals, needs and abilities of individuals or firms (Geroski 2000). This chapter develops a probit model of the diffusion of energy technologies and energy efficiency measures in urban areas. It is related to two strands of literature. The first one discusses other probit approaches of technology diffusion such as classic threshold models (Valente 1996), models of firms adaptation (Davies 1979), the Technology Acceptance Model (Davis 1989), the Theory of Planned Behavior (Ajzen
32
3 Private Actor Model
1991), and its application to innovation diffusion (Rogers 1995, Venkatesh et al. 2003). The second strand discusses bounded rational decision models and their empirical foundation. Simon (1956 and 1957) may serve as a starting point to bounded rationality decision making. Further, Gigerenzer et al. (1999) identified non optimizing decision heuristics and showed how they can benefit decision makers. Some further examples from a law context can be found in Gigerenzer and Engel (2006). Finally, Bettman et al. (1998) developed an integrated framework for bounded rational consumer choices. This chapter introduces a bounded rational decision model of energy technology and efficiency diffusion, which ideally would be parameterized using socio-demographic surveys. The bounded rationality approach recognizes that the information needed to take optimal investment decisions is not necessarily provided by the environment to building owners. Further, building owners might differ in their abilities to explore the information provided. Modeling proceeds by distilling the large number of individual decision problems into a number of representative decision problems by aggregating the technological and infrastructural data. This technological aggregation is complemented by a socio-economic clustering which allows the replacement of the large number of individual decision makers by stereotyped decision makers that are representative of the class to which they belong. This chapter presents a bounded rational decision model that enables researches and decision makers to estimate the development of energy demand within the residential building sector with respect to individual investments, indicates how the model parameters might be derived from socio-demographic surveys, and offers some results.
3.2 Private Energy Investment Decisions 3.2.1 Introductory Remarks There is evidence from numerous classic engineering-economic studies that potential investments in energy efficiency, which appear to be costeffective, remain unexploited (Jochem 1999, Interlaboratory Working Group 2000, Productivity Commission 2005, Jakob 2005). Researchers have sought to understand why the observed investment behaviors of building owners differ from estimated scenarios applying different frameworks. The three most important ones are: neoclassical economics, behavioral economics, and institutional economics (Sorrell et al. 2000 and 2004, Weber 1997, Jaffe and Starvins 1994). These frameworks offer a starting
3.2 Private Energy Investment Decisions
33
point to study both the decision maker and the properties and structures of their environments. Table 3.1 introduces each perspective. Table 3.1. Perspectives on energy efficiency investment Perspective Issues neoclassical imperfect information, asymmetric information, hidden cost, risk, heterogeneity of actors behavioral world does not permit optimization, problems are computationally intractable or poorly defined institutional organizational culture, management time and attention
Actors individuals and organizations conceived as rational and utility maximizing individuals conceived as boundedly rational, who apply identifiable rules and heuristics to decision making organizations conceived as social systems influenced by goals, routines, internal culture, power structures, etc.
Adopted from Sorrell et al. 2000
This chapter focuses on investment decisions related to retrofitted energy efficiency measures and energy conversion technologies in the residential sector. These decisions are mostly undertaken by individual households, private building owners, and property management companies. Despite the institutional character of property management companies, this perspective is not particularly relevant to the chosen topic and therefore not applied nor discussed in the remainder of the chapter 23. 3.2.2 Neoclassical Perspective Classic engineering-economic studies rely on the neoclassical perspective. Nonetheless, outcomes deviate from observed behavior patterns so that artificial constants such as a percentage of compliance to standards or fixed construction rates for renewable technologies have to be introduced. The need for those constants is motivated by the difficulties to include imperfect and asymmetric information, the heterogeneity of actors, and to account for the possible presence of hidden costs accurately in the analysis (Kleemann et al. 2000, Diefenbach et al. 2005). Below, some examples of imperfect and asymmetric information and heterogeneity are given.
23
Property management companies maintain and develop buildings as one of their core business. Those companies are rather small compared to large international companies supplying a wide product range to their customers. Thus institutional issues have miner influences on decision making.
34
3 Private Actor Model
Imperfect information: When considering energy efficiency upgrades, building owners are confronted with a wide range of complex products offered by an equally wide range of firms. Retrofitting houses and choosing among different energy supply technologies is a decision task carried out only infrequently by investors, and most technologies will have changed substantially since the previous purchase. Investors also find it difficult to evaluate the performance of those technologies because of their complexity, a lack of detailed energy consumption data, and feedback on current performance (Hewett 1998). In contrast, energy (in the form of fuel, heat, and electricity) is a simple, uniform and easy to understand product supplied from a manageable number of large, well-established and normally trusted firms. Viewing the purchase of energy efficiency and energy supply as different means to deliver energy services (heat, light, mobility, etc.), people tend to over-consume energy supply and under-consume energy efficiency (Sorrell et al. 2004). Asymmetric information: A well known example of asymmetric information is the split incentives problem between landlords and tenants. Landlords might not be willing to retrofit a house to reduce the energy demand because they would not be able to recapitalize their investments by increasing the rent. Adverse selection might also influence the energy service market when the owner believes that potential future tenants are not able to value the energy efficiency standard of an apartment in comparison to the additional rent burden. This can lead to situations where an investment is not made or reduced because of the perceived or actual inability of future tenants to value this investment appropriately. Heterogeneity: Building owners may have a range of independent goals when making investment decisions. These goals might include minimization of financial cost, maximization of comfort, or minimization of environmental impact. Decision makers might evaluate and adjust for the risks and inflexibilities associated with investments differently. These risks include technical risks (reliability, technical performance) and external risks (economic trends, energy prices, policy change). Low income households might face severe budget constraints and are not able to access the necessary capital. Or they might face well above average interest rates (Evry 1997, Sorrell et al. 2004). 3.2.3 Behavioral Perspective Following Simon (1956, 1990) and Gigerenzer and Selten (2001) a different picture emerges. Many real world problems are computationally intractable, poorly defined or involve a high degree of uncertainty, so that optimal solutions are unknown. People “act by habit, imitation of others, and
3.2 Private Energy Investment Decisions
35
trust in institutions, on reputation or a good name” (Gigerenzer and Engel, 2006, p.3) instead. From the behavioral perspective, heuristics and decisions rules are needed to cope with the complexity of the outside world. They may be highly robust and outperform optimization strategies. The term rationality does no longer define if decision makers align with predefined norms as to how they should solve a given problem; it refers to the ability of decision makers to select decision rules or heuristics which perform in a given environment. This so-called ecological rationality was introduced by Gigerenzer et al. (1999). In situations in which information is scarce or costly and where competitive markets are lacking or absent, hence the environment does not permit optimization, decision makers tend to make satisfactory decisions based on aspiration levels use heuristics, routines, or rules of thumb. Further, Bettman et al. (1998) point out that decisions tend to be made with regard to the following meta-goals: maximizing accuracy, minimizing effort, maximizing ease of justification, and minimizing negative emotions. In different decision contexts, one or more of these meta-goals predominate. Subject to the particular decision context, people select decision rules or respective heuristics. As stated earlier, domestic energy-related investment decisions are major decisions made infrequently — say once in every 20–50 years. The predominant meta-goals for this kind of technical decision may be maximized accuracy and minimized effort, which thereby suggest a more complex and alternatives-based decision algorithm relative to more frequently made decisions (Bettmann et al. 1998). In an industrial context, De Almeida (1998) notes that purchasers of electric motors in France tend to choose heuristics depending on the situation — for example, new equipment purchase, routine replacement or emergency replacement. In each setting, firms used different techniques to evaluate energy related investment decisions. Graham and Harvey (2001) interviewed 392 chief financial officers (CFO) and observed that the methods used for capital budgeting vary depending on the size of the firm and the tenure, education, and age of the CFO. Their “finding that payback period is used by older, longer-tenure CFOs without MBAs … suggests that lack of sophistication is a driving factor behind the popularity of the payback criterion” (Graham and Harvey 2001, p. 200). Finally, experiments undertaken by Samuelson and Zeckhauser (1998) and Kahneman et al. (1991) reveal that people prefer to stay with the status quo instead of changing. Stern (1986) noted that people respond to average prices or total costs instead of marginal costs, are more sensitive to changes than to stay with the status quo, and generally require a higher rate of return for smaller investment.
36
3 Private Actor Model
3.3 Bounded Rational Decision Models 3.3.1 Introductory Remarks The concept of bounded rationality was introduced by H.A. Simon with the aim of developing a descriptive model of human economic decision making (1956, 1957). Models of bounded rationality attempt to describe how a judgment or decision is reached by referring to the observed procedures that underlie the non-optimizing adaptive behavior of real people. To do so one has to study both the cognitive abilities of people who face the task of making a decision and the structure of the environment in which this task is carried out. This approach also allows uncovering how a particular choice mechanism is adopted in relation to the properties of the socio-economic environment in which it is made. Models of bounded rationality typically specify three classes of processes: search rules, stopping rules, and decision rules (Gigerenzer and Selten 2001). The search process is modeled as a step-by-step procedure for acquiring pieces of information. The process of searching distinguishes two classes of models — those that search for decision cues and those that search for alternatives. The first class of models is typically better suited to faster and more spontaneous decision processes whereas the latter class can embed a greater level of deliberation (Gigerenzer et al. 1999, Bettman et al. 1998). The search process is terminated by stopping rules. If two or more alternatives remain under consideration at that point, a decision rule is applied to select among the remaining alternatives. The decision model developed in this chapter will estimate the future insulation performance of residential buildings and the available energy conversion technologies. The model accounts for a range of decision characteristics as indicated in section two. Modelers will have to specify different types of agents who exhibit different distinct behavior patterns as well as an explicit set of technology and energy efficiency options which are potentially available to them. The model cannot be used to predict the outcome of a single decision, but it can be used to estimate the outcome of representative decisions of individuals who belong to the same group. The following sections introduce the goals, decision rules, and analysis tools which are used within the model. 3.3.2 Goals Building owners have a range of independent goals when making investment decisions. They face a multi-criteria decision problem; the values
3.3 Bounded Rational Decision Models
37
possible options have regarding one goal (e.g. cost) cannot easily be transformed and added to the values of options regarding another goal (e.g. comfort). Therefore, goals cannot be extinguished; moreover goals often work in opposition. The set of general goals G = {cost, environment, comfort} corresponding to the minimization of cost, minimization of environmental impact, and maximization of comfort is introduced. 3.3.3 Search Rules Search rules determine which alternatives are found by describing the different information gathering habits and abilities of decision makers. Naturally, search rules require additional parameters to be set. This section defines a set of search rules SR = {find_all, find_by_aspects, find_common, find_next} and briefly indicates these new parameters. It should be noted that some search rules can be used in combination. The find_all rule finds all available alternatives. It is especially useful in cases where the decision maker is likely to behave with high rationality. The find_all rule does not require additional parameters to be specified. find_by_aspects acts as a filter24. This search rule finds all alternatives which satisfy preset aspiration levels regarding each goal. Aspiration levels can be set by the internal constraints or requirements of the decision makers themselves, by legislation, or by referring to common practice within society or the decision makers’ reference group. Therefore, a set of reference domains RD = {internal, legislation, society, peer_group} and a set of aspiration levels AL = {alg | g ȯ G} is introduced. Moreover this rule requires having access to an inventory which stores the values taken to be common practice by society at large or by reference groups. find_by_aspects allows the search process to be restricted by defining upper and lower bounds for each goal. It also allows to include budget constraints, soft and hard standards, and similar aspects. find_common locates only those alternatives which are defined to be popular in relation to a given decision maker. To apply this rule, the notion of common needs some further investigation. Firstly, common alternatives can be those alternatives which have been widely selected by society, by the subset of the society the decision makers refer to, or by the decision makers themselves. Secondly, common alternatives can be those alternatives which are topical, that is mostly offered, sold, or advertised recently. On this account, a set of search domains SD = {peer_group, 24
find_by_aspects is related to Tversky’s (1972) decision strategy elimination by aspects.
38
3 Private Actor Model
status_quo, topical} is introduced. When applying find_ common, the search domain needs to be set. If the search domain peer_group is chosen the minimum market share a technology has to achieve to be perceived as common by a group of decision makers needs to be specified. This minimum market share can be different with regard to the peers and the location of the decision makers. The search domain status_quo includes all technologies used in the present state and topical refers to technologies perceived to be new. The rule find_common, also requires an inventory which stores information about the past decisions of society at large, of the relevant peer group, past decisions of each decision maker as well as the alternatives which have recently been mostly offered, advertised or sold. Applied in such a way, find_common allows a range of different information gathering habits of decision makers to be accommodated. The find_next rule finds one alternative which has a particular place in a hierarchy over the available alternatives, starting with the first alternative. The find_next rule requires an order among alternatives to be set. The order of searching can be defined as related to deviations from the status quo of the decision maker or related to the commonness of the alternatives. If it is not possible to define a distinct order over all alternatives available, the respective alternatives will be clustered in such a way that an order over clusters can be defined. The find_next rule will randomly select among the alternatives from the first cluster and then move to the next cluster. Hence, find_next allows the alternatives to be traversed in a preset order. 3.3.4 Analysis Tools The general goals of the decision problem in question include minimizing of cost, minimizing of environmental impact, and maximizing of comfort. Analysis tools determine how a particular set of goals will be assessed by an agent. Given the above mentioned general goals, a range of metrics are available to measure the goal fulfillment of a specific alternative. If decision makers want to select an alternative that minimizes cost, they might refer to different analysis tools to explore the costs associated with the alternatives under consideration, thus ATcost = {investment, operational, payback, npv}. Their decision can be based on investment cost information only. This information is mostly available when selecting among alternatives. No extended search effort or knowledge is needed to locate or process this information. In contrast, they might be interested to contain or minimize
3.3 Bounded Rational Decision Models
39
the operational cost of an investment. To estimate future cash-flows, they could either refer to past payments or calculate the operational cost from usage projections. More detailed information can be obtained if they calculate the payback period which compares investment cost with operational and maintenance cost. The payback period in years IJ is given by the following equation, where I and Iref are the investment costs of the regarded and the reference option, and likewise C and Cref stand for the annual operational cash-flows:
τ =
I − I ref
(3.1)
C ref − C
Using the net present value (npv), decision makers do not only include investment and operational and maintenance costs associated with the alternatives, but compare those costs at their net present value. To apply this method, they need to select a personal discount factor i, a time horizon T, and an estimate of the development of future costs. The net present value can thus be calculated by the following equation, where t = {1, …, T} is the time interval index and Ct the cash-flow at that point in time:
npv =
T
Ct
∑ (1 + i) t =1
t
−I
(3.2)
Generally, the net present value is reported relative to the status quo:
∆npv = npv − npvref
(3.3)
Similarly, the general goal of minimizing environmental impact offers different approaches to be calculated, ATenvironment = {qualitative, consumer_energy, co2}. Again each approach is able to assess the alternatives with different depth. The environmental impact of an alternative can be determined using simple heuristics (qualitative). This approach broadly ranks alternatives. Five categories are proposed ranging from one (low) to five (high) and assign each alternative to one category like, for instance, to the energy source which is used (e.g. oil < district heating < gas < solar, wood pellets). The calculation of consumer energy demand (oil, electricity, gas, etc.) offers a quantitative approach. This metric indicates how much energy enters the building, but it does not include the different impacts of each energy carrier. In contrast, the calculation of the overall CO2 emissions (co2) of each alternative differentiates energy carriers and includes the weighted impact on the greenhouse effect of their use.
40
3 Private Actor Model
The general goal of maximizing comfort will only be qualitatively assessed in this work. It is assumed that options which do not allow for a grid connected supply of the energy carriers have lower values than those which do. An increase of insulation is regarded as an increase in comfort. Further, the cogeneration plant has shorter maintenance intervals and emits noise so that the comfort is lower than those of simpler technologies. Hence, the alternatives are ranked with respect to the supply contracts (e.g. oil, pellets < gas, electricity < district heating), the indoor climate (e.g. no insulation < standard insulated < enhanced insulation), and the available technologies (cogeneration < gas boiler, oil boiler < district heating). Finally, AT = {ATcost, ATenvironment, ATcomfort}. The choice of analysis tools must align with the abilities of decision makers to explore the information accessible in the environment. The analysis tools depend on the goals under investigation and the tools or methods available to explore the accessible information. Different tools can be assigned to different goals reflecting the time, knowledge and motivations that a decision maker has. 3.3.5 Decision Strategies A decision strategy is used to select among the currently identified alternatives by combining one or more search rules and specifying the algorithm which selects alternatives. Note that search rules and decision strategies are linked; some strategies require certain search rules while others do not. Some strategies may require that a specific search rule is altered and the search process repeated. The set of decisions strategies is given by DS = {SAT, LEX, WADD}. The satisficing strategy (SAT) (Simon 1955, Selten 2001) considers each alternative sequentially in the order in which it is revealed in the search process. For each goal of the alternative currently under consideration, the respective value is compared to some predefined aspiration level. If any goal of the alternative fails to meet an aspiration level, that alternative is rejected and the next one is considered. The first alternative which satisfies all aspiration levels is selected. The SAT strategy requires the find_next search rule. The SAT strategy bundles different aspects of decision making behavior. It can, for example, be used to model traditional decision making behavior by ordering the search in relation to deviations from the status quo. The lexicographic strategy (LEX) requires the relative importance of goals to be set. The alternative with the best value on the most important goal is selected. If two or more alternatives are equal, the second most important goal is considered. This algorithm is run until an alternative is se-
3.4 Modeling Private Energy Investment Decisions
41
lected. The LEX strategy can be used in combination with search rules which provide more then one alternative (e.g. find_all, find_common, find_by_aspects, etc.). The LEX strategy can be relaxed by including the concept of a just-noticeable difference (JND). If several alternatives are within the JND of the best alternative of any goal under investigation, they are considered equal. The introduction of a JND allows for some compensation between goals. Selecting one alternative using the LEX strategy allows to model decision makers who have strong preferences over goals. Technology leading behavior can be modeled using the LEX strategy in combination with the find_common search rule while defining commonness as those alternatives perceived to be topical. The weighted adding strategy (WADD) requires the decision maker to assign a subjective utility uog to each option o regarding a goal g. Further, a weighting factor wg to reflect its subjective importance of a goal g is introduced. Thus, for each option:
uo =
∑w
g u og
g
(3.4)
This calculation is repeated for all alternatives. The decision maker then selects the alternative with the highest overall utility uo. In contrast to the previous strategies, WADD is a compensatory strategy. It allows for a good value on one goal to compensate for a poor value on another. This tradeoff process confronts conflicting goals by introducing weighing factors. WADD can be used in combination with a search rule which provides more than one alternative (e.g. find_all, find_common, find_by_aspects). Combining WADD with the find_all search rule will model the greatest degree of rationality.
3.4 Modeling Private Energy Investment Decisions 3.4.1 Introductory Remarks The decision model draws on the ideas of Gigerenzer and Selten (2001). The model combines the decision rules and analysis tools introduced in the previous section to build agents representing different types of actors. Fig. 3.1 illustrates the modeling process, which is structured as follows: 1. Aggregation of the technology and infrastructure information in a way that a general decision matrix consisting of all alternatives and general goals can be construed. Aggregation of the socio-economic
42
3 Private Actor Model
information in such a way that a set of representative agents for distinctive decision maker clusters can be built. 2. For each decision maker cluster, the general decision matrix is transformed into an agent-specific decision matrix, using tailored search rules and analysis tools. 3. Finally a selection is made applying a decision strategy.
Fig. 3.1. Modeling process – aggregation, transformation, and selection
3.4.2 Aggregation of Technology and Infrastructure Information The model is designed to project the future residential energy demand in cities. It is therefore assumed that each district of a city can be assigned to one of the neighborhood types described in Roth et al. (1980). If a district to be studied is composed of different neighborhood prototypes, then further subdivision will be necessary. This typology provides information about the size of and distance between buildings, the availability of different infrastructures (gas grid, district heating grid, electricity grid) and the costs associated with connecting to them. Each building in a given district is then assigned to a building prototype (Hake et al. 1999) which carries information about the annual heat energy demand by introducing a classification of size (single family house, semi-
3.4 Modeling Private Energy Investment Decisions
43
detached building, small residential building, large residential building, multi-story building) and construction period (before 1900, 1901–1918, 1919–1948, etc.). Each construction period refers to the dominating construction material and the technical regulations in place. The present state of the building is determined by the original building prototype and construction period, the energy efficiency measures carried out so far, and the heating system in place. Each typical building in each quarter can be retrofitted and its heating system can be replaced. To capture this evolution, a set of options is construed which reflects the important distinctive steps in the evolution to be modeled. The selection of options was based on observed retrofit investment behavior and the various technologies and configurations available today (Banfi et al. 2005). Three different retrofit options, status quo maintenance (maintenance — the status quo is maintained and only necessary maintenance work is done), standard efficiency measures (standard — all windows are replaced by state-of-the-art windows and the thermal efficiency of the building shell is improved), and enhanced efficiency measures (enhanced — all windows are replaced by state-of-the-art windows and the shell is considerably improved) were selected. The demand for electricity, hot water, and room heating can be supplied by eight different supply options: energy can enter the building in form of electricity, oil, gas, wood pellets or high temperature water by means of a district heating grid (Hea_Grid). Oil can be burned in a conventional boiler (Oil_BoiConv), gas in a conventional (Gas_BoiConv) or condensing boiler (Gas_BoiCond) or in a reciprocating engine (Gas_Cogen-Conv) providing heat and electricity. The hot water provision of the condensing gas boiler can be supported by a small solar thermal collector (Gas_BoiCondSolar), while the entire heat provision can be supported by a large solar thermal collector (Gas_SolarBoiCond). Wood pellets can be burnt in a conventional boiler (Pel_BoiConv) or the heat provision of the pellet boiler can be supported by a large solar thermal collector (Pel_SolarBoiConv). A superstructure of residential supply technologies which includes all supply options has been developed and is presented in the Appendix. Using the aggregation of infrastructure and buildings previously discussed, a number of neighborhood-specific general decision matrixes are construed, each of which holds the options available for typical buildings in their district. Depending on infrastructure and connection rate, for instance, district heating supply might not be available in a certain district or with regard to a given present state of the building, all standard efficiency measures may have already been realized.
44
3 Private Actor Model
3.4.3 Aggregation of Socio-economic Information The aggregation of technical and infrastructural information provides a number of representative decision problems, each differing with regard to the status quo and the set of options on offer, as restricted by technical and infrastructural limitations. Each of these decision problems, duly merged into a set of representative problems, is usually made by the individual building owner. In this section, it is shown how a large number of individual decision makers can be aggregated and modeled using representative agents. The outcome of the decision process for any given agent can vary, even if an identical decision task is faced. The aggregation relies on two concepts which help to select the appropriate goals, search rules, analysis tools, and decision strategies used to reproduce realistic behavior patterns. The first concept is that of social milieus (Bourdieu 1984) and is concerned with distinguishing different homogeneous groups of individuals who share similar aspirations in life, similar value systems, and similar lifestyles. The concept of milieus enables the modeler to perceive people in the richness of their life context and their attitudes towards society, work, family, leisure, money, consumption, and the environment. Therefore, milieus help to determine the available income, the decision goals and their relative importance, and the appropriate reference group. The second concept is that of rationality types 25 which enable us to distinguish between the different abilities of decision makers to choose and apply decision heuristics which perform in a given environment. The rationality type focuses on how the information needed to address a specific decision problem is gathered and processed. It provides insights as to where decision makers search for the relevant information and which techniques they use to assess this information. Further, conclusions on how decisions are reached can be drawn. Hence, the combination of social milieus and rationality types should allow decision rules to be specified for a set of agents in a way that the average outcome of a large number of individual decisions can be reproduced. To design, carry out and evaluate the required surveys which are prerequisite to an empirical determination of different agent models lies beyond the scope of this work. Instead, the specification of agents is carried out on the basis of expert judgments, with each agent type being allocated a short biography.
25
I would like to thank Martin Beckenkamp and Fritz Reusswig for helpful discussions.
3.4 Modeling Private Energy Investment Decisions
45
Parameters from Social Milieu
The milieu classification used in this work refers to the SINUS-MilieuTypology26 which was developed by SINUS-Sociovision. This typology is regularly updated and can be adapted to specific topics as required. To obtain a plausible categorization, some further abstractions and aggregations from the SINUS-Milieus are necessary. A brief illustrative description of the chosen milieus and the assigned parameters is given below. Technology leader: Susan regards herself as being successful in life. Profiting from a higher education, she is creative and inventive and loves to explore new opportunities and technologies. After she finished university, she carefully worked on her career and succeeded. She is around 40 years old and has a partner, Tom. Their household income is high. The couple has friends who share similar values. Susan and Tom spend their income selectively on high quality products. Their circle of friends is interested in technology and chat about innovations regularly. When purchasing, they look for leading products which are discussed in the relevant magazines and so forth. When planning a bigger investment they carefully read the relevant publications and consult experts and friends. They are always willing to experiment. Having a professional background, Susan knows how to evaluate investments. She is concerned with costeffectiveness, but is also willing to spend more than necessary if the product is innovative, has good press, and/or the potential to become a new standard. Susan was attracted to the energy field when she learned about the liberalization of the energy market and the ensuing potential of small companies selling small and innovative technologies. She is also aware of peaking conventional energy sources and the dependency their usage imposes. She understands that climate change is a serious threat and that both society and the individual have different options by which to address it. Traditionalist: Bill is married and has adult children and perhaps grandchildren. He is close to retirement. He has a secondary education, works as an employee, and has a medium income. He and his wife have always worked hard and cared for the family. They try to secure a reasonable living standard for themselves and their children. They bought a house and saved money in the bank. Bill and his wife are members of local associations and contribute to their community. Bill does not like to spend his money, but if he must, he does so very selectively. He tries to get good quality at a low price; being older he is also interested in comfort. Having a long experience in purchasing, he generally knows what he wants. Bigger investments are only made if they are really necessary. Bill then tries to replace the product he had. He is interested in well established, but also 26
See http://www.sociovision.com/ following Sinus Milieus for further information.
46
3 Private Actor Model
current technologies. Bill knows what he has to pay for his monthly energy bills. He tries to minimize this expenditure by saving energy daily. If Bill has to replace an energy technology he consults manufacturers directly. He is neither willing to spend much of his savings nor to indebt himself. Established: Roger and his wife Paula have known each other since university. They have good educational backgrounds. After finishing university, Roger worked hard, achieved a good reputation, has responsibilities in the local community, benefits from an above average income, and has a secure living standard. After they had children, Paula left her job and stayed at home to raise the children. Roger and Paula spend their income selectively on high quality products. They love good wine and food, meeting friends, enjoying cultural events, and traveling. When purchasing products, they try to improve their level of comfort. When Roger faces a bigger investment, he evaluates it carefully. He and his wife consult experts and discuss the relevant facts intensively. They purchase well established, state-of-the-art technologies. Roger loves winter Sunday afternoons at home. He knows that he has to pay for his well heated house but he enjoys the comfort. Investment in energy technology or energy efficiency upgrades is not a joyful task to him. He has to spend time acquiring knowledge in an unattractive domain. He is interested in decreasing his annual energy bill and relying on clean and easy to operate technology. The social milieus approach enables us to derive some of the required parameters for the decision model. This includes the decision goals and their relative hierarchy (as required), the search domain, the financial, environmental and comfort constraints, as well as the rationality type which can be used in combination. The parameters are given in Table 3.2. Table 3.2. Parameters derived from social milieus
goals (ordered)
constraints
search domain budget debt comfort (1-5) environment rationality types
technology leader environment, cost, comfort topical considerable yes 3 legislation medium, high
traditionalist cost, comfort status_quo limited no 3 – low, medium
established agent comfort, cost, environment peer_group some yes 4 legislation medium
3.4 Modeling Private Energy Investment Decisions
47
Parameters from Rationality Types
The rationality type approach draws from Gigerenzer et al. (1999) and Gigerenzer and Selten (2001). The underlying idea is to categorize decision makers by their ability to select heuristics which perform in a given environment, thus by their ecological rationality. The following rationality types are distinguished. Low rationality: People exhibiting low rationality use a recognition heuristic to search for alternatives. When looking for options, they tend to rely on past decisions and are therefore focused on the status quo. In addition, these people use simple analysis tools to assess and compare alternatives regarding their goals. Medium rationality: People with medium rationality also use a recognition heuristic to search for alternatives. They gather information by asking friends, consulting recommended experts, and reading newspapers. In addition, they use sophisticated analysis tools to assess and compare alternatives regarding their goals. High rationality: Decision makers with high rationality find all alternatives available. In addition, they use advanced analysis tools to assess and compare alternatives regarding their goals. In contrast to the social milieu model, the rationality type model allows for the specification of search rules, analysis tools and decision strategies. The chosen decision rules are shown below in Table 3.3. Table 3.3. Parameters derived from rationality types
analysis tool
low rationality medium rationality high rationality search rule find_next find_common find_all cost investment payback period npv environment qualitative consumer_energy co2 comfort qualitative qualitative qualitative decision strategy SAT LEX LEX SAT: satisficing strategy, LEX: lexicographic strategy, npv: net present value, co2: CO2 emissions Specification of Agents
Various combinations of social milieu and rationality type enable and restrict the different bounded rationality decision models available to each representative agent. Subsequently, six types of building owners are exemplarily distinguished. Five represent building owners who live in the building they own and one represents landlords. It is assumed that traditionalist have low or medium rationality because of their focus on status quo and
48
3 Private Actor Model
their peer group. Established actors were combined with the medium rationality type assuming that they do not carry out advanced analysis but are open to options which might not be found in their status quo. Finally, technology leaders were combined with the medium and high rationality type because of their interest in energy and energy technology purchase. The types and the related parameters are shown in Table 3.4. Matrix Transformation and Selection
The agent decision model constructs an agent-specific decision matrix by searching the general decision matrix for all admissible alternatives. The agent-specific decision matrix thus holds only the subset of options which can be known to the considered agent type. Each of those options is then evaluated with regard to each goal by applying the analysis tool specified. The completed agent-specific decision matrix holds all alternatives available to an agent as well as the values of each alternative regarding each goal. The agent model finally chooses one alternative from the agent-specific decision matrix by applying a decision strategy. The strategy may require that aspiration levels are relaxed and the search and evaluation process is repeated if no feasible alternative is found on the first iteration.
3.4 Modeling Private Energy Investment Decisions
49
constraints
decision strategy
analysis tool
search rule
Table 3.4. Parameters of the building owner decision model as assumed. traditionalist, low rationality
traditionalist, technology leader, medium rational- medium rationality ity
search rule
find_next
find_common
find_common
search domain
–
peer_group, status_quo
topical
market share society –
15%
–
market share location –
7.50%
–
market share peer
–
7.50%
–
search order
status quo
–
–
cost
investment
payback period
payback period
discount rate
–
–
–
time horizon
–
–
–
environment
–
–
consumer energy
comfort
qualitative
qualitative
qualitative
strategy
SAT
LEX
LEX
goal ranking
–
cost, comfort
environment, cost, comfort
JND
–
5%
5%
AL cost
–
10 years
25 years
AL environment
–
–
–
AL comfort
3
3
3
budget
125€/m²
125€/m²
250€/m²
debt
no
no
yes
comfort
3
3
3
environment
–
–
legislation
AL: aspiration level; JND: just noticeable difference; Note that the parameters shown here serve as an example. A real application of the approach would require suitably designed surveys to determine the agent types and associated parameters.
50
3 Private Actor Model
Table 3.4. (cont.)
decision strategy
analysis tool
search rule
technology leader, established agent, real estate manhigh rationality medium rational- agement company ity search rule
find_all
find_common
find_all
search domain
–
peer_group, status_quo
–
market share society –
10%
–
market share location –
5%
–
market share peer
–
5%
–
search order
–
–
–
cost
net present value payback period
net present value
discount rate
3%
–
6%
time horizon
15 years
–
10 years
environment
CO2-emissions
consumer energy CO2-emissions
comfort
qualitative
qualitative
qualitative
strategy
LEX
LEX
LEX
goal ranking
environment, cost, comfort, cost, en- cost, comfort, encomfort vironment vironment
JND
15%
5%
5%
AL cost
–5000€
15 years
0€
AL environment
–
–
–
AL comfort
3
4
4
constraints
budget 250€/m² 250€/m² 500€/m² debt yes yes yes comfort 3 4 3 environment legislation legislation legislation AL: aspiration level; JND: just noticeable difference; Note that the parameters shown here serve as an example. A real application of the approach would require suitably designed surveys to determine the agent types and associated parameters.
3.5 Results
51
3.5 Results 3.5.1 Introductory Remarks A prototype software of the decision model was developed, parameterized and applied. This section discusses some exemplary decision outcomes. At first, the general decision matrix is presented, followed by the single decision outcomes of the model for each type of agent for one representative building. Finally, the aggregated outcome of the model, which was applied to a prototype city, is given. 3.5.2 General Decision Matrix The set of options consists of nine technology options and three efficiency options, as described above. Each efficiency option can be combined with each technology option, which leads to a set of 27 options in the general decision matrix. The general goals of minimum cost, minimum environmental impact, and maximum comfort are calculated using three analysis tools for the first two goals, and one for the last. Table 3.5 presents the completed general decision matrix including the operational costs. Each possible analysis tool was applied. The matrix was generated using the energy system optimization software tool deeco (Bruckner et al. 2003) coupled to a decision modeling software tool. The matrix is based on demand and cost data for a typical single family building built between 1958 and 1968 and located in a quarter where gas and district heating grids are available. The present state supply option is a conventional gas boiler (Gas_BoiConv) with no insulation upgrades realized so far (maintenance). Note that an equivalent matrix can be calculated for each building type having access to different local infrastructures. The general goal of minimum cost is determined using three different analysis tools. The investment costs and operational costs can be understood intuitively by recognizing that additional insulation increases investment costs but saves energy and therefore operational costs. The calculation of the payback period and the net present value is done with reference to the present state of the building. The payback period of the present state is set to the aspiration level of the agent who carries out the analysis. A negative net present value indicates that the regarded option is not profitable in reference to the present state option; a positive value represents the present cash-flow, when selecting that option in relation to the status quo option.
52
3 Private Actor Model
The general goal of minimum environmental impact is assessed using three different analysis tools. The qualitative tool reflects a perceived hierarchy, the consumer energy indicates how much energy will be consumed, and the CO2 emissions reflect the total emissions of an option. Note that the CO2 emissions from wood pellets were set to zero and that the resulting emissions for those options derive from grid electricity consumption. The higher emissions of the Pel_SolarBoiConv options are related to the ancillary electricity required to operate the solar collectors. The general goal of maximum comfort is only expressed qualitatively. Options which do not allow for grid connected supply of the energy carriers have lower values than those which do. An increase of insulation is regarded as an increase in comfort. Further, the cogeneration plant has shorter maintenance intervals and emits noise so that the comfort is lower than with simpler technologies. The decision matrix in Table 3.5 shows that the hierarchy of options changes relative to different goals, but also in regard to different analysis tools. For example, option G (Gas_BoiConv/maintenance) has the lowest investment cost, but A (Gas_BoiCond/maintenance) wins on payback period and P (Hea_Grid/maintenance) has the highest (most attractive) net present value. Regarding environmental impact, option O (Gas_SolarBoiCond/enhanced) outperforms on consumer energy, but X (Pel_BoiConv/enhanced) has the lowest CO2 emissions. A large set of options is perceived to be leading when the analysis is only qualitative. In terms of comfort, most of the options with grid supply and enhanced insulation offer the highest qualitative comfort. The matrix also shows that not only the best option changes when looking at different goals and applying different analysis tools, but the whole hierarchy of options is normally affected. The application of different analysis tools enables us to model different levels of sophistication. For instance, in terms of environmental impact, the consumer energy metric reflects only the amount of energy which is used and therefore favors technologies with a high efficiency and direct renewable input, whereas the use of CO2 emissions additionally recognizes the different specific CO2 content of fuels. A similar effect arises when comparing payback period with net present value. The payback period does not distinguish between present and future payments, whereas the net present value discounts future payments and thereby gives a different picture regarding cash-flow.
3.5 Results
53
Table 3.5. Completed general decision matrix showing all goals and the results of all possible analysis tools. technology
insulation
A Gas_BoiCond maintainance B Gas_BoiCond standard C Gas_BoiCond enhanced D Gas_BoiCondSolar maintainance E Gas_BoiCondSolar standard F Gas_BoiCondSolar enhanced G Gas_BoiConv maintainance H Gas_BoiConv standard I Gas_BoiConv enhanced J Gas_CogenConv maintainance K Gas_CogenConv standard L Gas_CogenConv enhanced M Gas_SolarBoiCond maintainance N Gas_SolarBoiCond standard O Gas_SolarBoiCond enhanced P Hea_Grid maintainance Q Hea_Grid standard R Hea_Grid enhanced S Oil_BoiConv maintainance T Oil_BoiConv standard U Oil_BoiConv enhanced V Pel_BoiConv maintainance W Pel_BoiConv standard X Pel_BoiConv enhanced Y Pel_SolarBoiConv maintainance Z Pel_SolarBoiConv standard A1 Pel_SolarBoiConv enhanced The present state is represented by option G.
investment cost
operational cost
[€] 28,113.90 35,901.60 55,768.80 32,062.10 39,849.80 59,717.00 27,824.30 35,667.80 55,572.10 48,335.70 52,231.70 69,504.40 51,797.00 59,584.70 79,451.90 31,164.70 40,289.80 61,048.60 28,534.80 36,434.00 56,375.40 35,391.30 41,778.50 60,712.00 59,074.40 65,461.70 84,395.10
[€/year] 3,527.20 2,743.80 2,221.40 3,372.50 2,589.00 2,066.70 3,799.90 2,945.10 2,375.20 2,044.50 1,620.80 1,338.40 3,239.20 2,473.20 1,969.20 2,923.70 2,274.70 1,848.90 3,439.40 2,678.90 2,171.90 2,602.40 2,061.00 1,700.00 2,397.60 1,873.90 1,529.30
54
3 Private Actor Model
Table 3.5. (cont.) net present payback value period
environment consumer energy
[€] [year] [ranked] A 2,965.80 1.1 2.0 B 4,531.30 7.6 2.0 C -9,100.30 17.7 2.0 D 864.6 9.9 3.5 E 2,430.30 9.9 3.5 F -11,201.30 18.4 3.5 G 0 10 2.0 H 2,361.80 9.2 2.0 I -10,739.00 19.5 2.0 J 444.5 11.7 3.0 K 1,606.50 11.2 3.0 L -12,294.10 16.9 3.0 M -17,278.90 42.8 5.0 N -15,922.20 23.9 5.0 O -29,772.90 28.2 5.0 P 7,120.40 3.8 3.0 Q 5,743.00 8.2 3.0 R -9,932.60 17.0 3.0 S 3,593.00 2.0 1.0 T 4,772.70 7.7 1.0 U -9,116.00 17.5 1.0 V 6,728.80 6.3 5.0 W 6,805.60 8.0 5.0 X -7,818.50 15.7 5.0 Y -14,509.30 22.3 5.0 Z -14,644.80 19.5 5.0 A1 -29,464.00 24.9 5.0 The present state is represented by option G.
[kWh/year] 51,326.90 38,627.20 30,160.80 48,640.80 35,941.40 27,474.90 55,675.00 41,837.50 32,612.20 59,645.00 44,768.60 34,850.80 46,138.60 33,728.90 25,566.90 50,300.60 37,869.40 29,582.20 55,675.00 41,837.50 32,612.20 57,635.80 43,285.30 33,718.10 51,190.30 37,325.80 28,207.20
CO2 emissions
comfort
[kg/year] 11,438.60 8,880.80 7,175.50 10,931.10 8,373.30 6,668.00 12,327.90 9,537.30 7,676.90 4,813.30 3,989.30 3,440.00 10,491.50 7,990.80 6,345.50 10,667.90 8,311.70 6,740.90 15,854.40 12,140.90 9,665.20 1,805.80 1,768.90 1,744.20 1,938.00 1,900.00 1,874.30
[ranked] 4.0 4.3 4.7 4.0 4.3 4.7 4 4.3 4.7 3.9 4.1 4.4 4.0 4.3 4.7 4.0 4.3 4.7 3.3 3.7 4.0 3.3 3.7 4.0 3.3 3.7 4.0
3.5 Results
55
3.5.3 Single Decision Outcomes The decision outcome for the different agents is determined by the assigned analysis tools and decision rules. Each decision process requires the generation of an agent-specific decision matrix, which can deviate from that shown in Table 3.5 depending on the building characteristics and demand profiles, the selected search rules, aspiration levels, and analysis tools and their related parameters — including discount rate and time horizon. Table 3.6 gives a single decision outcome for a typical single family building, erected between 1958 and 1968, of all agents specified in Sect. 4. Further, it shows whether the aspiration levels (AL) affected the selection and the decisive goal and its rank. Table 3.6. Single decision outcomes agent choice selection process milieu rationality technology insulation AL goal rank maintetraditionalist low Gas_BoiConv no investment 1 nance maintetraditionalist medium Gas_BoiCond no payback 1 nance established medium Gas_BoiCond standard yes payback 2 agent technology consumer medium Gas_BoiCondSolar enhanced yes 1 leader energy technology maintehigh Pel_BoiConv yes npv 2 leader nance real estate mainteHea_Grid no npv 1 manager nance Results were determined by AL: aspiration level; goal: decisive goal; rank: rank of the decisive goal.
A range of different technologies was selected and some agents carried out efficiency upgrades. Keeping in mind that a conventional gas boiler in a non-insulated building (Gas_BoiConv/maintenance) defined the present state, the selection of the traditionalist with low rationality can be easily understood. The present state option was feasible within the aspiration levels and therefore selected. Both the traditionalist and the established agent with medium rationality selected the condensing gas boiler. In contrast to the traditionalist who selected the option with the lowest payback period, the established agent accounted for comfort in the first place, which ruled out low comfort options. The LEX strategy then moved to the second goal and selected the standard insulation option due to its short payback period and high comfort. The highest comfort option had an investment cost above the aspiration level.
56
3 Private Actor Model
The selection of the technology leader differs with regard to the insulation standard and the technology option. While the technology leader with medium rationality selected the solar supported water heating, the agent with high rationality selected the wood pellet boiler. This difference is due to the different tools used to evaluate environmental impact. Due to the investment cost aspiration level the technology leader with medium rationality selected the gas boiler with solar supported water heating and enhanced insulation (Gas_BoiCondSolar/enhanced). In contrast, more than one alternative was within the just noticeable difference of the technology leaders with high rationality with regard to their first important goal, minimum environmental impact. They then moved to the minimum cost goal and selected the pellet boiler with no efficiency upgrade due to the highest net present value. The district heating grid was selected by the real estate managers. They based their decisions on net present value in the first place. The decision rules for the real estate managers assume that the investment costs cannot be passed on to tenants by means of higher rents. In this example, none of the determined decision outcomes would change if the search rules were altered, for instance, if find_all was applied throughout. The choice of search rules becomes significant for this example if fuel and technology prices change over time. Assuming a steady rise in gas and oil prices and further technical and cost improvements for solar collectors and wood pellet boilers, the established and traditional agents might not find these topical, good comfort, low cost options. In other words, they remain late adopters. A change in adoption behavior can be modeled by making options common which are applied by e.g. over 5% of the agents in the regarded district. 3.5.4 Aggregated Decision Outcomes Finally the model was applied to a prototype city consisting of nine prototype single family houses which could have one of the nine supply technologies introduced above in their present state (see Table 3.7). Further, the model distinguished six agent types as described in the previous section. A combination of all prototype buildings, status quo technology options and agent types generated 369 agents, representing 8,991 building owners of the city27. It is assumed that 30% of the population are technology leaders (7.5% with high rationality and 22.5% with medium rational27
Note that it is assumed, that not all agent types have all status quo technology options, e.g. the traditionalist with low rationality does not initially live in a building with a wood pellet boiler. Therefore the agent number is lower than 9•9 •6 = 486.
3.5 Results
57
ity), 40% are established agents, 15% are traditionalists (10% with medium rationality and 5% with low rationality) and 15% are real estate managers. The maximal possible connection rate of buildings to the electricity grid was set to 100%, to the gas grid to 90% and to the district heating grid to 30%. Further, it is assumed that the prices for electricity, heat, oil and gas will increase at a constant rate of 2% per year, wood pellet prices will rise at 1.5% per year. The average lifetime of technologies was set to 15 years, the renovation cycles of the building shell to 25 years. The model was executed over a 25 year period. The results are stated and discussed below. Table 3.7. Distribution of technology options among agent types in the present state technology established traditionalist real estate total populeader agent manager lation Gas_BoiCond 25.00% 23.75% 20.00% 20.00% 23.00% Gas_BoiCondSolar 10.00% 1.00% 5.00% 4.15% Gas_BoiConv 25.00% 23.75% 30.00% 25.00% 25.25% Gas_CogenConv 5.00% 1.00% 5.00% 2.65% Gas_SolarBoiCond 2.00% 1.00% 1.00% Hea_Grid 2.00% 23.75% 10.00% 20.00% 14.60% Oil_BoiConv 25.00% 23.75% 40.00% 25.00% 26.75% Pel_BoiConv 4.00% 1.00% 1.60% Pel_SolarBoiConv 2.00% 1.00% 1.00%
Fig. 3.2. Diffusion curves for technologies as calculated by the simulation.28 Detailed diffusion curves for all agent types are shown in the Appendix. 28
The relatively high variation of some curves between time steps is due to the fact that all buildings of one specific prototype are assumed to need renovation
58
3 Private Actor Model
To understand the resulting diffusion curves given in Fig 3.2, a look at search rules might yield some insight. Basically one can distinguish between agents who can make options common – technology leaders are looking for topical options and high rationality agents e.g. the real estate manager will find all option – and agents who consider only common options – established agents or traditionalists with medium rationality apply the find_common rule referring to their peer group, the society and the location. Therefore, firstly new technologies always start entering the system through the find_all and the find_common (with the search_domain topical) rule and secondly only percolate through the system if they become common and agents’ valuation favors such new technologies. In the chosen example, new technologies such as solar thermal collectors, cogeneration units, and pellet boilers (e.g. Gas_BoiCondSolar, Gas_CogenConv, Pel_BoiConv) are initially only selected by technology leaders. Further diffusion is only possible if agents who only find technologies which have achieved a specific market share (e.g. 5% for the established agents and 7.5% for traditionalists) in a location or among the whole society judge them positive and consequently apply them. The Gas_BoiCondSolar, the Pel_BoiConv, and the Gas_SolarBoiCond options belong to this group. In contrast, the Gas_CogenConv never reaches the commonness threshold and is therefore never evaluated by agents searching with find_common. Having chosen the commonness of technologies as the first rationale to explain the diffusion curves produced by the model, a focus on analysis tools is useful. Agents base their decisions on the evaluation of costs, environmental impact, and comfort of an option. The results from the analysis of the comfort and environmental impact of options differ with respect to the applied analysis tool and the status quo, but do not change over time. Occurrence of the diffusion of renewable technologies such as solar thermal collectors and wood pellet boilers among technology leaders is predominant, because the medium and high rationality types minimize the environmental impact. But simulation outcomes get more complex when energy prices are assumed to rise at different rates over the whole scenario timeframe, increasingly favoring high conversion efficiencies, energy efficiency upgrades, renewable energy input, and low price energy carriers. Further, investment costs for novel technologies are assumed to fall making novel technologies easier29. Technologies benefiting from these developments (such as Gas_BoiCond, Gas_BoiCondSolar, Gas_Solar within the same time interval. Naturally, the renovation in reality would occur over a longer time period and flatten the curves. 29 The investment costs for solar thermal collectors used for room and water heating (Gas_SolarBoiCond) and for pellet boilers (Pel_BoiConv, Pel_SolarBoiConv) were assumed to decrease at 1% per year. The invest-
3.5 Results
59
BoiCond, Gas_CogenConv, Pel_BoiConv, and Pel_SolarBoi Conv) might be chosen by the established agents, the traditionalists or the real estate managers as soon as they become attractive. Further, the results in Fig. 3.2 show two losing options, the conventional oil and gas boiler (Oil_BoiConv and Gas_BoiConv). Initially used by around 50% of the agents both options drop to 3% and 5% respectively over the last time interval. The conventional gas boiler is mostly replaced by a condensing gas boiler and is only kept by some traditional agents with low rationality preferring to maintain the status quo. Likewise, the conventional oil boiler is on its way out. Finally, the Gas_CogenConv option deserves some attention. Over the first five years of the simulation this option is selected by the technology leader with medium rationality just occasionally and therefore takes up slowly. Because of its high cost and its context dependent performance, only some buildings are economically attractive environments for micro-cogeneration. Later solar options are favored because of their lower consumer energy demands and economic attractiveness. As it is assumed that unit production costs decrease and prices for fuels increase, the micro-cogeneration unit becomes an interesting option to the real estate manager towards the end of the simulation and therefore the market share rises again. The diagrams in Fig 3.3 and 3.4 show how the diffusion curves for gas fired condensing boilers (Gas_BoiCond, Fig 3.3) and gas fired condensing boilers with solar thermal water heating (Gas_BoiCondSolar, Fig 3.4) are obtained from the different agent specific diffusion curves. The Gas_BoiCond option spreads at a slowly declining rate into the city, leading to a period of saturation before it starts to fade out towards the end of the simulation. One can easily identify that the real estate managers, the traditionalists and the established agents sustain this diffusion while the technology leader chooses different options. The decline at the end is due to the stepping out of the established agents who increasingly select the Gas_BoiCondSolar option and the real estate manager who perceives the Gas_CogenConv options to be attractive. Turning to Fig 3.4, the technology leader favors the Gas_BoiCondSolar option for the first half of the scenario time frame making it common to the established agents (their threshold for commonness is 5%). But they would not pick it up until it becomes economically attractive as well. In the second half of the simulation time frame the further diffusion of solar thermal water heating is driven by established agents, who overcompensate the stepping out of the technology leaders. Similar trajectories are obtained for the diffusion of ment costs for micro-cogeneration (Gas_CogenConv) decreased at 1.5% per year.
60
3 Private Actor Model
Fig. 3.3. Behind the diffusion curves for gas fired condensing boilers without solar thermal water heating (Gas_BoiCond)
Fig. 3.4. Behind the diffusion curves for gas fired condensing boilers with solar thermal water heating (Gas_BoiCondSolar)
pellet boilers (Pel_BoiConv), which are selected by traditionalists as soon as they are made common by technology leaders and become economically attractive. The established agents never choose pellet boilers because of their low comfort value.
3.5 Results
61
Fig. 3.5. Combination of technology choice and efficiency upgrades in combination with the condensing gas boiler (Gas_BoiCond) option
Fig. 3.6. Combination of technology choice and efficiency upgrades in combination with the condensing gas boiler with the solar thermal water heating (Gas_BoiCondSolar) option
In addition, the combination of efficiency upgrades and technology choice deserves some attention. Fig 3.5 and Fig 3.6 show which efficiency standards were combined with the gas fired condensing boiler and the gas fired condensing boiler plus solar supported water heating. To understand
62
3 Private Actor Model
the trajectories, it is important to note that variations occur if an agent carries out an efficiency upgrade and stays with the regarded technology option or if an agent either changes to or from the regarded technology option. In the first two thirds of the scenario time frame the gas fired condensing boiler option (Gas_BoiCond) diffuses into the system, then starting to fade again. During the whole simulation, agents using the gas fired condensing boiler have carried out efficiency upgrades, mostly standard ones. Nearly half of the Gas_BoiCond option is used together with the simple maintenance option, though. Towards the end of the simulation, agents having applied the enhanced efficiency upgrade tend to change to a different supply technology. The picture changes if the gas fired condensing boiler with solar supported warm water heating is regarded (Gas_BoiCondSolar). Most of the diffusion of this technology option is realized in combination with an enhanced efficiency upgrade simultaneously driving out the simple maintenance option. The differences can be understood when looking at the agents who select the technology options. As shown in Fig 3.4 the diffusion of the gas fired condensing boiler with solar thermal water heating is driven by the established agents and the technology leaders only. While the established agents prefer options which maximize comfort, the technology leaders try to minimize the environmental impact in the first place. Both the comfort and the environmental impact properties increase if energy efficiency upgrades are carried out, therefore building shells are insulated. In contrast, the diffusion of gas fired condensing boilers is driven by all agents. Similar to the diffusion of the Gas_BoiCondSolar option, the established agents and the technology leaders tend to carry out efficiency upgrades. Towards the end of the simulation, both agents start to install technologies other than gas fired condensing boilers in their well insulated building which causes a decrease in the share of energy efficient buildings with gas fired condensing boilers. In contrast, traditionalists and real estate managers value cost in the first place. Both only carry out efficiency upgrades if they are perceived to be cost-effective. Despite some economically attractive upgrades realized, a considerable share of non insulated buildings with Gas_BoiCond options exists throughout the scenario timeframe. Finally, different trajectories for the demand for gas, oil, heat and wood pellets were generated by the simulation and are shown in Fig. 3.7. The overall energy demand of the prototype city keeps falling constantly throughout the scenario time frame. Further a shift from oil to gas and an increasing use of district heating and wood pellets can be observed.
3.5 Results
63
Fig. 3.7. Development of primary energy demand in the prototype city
Fig. 3.8. Energy savings in the prototype city
The decreasing overall demand is mainly due to the fact that building owners carry out efficiency upgrades of the building shell during the simulation30. A smaller share is due to an increase in conversion efficiency and renewable energy input (see Fig. 3.8). Despite the overall decrease of en30
Note that it was assumed that in the initial state of all buildings at the beginning of the simulation no efficiency upgrades were carried out.
64
3 Private Actor Model
ergy demand, the demand for wood pellets and district heating slowly increases. Traditionalists, established agents, and real estate managers increase their share of district heating throughout the scenario time frame. While established agents mostly carry out efficiency upgrades because of their comfort preference, both traditionalists and real estate managers also find the simple maintenance option suitable if it pays. Therefore the efficiency upgrades carried out do not compensate the increasing diffusion of district heating and consequently the heat demand within the district heating grid grows. Wood pellets are burned in relatively expensive pellet boilers which are mostly installed by technology leaders. Because of the high investment costs, most technology leaders cannot afford the enhanced efficiency upgrade as well. Thus, pellet boilers are mostly used with the simple maintenance or the standard insulation option. In consequence, the diffusion of pellet boilers leads to a similar increase of the demand for pellets. The oil demand mostly decreases because oil boilers fade out of the system. Finally, despite the rising share of gas-fired technologies from ca. 30% to ca. 50% during the simulation, the overall gas demand decreases. This occurs firstly because gas-fired technologies primarily diffuse among agents who value either environmentally friendly (technology leaders) or high comfort options (established agents) and therefore prefer to insulate their buildings, secondly because the relatively low investment costs for gas boilers enables agents to realize efficiency upgrades, and finally because of the increasing share of efficient condensing boilers and solar supported water heating.
3.6 Discussion The presented decision model offers a new approach for understanding and estimating energy demand and technology diffusion trajectories within the residential housing sector. The modeling concept relies on three steps: the aggregation of technologies and infrastructures to provide representative options, the aggregation of socio-economic data to yield representative agents, and the development of a set of search, analysis, and selection processes by which the agents can make their investment decisions. The proof of concept application provided demonstrates that a rich set of decision outcomes can result from this form of simulation. Both positive and negative diffusion curves mostly showing an S-shape were obtained. In addition, some curves never took up, indicating a none-attractive option. Further, results can be understood referring to the assumptions made and thus are sensible. Compared to empirical analysis relying on regression models, the decision outcomes of the model seem to be reasonable
3.6 Discussion
65
(Lutzenhiser 1993, Schuler 2000). Further, comparing the results to diffusion curves which were generated using a single agent who applies the weighted adding strategy, showed that the bounded rational decision model has some unique advantages (see the Appendix for the wadd diffusion curves). Firstly, traditional agents soften fade out curves and might still apply technologies which would completely fade out otherwise. Secondly, technology leaders introduce novel technologies referring to two environmental performance measures, their final energy consumption and their CO2 emissions. This enables to model market entries of high efficiency options, perhaps with additional solar input, and options which use fuels with low CO2 emissions. Thirdly, the traditional and the established agents both using the find_common rule are late adopters. Thus, they slow down diffusion rates for novel technologies. Finally, the heterogeneity of agents used for the bounded rational decision model enables us to account for different performance features of the technology and efficiency options in different ways, leading to a rich set of diffusion curves. If this is done by introducing different agents using different weight factors for weighted adding, the problem of specifying the weight factors remains. The bounded rationality approach challenges this by providing a set of search rules, analysis tools, and decision strategies from which different agents can be parameterized using socio-demographic surveys. The combination of sociological and technological dynamics within a single energy system simulation provides some unique benefits. Both aspects are well established in their own right, with the sociological dynamics being based on the social milieu methods. These methods were developed to support applied sociological investigations for direct marketing, product and services design and placement, voter analysis, and related issues. The approach is empirical and can be conducted at a high geographical resolution. Good base data sets exist, although most are proprietary. Further effort would be required to refine and particularize this information for use in energy system investment models.31 More work is also needed to validate the assumed agent types used in this analysis. This may comprise computer assisted telephone interviews, focus group discussions and/or questionnaire surveys in order to better understand how house owners and managers make energy investment decisions. Public interest applications for this model include the assessment of domestic sector policy measures, particularly in terms of effectiveness and robustness. It might further yield insights in how policy measures are applied with regard to different technical and socio-economic structures of cities or single districts. Private applications could include improved in31
See http://www.sociovision.com for additional information on services. Microm GmbH (http://www.microm-online.de/en) offers spatially resolved data.
66
3 Private Actor Model
sight into existing and new energy technology markets, spatially highly resolved infrastructure and utilities planning, and the identification of robust business strategies. From a research perspective, it becomes increasingly evident that the energy decisions of investors and households need to be included in energy system models. The proposed model can be coupled with an energy system model with embedded price discovery and is therefore sensitive towards energy prices.
4 Commercial Actor Model
4.1 Introduction Energy sector deregulation has introduced generation and retail competition among energy firms. Urban areas are an important target market for commercial actors. These firms are heterogeneous. Some only trade energy on wholesale markets and offer supply contracts to local consumers, others generate heat and/or electricity, invest in technologies or operate and develop energy infrastructures as part of their core business. Commercial actors have an adequate knowledge of their markets, develop and adjust strategies and perspectives, and dispose of a sufficient budget. Commercial actors, as defined in this work, include local utilities, gas suppliers, heat suppliers, independent energy producers, and also large commercial energy consumers. The future structure of urban energy systems is likely to depend considerably on the types of business model firms adopt to target urban areas. Some commercial actors might maintain their generation capacities and further develop the existing infrastructure; others might focus on distributed technologies as an entry strategy. Price competition and promotional strategies for contracts or technologies can help to win new customers and additionally increase the overall market size and the utilization of network infrastructure. In contrast, an increase in gas, oil or heat prices might affect the investment decision of building owners toward higher efficiency or renewable energy input and therefore decrease the market size for certain fuels. In this context, a good understanding as to how the competition of heterogeneous energy firms in urban areas will affect the performance of each of them, influence the decisions of clients, and consequently change the structure of the entire system is essential to stakeholders. For instance, the efficiency of competition on wholesale and retail markets depends on the number of firms participating in it; the number of firms itself might considerably increase as distributed technologies become cost-effective, enabling new entrants to operate generation capacities. Further, the overall CO2 emissions of urban areas can be affected by both the number and type of generation capacities operated by commercial actors and the technolo-
68
4 Commercial Actor Model
gies and efficiency standards applied by buildings owners. Bearing in mind that technology selection by building owners partially depends on retail price levels and is therefore indirectly influenced by competition, a complex environment needs to be studied. This chapter proposes a quantitative model for the energy investment decisions of firms which fits in the framework developed in Chap. 2. The remainder is structured as follows: after a review of the relevant literature, the selected concept to aggregate options and firms is presented, followed by a description of the applied decision model. The next section presents some illustrative outcomes from the application of the model. This was combined with the private agent decision model presented in Chap. 3. Finally, some closing remarks are provided.
4.2 Commercial Energy Investment Decisions 4.2.1 Introductory Remarks The modeling concept relates to two strands of literature. The first discusses theoretical concepts explaining how firms achieve and sustain a competitive advantage comprising the market-based (Porter 1980) and resource-based views of a firm (Penrose 1959, Wernerfelt 1984, Barney 1991). The second, more empirical strand, studies how firms, which cope with changes in technologies, regulations and institutional settings, select strategies (Tirole 1988, Levy and Rothenberg 2002, Christensen 2000). This section briefly introduces theoretical concepts and gives an overview using recent case studies. 4.2.2 Theoretical Background Economic theories that attempt to explain how firms develop a competitive strategy and thus build and sustain a competitive advantage can broadly be divided into theories which analyze the opportunities and threats arising from a specific environment on competitive positions and those which focus on internal structures by exploring strengths and weaknesses. The first theory, the so-called market-based view suggests that firms analyze their competitive environment, choose their strategies, and then acquire the resources needed to implement the strategy. This view is built on the assumption that firms within a given industry are identical in terms of the strategically relevant resources they can control and the strategies they can pursue. In addition, if resource heterogeneity developed (e.g.
4.2 Commercial Energy Investment Decisions
69
through mergers or new entrants) it would not be long-lived because the resources that firms require to implement their strategies are highly mobile (Porter 1980). The second theory, the so-called resource-based view, focuses on the link between a firm’s internal characteristics and its performance. Therefore, this view assumes that firms within an industry may be heterogeneous regarding their strategic resources and that those resources may not be perfectly mobile so that heterogeneities can persist. The implications which arise from both assumptions are employed to explain the reason for sustained competitive advantage (Barney 1991). The Market-based View
The market-based view was introduced and formalized by Michael E. Porter. Porter identifies three generic strategies – cost-leadership, differentiation, and segmentation – commonly used in business. A strategy should fit the strategic scope, which refers to the size and composition of the market a firm intends to target, and the strategic strength, which refers to the core competencies of the firm. The marked-based view explains the emergence of a competitive advantage of a firm by the forceful implementation of one of these generic strategies along its value chain. The value chain distinguishes between primary activities (such as inbound logistics, operation and production, outbound logistics, sales and marketing, and service and maintenance) and support activities (such as administrative infrastructure management, human resources management, research and development, and procurement; Porter 1980). A cost-leadership strategy aims at minimizing the production cost, while simultaneously guaranteeing sufficient quality. Products should be almost homogeneous to permit producing high quantities and serving a large number of consumers to benefit from economy of scale and experience curve effects. Cost-leadership strategies may be successfully implemented when the firm possesses substantial market shares and privileged access to raw material, technologies, labor or other important input factors. The differentiation strategy develops products which are perceived as unique by consumers. Those products should offer additional features such as high quality, excellent service, additional utility, or advantage of location. A differentiation strategy will enable firms to place products with a higher price on the market, and to win costumers’ loyalty. Differentiation strategies might be implemented by firms which are innovative, have good research and development skills, good marketing strategies and attract creative and skilled employees. A segmentation strategy focuses on just one or a few target markets. Products are especially designed to meet the needs of consumers in that
70
4 Commercial Actor Model
specific niche. Segmentation strategies are likely to be attractive to small specialized firms trying to gain a competitive advantage through effectiveness rather than efficiency. A typical target market has a weak competition level, is little vulnerable to substitutes, and offers an above average return. Large energy generation companies are likely to follow a costleadership strategy. They operate large generation units and provide standardized products to wholesale energy markets. Further, energy traders targeting the retail market, which is often dominated by the incumbent local utility, might implement a cost-leadership strategy as well by offering low price contracts to private consumers. In contrast, differentiation strategies might be attractive to local utilities, offering a large number of different services to their clients. Being on their doorstep, they are likely to be perceived as a multi-utility company not only engaged in electricity and gas supply but also operating the required infrastructure, and offering a high quality service accompanied by some commitment to the local community. Small companies operating cogeneration units or renewable energy technologies may have implemented a segmentation strategy by focusing on the installation and operation of just one or two small and innovative technologies. Energy traders offering electricity generated from renewable sources to the green niche market may also follow a segmentation strategy. They thereby target a limited and very selective market. Segmentation strategies might also enable firms to enter a market, exploit its potential, and, assuming technology, regulation, or competition changes are favorable, enlarge their activities. The Resource-based View
The resource-based view was initially formulated by Penrose in the late 1950s (Penrose 1959). It was recently revived and further developed by Birger Wernerfelt (1984) and Jay Barney (1991). Following Barney (1991), firms may be portrayed as a bundle of resources which can be classified into three categories: physical, human, and organizational capital resources. Physical capital resources are technologies, infrastructures, plants and equipment. Further, human capital resources are training, judgment, intelligence, relationships, and insights of individual managers and workers. Finally, organizational capital resources are the reporting structure, the formal and informal planning, controlling, and coordination systems of a firm. Barney explains the emergence of sustained competitive advantages among firms by the heterogeneity and immobility of some crucial resources. Those resources must be valuable, rare among firms, and imperfectly imitable. Further, there must not be strategically equivalent substitutes that are valuable and neither rare or imperfectly imitable for these
4.2 Commercial Energy Investment Decisions
71
crucial resources. Firms seeking to maintain and increase their competitive advantage need to develop their resources in a way that future markets can be exploited. A firm has a sustained competitive advantage over its competitors and all other competitors poised to enter the market if this advantage can be sustained even after the efforts to duplicate this advantage have ceased (Rumelt 1984). Adopting this definition, sustained competitive advantages may, on average, persist over a long time. Under some circumstances competitive advantages may not be sustained due to unanticipated changes or structural revolutions within the economic setting of an industry. Valuable, rare, imperfectly imitable, and not easily substitutable resources of energy firms include access to capital markets, knowledge and skills on the part of the management, generation capacities, networks, and the costumer base. For example, the construction of large generation units such as coal or nuclear power stations, and combined cycle power plants requires considerably high investments 32 which can only be supplied by capital markets. Firms targeting such investments need to group the required engineering and management skills and benefit from the trust of the capital markets. Further, the deregulation of energy markets has enabled private consumers to choose their energy suppliers. Despite increasing competition within these markets, few consumers have exercised this choice33. Therefore, the existence of a large costumer base may be interpreted as a crucial resource. 4.2.3 Empirical Evidence Empirical work has shown that management boards facing new technology options, ongoing or expected changes in regulations, or new economic, social, or environmental threats have problems to design sophisticated longterm strategies. In fact, companies temporize, imitate past successful actions, and align their strategic consideration with their perceived consumer expectations, company culture, and personal perspectives regarding the future development of the world.
32
Gas turbines require around 250€/kW, combined cycle power plants 500€/kW and coal fired power plants 1,200€/kW investment (Heuck et al., 1999, p. 616). 33 In the UK nearly one quarter of private consumers have changed their energy supplier since the deregulation. In contrast, the liberalization of the German market has encouraged only 3% of the consumers to change their supplier. A common strategy to increase the market shares is extensive marketing. In addition, some energy firms have chosen to buy local retail companies to gain direct access to retail markets.
72
4 Commercial Actor Model
Geroski (2000) surveyed the technology diffusion literature and identified a couple of factors influencing the diffusion of new technologies. Firm size, for example, may determine the daily runtime of a technology, thereby affecting its potential to reduce staff and influencing the cost of capital. The provision of information and the marketing of a new technology by suppliers also have an impact on diffusion rates. Further, technology expectations like current or near-future improvements in either the old or the new technology are likely to inhibit diffusion. In addition, learning and switching costs might deter firms from adopting. Finally, opportunity costs, e.g. created by previous investment in machinery not fully depreciated, are also of importance. Further, two effects – the pre-emption effect and the rent displacement effect – leading to sequential adoption have been identified by economists (Oster 1982, Tirole 1988, Stoneman and Kwon 1994). The pre-emption effect can occur if a technology complements the existing activities of a firm more than those of another. This gives an incentive to that firm to adopt earlier than its rivals. For example, local gas suppliers might view microcogeneration units as a technology to increase the local gas market size and foster their diffusion. In contrast, rent displacement occurs if a technology interferes with other activities of the firm, making adoption more costly than it would be in the absence of such activities. Energy generators who simultaneously supply commodity markets with fossil fuels such as natural gas or hard coal might resist renewable technologies for fear of decreasing the demand for those commodities. Likewise, network operators might prefer central technologies in order to secure a high utilization of their infrastructure. Levy and Rothenberg (2002) used a case study approach to investigate the strategic responses of two major American automobile manufacturers, GM and Ford, and the European companies Daimler-Chrysler and Volkswagen toward climate change. Among their major finding was that “each firm’s history influenced the degree to which future technological options were viewed as an opportunity or a threat”34. Company strategies differed with regard to the development of fuel-efficient motors, the reduction of car weights, research and development investments in the search for new engines such as fuel cells and gas motors, and pollution controls for pollutants such as NOx and SO2. Another example of adaptation behavior of firms facing fundamental breaks in technology is given by Christensen (2000). He draws a distinction between so-called sustaining and disruptive technologies and gives the competition between electric utilities and firms operating distributed tech34
See Levy and Rothenberg (2002, p. 188).
4.3 Aggregation of the Firm
73
nologies as an example. He argues that some firms resist disruptive technologies for fear of causing too much disruption for their customers, or fail to adopt them because those technologies are not well-matched to their consumer’s current needs. The initial low interest in solar thermal collectors by well-established boiler producers can be interpreted in this way. Further, the emergence of independent producers operating small cogeneration units close to consumers with large heat demands can to some extent be attributed to the immobility of local utilities.
4.3 Aggregation of the Firm 4.3.1 Introductory Remarks Firms offering energy services in urban distribution systems are heterogeneous. They posses different resources, they operate different technologies, their perspectives regarding future developments differ and they pursue different strategies to sustain or build a competitive advantage. Further, firms operate under a large degree of uncertainty regarding technological change, new entrants, and future changes in regulations and climate protection strategies. This section presents an approach to aggregate energy firms so that a quantitative model of investment decisions can be built that fits within the framework of this work. The approach is based on ideas and concepts from the resource-based view as well as the market-based view and relies on empirical findings. The heterogeneity of resources and perspectives and the consequential heterogeneity of actions are modeled by aggregating possible actions to consistent options, dividing the firm into business units, attributing different options to one or more of those units, assigning different strategies to each unit, and introducing perspectives. This approach enables us to capture the effects of different internal characteristics of firms on their performance and the evolution of the overall system. In the following section, the aggregation of options, business units, and the modeling of strategies and perspectives is explained. 4.3.2 Aggregation of Options A wide range of actions can potentially be realized by commercial actors in energy distribution systems. New central and dispersed generation capacities can be added, the existing infrastructure might be extended or dismantled, or plants may be decommissioned. Further, new supply offers
74
4 Commercial Actor Model
can be introduced, or existing ones can be altered to affect the business relations between actors. Table 4.1. Possible set of options O Option Market/Network (Gas, Heat, Electricity, Oil and Pellets)
Description Altering energy carrier prices and simultaneously promoting that energy carrier in order to increase the market size, market share, and/or the utilization of infrastructure.a Advertising technology X in order to increase the Advertising Technology X demand for that technology and thereby increase or reduce the demand for a specific fuelb. Infrastructure Expansion Investment in energy infrastructure in order to reach new consumers. Power Station Investment Investment in central power stations in order to sell electricity on markets. Cogeneration Islands Construction of cogeneration units in order to supply neighborhoods with heat and to sell electricity on markets. Contracting MicroOperation of small cogeneration units in residenCogeneration tial buildings in order to supply heat and to sell electricity. Wind Power / PV Investment in wind energy or PV in order to sell electricity. Virtual Power Station Coordinated operation of distributed technologies in one control domain. Upgrading in Analysis and Ac- Switching from typical day load profiles to syncounting Tools thetic year load profiles for the analysis of options; switching from aggregated energy flow to single technology accounting. a
Infrastructure may be an existing gas, heat or electricity network, it may also refer to the utilization of a fleet or other fixed cost based investment. b The promotion of gas-fired condensing boilers is likely to increase the gas demand of condensing boilers in the heat market. Further, assuming that the increase in demand for gas due to fuel switching outweighs the decrease due to efficiency gains, the size of the gas market is increased. In contrast, the promotion and joint diffusion of solar thermal collectors is likely to reduce the demand for fossil fuels such as gas and oil.
A sophisticated aggregation of actions is deemed to construct a limited set of consistent options for energy firms competing in urban areas. Each option groups a number of actions which are likely to be jointly carried out by a firm which pursues a specific strategy. Table 4.1 introduces possible options which duly form the set of options O.
4.3 Aggregation of the Firm
75
Each of these options is associated with a technology, a network extension, a change in unit commitment protocol, a price differential for a contract offer, or an advertising campaign. Further, the realization of one option requires a fixed amount of investment. The concept of options allows to distinguish and parameterize a set of actions a firm can carry out to change the technical energy system or the interactions of actors on retail markets and thereby improve its financial performance. 4.3.3 Aggregation of Business Units The aggregation of business units focuses on the firm itself. Energy firms typically consist of different divisions which have their own accounting and reporting structure and pursue different tasks. An aggregation of the firm into business units will assist the modeling of the heterogeneity of resources and perspectives. For instance, the set of business units BU = {Central_Generation, Distributed_Generation, Renewable_Generation, Networks, Clients_and_Markets} is considered and explained below: The Central_Generation business unit includes the investment and decommissioning of large power stations, their operation and maintenance, as well as the required infrastructure such as fuel supply, waste discharge and product delivery. The central generation business unit includes all power stations with a capacity well above 10MWel (Reisinger 2002) whereas distributed generation has capacities down to some kWel. Usually, this business unit also includes cogeneration units or boilers which are designed to supply a large district heating grid, gas turbines, and pressurized air storages. Technologies belonging to the central generation business unit require large investments and are designed to operate for time periods well above 20 years. The Distributed_Generation business unit contains all generation technologies not covered by the central generation business unit, including investment and dismantling, fuel supply, operation, and product delivery. Examples are small cogeneration units with capacities up to 10MWel, micro-cogeneration, heat pumps, and fuels cells. Usually, the operation of those technologies neither requires a fully developed high-level infrastructure35 nor does it affect the operation of the remaining energy system to any degree. Dispersed technologies are usually operated close to
35
The connection and failure protection of distributed technologies toward networks for supply (e.g. gas grid) and delivery (e.g. electricity grid) is simple in contrast to the connection and failure protection of large generation units.
76
4 Commercial Actor Model
demand, require a limited amount of investment, and have typical life times of up to 20 years. The Renewable_Generation business unit especially focuses on decentralized, renewable technologies such as wind generators, solar cells, biogas engines, etc. It comprises the investment and dismantling, the operation, and product delivery of such technologies. In general, the operation of dispersed, renewable technologies neither requires a fully developed high-level infrastructure nor does it affect the operation of the remaining energy system to any degree. The required capital for construction is limited and the average life time is around 20 years. The Network business units can be further distinguished into gas, electricity and district heating grid business units. A network business unit covers a single network infrastructure. Network extension and dismantling as well as an increase in utilization are strategic targets for business units of this type. The Clients_and_Markets business unit covers all contracts for energy that a firm has concluded and its potential future consumers. Contracts may be issued for the delivery of electricity, gas, heat, coal, wood pellets, etc. A firm may reduce the price for supply contracts in order to increase the number of clients, or it may raise the price to increase the profit. Further, it comprises the supply chain of the company and how purchases are managed. A firm f is composed of at least one of the business units BU introduced above, so
BU f ⊆ BU .
(4.1)
Business units should be distinguished in such a way that the firms to be modeled differ with regard to the number of business units they operate and the actual resource endowment of each unit. Further, the distinction of business units enables us to introduce a subset Ofb of the option set O which groups all options belonging to the business unit b of a firm f, so
O fb ⊆ O | b ∈ BU f .
(4.2)
Table 4.2 shows how options may be assigned to one or more business units introduced above. Firms who do not operate a specific business unit cannot realize the options unique to it. Finally, the distinction enables us to assign different strategies to each business unit; options which occur in more than one business unit will be evaluated regarding different strategies and can be realized for both.
4.3 Aggregation of the Firm
77
Table 4.2. Assignment of options to business units Central Generation Market/Network Promotion of Technology Xb Infrastructure Expansion Central Power Station Investment Cogeneration Islands Contracting MicroCogeneration Wind Power / PV Virtual Power Station Upgrade of Tools
X
Distrib- Renew- Neta uted Gen- able Gen- works eration eration X X X
Clients and Markets X X
X X X X X
X X X
a
The network business unit can be further differentiated into the electrical grid, the gas grid and the district heating grid. Assigned options should target the specific grid. b This option can be parameterized for each technology which might be promoted by a firm in order to increase network utilization or the number of contracts made with customers.
4.3.4 Definition of Strategies and Perspectives As stated earlier, firms are assumed to be heterogeneous with regard to their current resources and have different perspectives on tomorrow’s requirements to enable them to compete. For example, a well-established regional energy supplier might expect the wholesale market energy prices not to rise substantially and the purchase costs for distributed generation technologies to stay high. Therefore, the supplier might concentrate on the central generation capacities and try to compete by offering good service and low prices. In contrast, a regional gas supplier without generation capacities might view distributed technology options as an opportunity to become an electricity supplier and simultaneously extend the market size of the local gas market. This gas supplier might therefore invest in distributed generation early despite higher investment costs and lower initial profits. The model accounts for these differences in perspectives and consequential actions by introducing a strategy for each business unit of a firm. A strategy defines which share cs of the firm’s total capital CP available for investments in a time interval t may be invested in that specific business unit. Further, two types of capital, research and development capital CPR&D and conventional capital CPconventional, both with different interest
78
4 Commercial Actor Model
rates i and time horizons T are distinguished. A single strategy sˆ fb of firm f for the business unit b consists of six values:
sˆ fb
⎛ cs conventional ⎞ ⎟ ⎜ ⎜ iconventional ⎟ ⎟ ⎜T = ⎜ conventional ⎟ ⎜ cs R & D ⎟ ⎟ ⎜ ⎜ i R&D ⎟ ⎟ ⎜ T R& D ⎠ ⎝
(4.3)
The set of strategies for firm f for all business units is
S f = {sˆ fb | b ∈ BU f } .
(4.4)
Further, a firm has expectations regarding future energy prices on the wholesale market. A set of energy carriers E = {electricity, gas, coal, oil, wood_pellets} is introduced. Firm f holds a time series pfe(t) for the expected prices of each energy carrier e ∈ E covering each decision interval t of the scenario timeframe. Finally, the firm makes assumptions about the response of clients to promotions schemes, contract price changes and network extensions. Therefore, the sensitivity Įfc reflecting the expected responses toward advertisements for a certain contract c, ȕfc reflecting the expected sensitivity toward price changes for contract c and Ȗfc reflecting the expected variability towards network extensions enlarging the market for contract c are introduced. The initial set of business units BUf and their present state resource endowment, the parameter of the strategy of each business unit, the price time-series, and the sensitivities may be obtained from the firms themselves. A structured interview with the relevant heads of department and the strategic management division should help to reveal the different perspectives of a firm with regard to different business units. Minimum requirements for interest rates, research and development engagements, maximal time horizons, and the available capital for investments per year can be sought. In a second step, all interviews should be evaluated and interpreted jointly to derive a consistent set of strategies for each company, which adequately reflects the differences among firms. Further, data from the relevant business associations or financial institutions etc. can be included. Further, some parameters might be estimated on the basis of expert judgments.
4.4 Decisions Model
79
4.4 Decisions Model 4.4.1 Basic Concepts Commercial agent decisions are divided into operational, low-stake structural and high-stake structural decisions. Operational decisions determine the operation of each technology in the control domain under the control of an agent. They are simulated in the operational timeframe for an entire year with a time resolution of one hour. In contrast, structural decisions change the structure of a control domain or change prices or promotion schemes and are carried out within the structural timeframe. Energy firms can carry out a range of different structural decisions ranging from contract price selection, the installation of small generation facilities and incremental network extensions to investments in large generation facilities and the development of new infrastructure. All decision options are specified in the option set O. Decisions differ with regard to the required investment, the frequency of occurrence, the firm’s business units BUf involved and the expected impact they might have on the future performance of the firm. Therefore, a distinction is drawn between high-stake structural decisions, which involve a large degree of strategic and political considerations of the management board, and low-stake structural decisions, which are regularly carried out by the middle management, and do not alter the asset profile of a firm to any degree. High-stake and low-stake structural decisions are addressed by two different modeling approaches. Low-stake decisions are modeled endogenously using heuristics based on real-world observations and a rational choice approach. High-stake decisions will be addressed exogenously either by treating each decision option as a new scenario or by including human experts in the run-time decision loop. Each of the introduced options can be assigned to either the low-stake or the high-stake decision model. Low stake-decisions involve investments in small plants like cogeneration units, boilers and other dispersed technologies, small infrastructure extensions, plant decommissioning and contract price fixing. High-stake decisions involve investments in large power stations, and the development of a new infrastructure. 4.4.2 Operational Decisions The technology dispatch within the control domains of commercial agents is determined by a unit commitment protocol. Each control domain is treated independently; its operation is constrained by each component’s capacity, the energy demand within the control domain and the supply ob-
80
4 Commercial Actor Model
ligations for the upstream control domains. Different unit commitment protocols can be applied to simulate the operation of the energy system in a control domain, ranging from simple heuristics to more sophisticated rules, like merit order or the application of optimization tools. This work uses the dynamic energy, emission and cost optimization model deeco36 (Bruckner et. al. 2003) to operate control domains37. Energy demand profiles are specified by time-series of a one hour resolution and provided by demand processes or from agents with whom supply contracts were made. Further, scenario energy prices pe(t) for each energy carrier e are provided by time-series for each interval t. The cost minimal share of each component of a control domain which meets the energy demand is determined by a linear optimization solver taking ambient conditions and energy carrier intensities into account. Deeco is executed once for each control domain in each operational timeframe. Simulation results are included in an accounting procedure and enable agents to evaluate the performance of their technologies, to inform future structural decisions and to adjust perspectives. 4.4.3 Low-stake Structural Decisions Low-stake structural decisions are modeled endogenously for the entire scenario timeframe. The operation of the energy system and the selection process are simulated, alternating the operational timeframe and the structural timeframe. After each operational timeframe, all options Ofb in a business unit b of a firm f are evaluated regarding the strategy vector sˆ fb and, if profitable, realized. Commercial agents can choose among options once in every structural timeframe. Therefore, each single option in a business unit b of firm f and their meaningful combination is analyzed with regard to the strategy vector sˆ fb. The profitability of options is calculated using the net present value npv =
T
t =1
36
Ct
∑ (1 + i)
t
−I
(4.5)
The software tool deeco was developed as a PhD project at the University of Würzburg in Germany by Thomas Bruckner (Bruckner 1997). It serves as an analysis tool for different research project and provides decision support to energy companies. Presently, the software is maintained and developed at the Technical University of Berlin, Germany. 37 Other energy system models having similar features to deeco could be used as well.
4.4 Decisions Model
81
where I is the initial investment required, Ct the expected cash-flows in each time interval t, i is the interest rate, and T the allocated economic lifetime of the investment. Ct is determined applying the simulation tool deeco and assuming that the prices will develop according to the expected price trajectory pfe(t) for energy carrier e and firm f. T and i depend on the strategy sˆ fb and if conventional or research and development capital is used. The npv of each option is calculated twice, first using the conventional capital parameters and second applying the research and development capital parameters. Finally, the option with the highest positive conventional npv, which requires a lower investment than the available conventional capital in the respective business unit, is selected and removed from the set of available options for the current time interval. If the selected option is a combination of single options, all single options included in this combination are removed from the set of available options as well. Further, all other combinations of options including at least one single option which was also included in the selected option are removed. The selection process is repeated until all conventional capital available for the respective business unit in the current time interval is spent or no more options with positive npv are available. The selection process is repeated starting with the remaining options with the highest positive npv calculated using the research and development parameters. Any remaining capital is transferred to the next time interval. If options promote a certain technology in order to increase the demand for a specific fuel and network expansion options or options which change contract prices are evaluated, the calculation of Ct includes an estimate of the response of customers. The influence that such an option has on the revenue Ctc from a specific supply contract c in a time interval t is calculated according to
(
)
Ctc = nc + α fc I c + β fc ∆p c + γ fc Ac p c
(4.6)
where nc is the number of customers who actually hold a contract c. Įfc, ȕfc, and Ȗfc are elasticities: Įfc is the number of clients who are expected to respond if one monetary unit is invested in the promotion of contract c or a related technology, and Ic is the capital actually invested in promotion. ȕfc is the assumed sensitivity of customer response if contract prices are changed according to ǻpc. Finally Ȗfc is the assumed number of people who will decide to make a contract c if the network offers additional access and Ac is the actual number of additional accesses provided. For instance, if a new district heating grid is built in a street, enabling Ac additional building to be connected, Ȗfc is the expected share of buildings which will actually connect.
82
4 Commercial Actor Model
As the expected development of energy prices of each firm may deter from the overall scenario energy price development and the expected response of consumers to advertisement, contract price changes and network expansion may differ from those generated by the private agent model, firms should be enabled to adapt perspectives. Therefore, an adaptation phase could be added, which may be started after a fixed number of consecutive operational and structural timeframes. A description of how this may be included in the model is given in the Appendix. 4.4.4 High-stake Structural Decisions High-stake structural decisions are not modeled endogenously but supplied exogenously instead. High-stake structural decisions are defined as decisions which alter the asset profile of a firm considerably. Those decisions are rarely carried out in urban energy systems and involve a high degree of strategic and political consideration by the management. The investments in central generation capacity, the development of new infrastructure, and also the decision to enter a new market belong to this decision type. Therefore, all options which are in the Central_Generation business unit are regarded as high-stake decisions. Options from other business units may also be included. Nonetheless, the impact of high-stake structural decisions can be addressed with the decision model as well. Two different approaches are considered: Firstly, different scenarios can be built to test each single and each combination of high-stake decision options. This approach is suitable if only a small number of such options are to be tested and data and computing time needs can be contained. Secondly, it is possible to include human experts in the decision loop. Experts can interpret different business indicators in each time interval and may decide if they want to implement a high-stake decision option.
4.5 Application and Results The modeling concept, as described above, was realized in a prototype computer software model, which was combined with the private agent decision model presented in Chap. 3. The infrastructure, the present state data and the private agent types of the chapter were used. Further, two firms were parameterized and may interact with the building owners. Interaction between firm models and private agent models occurred by supply contract offers and advertisements aiming to increase the commonness of offers or connected technologies. The adaptation phase described in the
4.5 Application and Results
83
appendix was not implemented in the prototype model. As a consequence, firms could not recognize the feedback from private agents by counting the number of contracts they had made, calculating the amount of energy which was delivered to or supplied by private agents, and by updating their financial performance. Instead, different scenarios with fixed trajectories for the contract price development of each firm and their advertising campaigns were used. Nonetheless, the commercial agent model can be used to determine the cash-flow from each contract to evaluate the performance of business units. Firm I is the incumbent utility, which operates a large cogeneration plant to supply a district heating grid, offers heat and electricity supply contracts to consumers and operates the relevant distribution infrastructure. In its initial state, the firm has already built both the cogeneration plant and the district heating grid. Further, it supplies all local consumers with electricity and 16% of the consumers with heat over the district heating grid. Therefore, it is the most visible commercial actor in the urban market. The strategy of Firm I is to maintain its dominant position in the local market and to increase the utilization of the district heating grid. Firm I consists of the following business units: BUI = {Central_Generation, Clients_and_Markets, District_Heating_Grid}. Firm II is a new entrant, relying on a markedly different business model. It promotes micro-cogeneration among building owners. Such small technologies, which simultaneously generate heat and electricity, are used in buildings to supply their heat and electricity demand. Primarily driven by the heat demand, micro-cogeneration units typically generate more electricity than is consumed inside the building. Thus, surplus electricity can be sold on the market. Firm II offers to buy this electricity at a premium price, if it can control the operation of the micro-cogeneration plant. Pursuing this strategy, Firm II makes micro-cogeneration more common among building owners using a promotion campaign and contributes to its costeffectiveness by offering to buy the surplus electricity. Furthermore, Firm II offers electricity supply contracts to local consumers and participates in the wholesale electricity market to balance supply and demand. Firm II’s long-term strategy is to build a virtual power station consisting of several micro-cogeneration units. Having a low capital structure, it encourages building owners to undertake the required investments and focuses on promotion, operational control and trading. It is assumed that the strategy of Firm II benefits from a government support scheme targeting microcogeneration38 and the fact that for electricity which is bought and sold
38
Such a support scheme might be similar to the Cogeneration Act in Germany, which rewards the feed-in of 1kWh of electricity generated by micro-
84
4 Commercial Actor Model
within the same low voltage distribution network no high voltage grid charges need to be paid. Firm II consists of the following unit: BUII = {Clients_and_Markets}. Generally speaking, a centralized concept and a decentralized concept compete in an urban area. Firm I focuses on central generation, operating a large cogeneration plant, supplying a district heating grid and selling both heat and electricity to consumers. This business model relies on a sophisticated electricity and heat infrastructure and depends on a high utilization rate of both grids. In contrast, Firm II’s business model introduces the distributed generation paradigm. Generation investment is carried out by a range of private agents, not just by a small number of firms. Each private generator supplies her demand for heat and electricity first; only additional electricity is imported from or exported to the gird. The commercial target of Firm II is to balance the demand and supply of a large number of those private generators and thereby to create added values from distributed generation. The prototype model of commercial agents was coupled with the private agent decision model described in Chap. 3, and executed. Private agents could choose a supply technology and among the offers which were supplied by commercial agents. Private agent decisions were dependent on the agent type, costs and commonness of technologies, and offers available on the market. The model yields two results. Firstly, the technology endowment of private agents changes over time. These changes should be compared with the results given in Chap. 3, where no competition of firms was included, therefore only a single contract offer was available and no advertising was carried out. Secondly, the simulation calculated cash-flows of each firm from the contracts they made with their clients and energy trades on markets. Both results are presented and discussed next. 4.5.1 Impact on the Decisions of Private Agents To understand the impact of the competition between Firm I and Firm II on the technology diffusion among building owners it is important to note that Firm II additionally introduces two types of contracts which could be selected by building owners. The first one offers to buy surplus electricity at 0.07€/kWh from micro-cogeneration units which are installed by building owners. The purchase price was assumed to increase by 1% per year. Further, this contract was promoted among building owners. Two scenarios with small and with large advertising campaigns of micro-cogeneration cogeneration units with a capacity lower than 50kW with 0.0511€ per kWh, allowing generators to additionally sell the electricity.
4.5 Application and Results
85
by Firm II were simulated. The small campaign made the microcogeneration technology common to 15%, the large advertising to 25% randomly selected building owners each year. Fig. 4.1 shows the resulting diffusion curves for the small advertising campaign, Fig 4.2 presents the results for the large advertising campaign. The diffusion curve of microcogeneration (Gas_CogenConv) is indicated by the black broken line.
Fig. 4.1. Diffusion curves for supply technologies assuming 15% of consumers are reached by advertisements39
Fig. 4.2. Diffusion curves for supply technologies assuming that 25% of consumers are reached by advertisement 39
It is assumed that Firm II offers to buy surplus energy from micro-cogeneration for 0.07€/kWh. Price offers increase by 1% each year.
86
4 Commercial Actor Model
The diffusion paths shown in Fig. 4.1 (small promotion campaign) do not differ much from the results given in Chap. 3 where no competition was included. A diffusion of gas-fired condensing boilers, solar water heating, and pellet boilers occurs, but the market share of microcogeneration does not increase. Keeping all other parameters constant, but launching the large advertising campaign by Firm II yields a different picture. If micro-cogeneration is advertised to 25% of the building owners, who are randomly selected in each decision interval, the market share of the technology increases. To understand the difference, one must recall that some building owners apply the find_common search rule, which only finds those technologies which have a market share above their commonness threshold (the commonness threshold is 5% for established agents and 10% for traditional agents, respectively). An advertising campaign makes a specific technology common to randomly selected building owners regardless of its actual market share. Thus, only the large advertising campaign made micro-cogeneration common to a sufficient number of building owners who apply the find_common search rule and finally select micro-cogeneration to raise the market share considerably. Fig. 4.3 and Fig 4.4 show how the different agent types contribute to the diffusion of micro-cogeneration. Initially, the technology is installed by real estate managers, technology leaders and established agents. Later, technology leaders and real estate managers, however, fade out of microcogeneration over the first 20 years in both scenarios. Technology leaders focus on supply technologies with renewable energy input and real estate managers have high requirements for the cost-effectiveness. Only the established agents, who have been informed by the advertising campaign, select micro-cogeneration if it is attractive in terms of comfort, cost, and environment, and then foster diffusion. Despite the advertising, established agents do not select micro-cogeneration until 2013 because of its high costs and comparatively low comfort. As energy prices rise and investment costs fall, after 2013 at least some established agents who know about micro-cogeneration select it. The small advertising campaign does basically not reach enough established agents who favor micro-cogeneration to raise the market share above 5%. In contrast, a market share above the commonness threshold is reached when the large advertising campaign is launched. This enables all established agents to find micro-cogeneration and therefore its diffusion rate takes up.
4.5 Application and Results
87
Fig. 4.3. Agent-specific market shares of micro-cogeneration with small advertising campaign
Fig. 4.4. Agent-specific market shares of micro-cogeneration with large advertising campaign
The example shows how the commonness of technologies and advertising campaigns can be included in the model. A technology which does not have the cost, environment, and comfort properties which are valued by agent types using the find_all search rule can only spread if it is common to agent types who would favor it, either by advertising or by a given market share above the commonness threshold of the respective agent
88
4 Commercial Actor Model
type. In this example, after 2013 some established agents would select micro-cogeneration if they knew about it. Firm II’s offer to purchase surplus electricity from micro-cogeneration at a price above the wholesale market prices and the accompanying advertising campaign make the diffusion possible. The next section discusses whether Firm II is able to benefit from this investment and how this affects the success of Firm I. 4.5.2 Impact of Competition The strategy of Firm II might lead to a diffusion of micro-cogeneration into the energy system. In the second scenario (large advertising campaign), Firm II is able to make a considerable number of purchase contracts for surplus electricity from distributed generation at an above wholesale market price and thereby stimulate the diffusion. Simultaneously, it offers supply contracts for electricity to local consumers, trying to sell the electricity from distributed generation within the urban area. This business model enables Firm II to save high voltage transmission charges on all electricity which is generated and consumed locally within the same trading interval. As the overall electricity demand by local consumers is assumed to stay at around 25GWh per year throughout the whole simulation, both the increase in distributed generation and the retail market entry of Firm II reduce the overall market size and the market share held by Firm I.
Fig. 4.5. Electricity generation, purchase and sales of Firm I in the second scenario
4.5 Application and Results
89
Fig. 4.6. Electricity generation, purchase and sales of Firm II in the second scenario
Above, the electricity generation, purchase and sales trajectories of Firm I (Fig 4.5) and Firm II (Fig. 4.6) in the second scenario (large promotion scheme) are shown. The broken grey line indicates the amount of electricity each firm sells to local consumers on the retail market. Throughout the scenario, Firm I loses customers, and subsequently the electricity demand decreases by around 3.8GWh (15%) during the scenario. In contrast, Firm II makes electricity supply contracts with local consumers and thereby increases the demand by 1.9GWh (8%) in the same time period. Thus, Firm I loses market shares to Firm II and additionally suffers from reduction of the market size by 1.9GWh, due to the distributed generation and self-sufficiency of building owners. To supply demand, Firm I operates a central cogeneration plant and trades electricity on the wholesale market. The broken black line in Fig. 4.5 indicates that the electricity supplied by central cogeneration decreases over the first five years, and then starts to increase again. The model assumes that the share of heat and electricity generated by the cogeneration plant is constant. Therefore, a given heat demand in the district heating grid determines its electricity output. The market share of district heating as shown in Fig. 4.2 is nearly constant for the first five years, and then starts to increase slowly. As the total demand in the district heating grid not only depends on the number of connected buildings, but also on their insulation standards, the total demand in the grid initially decreases, because building owners undertake efficiency upgrades. Thus, the heat and subsequently the electricity output of the cogeneration plant decreases until 2010, followed by constant increase which is triggered by a positive diffusion rate of district heating in the residential sector. Facing a constant elec-
90
4 Commercial Actor Model
tricity demand reduction on the local retail market and an increase in the heat demand in the district heating grid, Firm I changes from a buyer to a seller of electricity on the wholesale market (black and grey lines, Fig. 4.5). Firm II is able to balance both the supply obligations (broken grey line, Fig. 4.6) and the feed-in from distributed generation (broken black line, Fig 4.6) over the first 20 years. Only after 2025, the increasing diffusion rate of micro-cogeneration cannot be accompanied by an equal increase in supply contracts, with local clients leading to a significant export of electricity to the wholesale market (grey line, Fig 4.6) by Firm II.
Fig. 4.7. Economic performance of Firm I in the second scenario
Fig. 4.8. Economic performance of Firm II in the second scenario
4.5 Application and Results
91
Finally, the model results can help to estimate the impact of microcogeneration and the increasing demand for district heating on the economic performance of both firms. Fig. 4.7 and Fig. 4.8 show how expenditures, revenues and the total cash-flow of each firm will develop40. Expenditures are payments of a firm in order to purchase an energy carrier from a market. Revenues are the net incomes from supply contracts. Finally, the total cash-flow is revenue minus expenditures. A firm might use its total cash-flows to operate its business and create a profit. The total cash-flow of Firm I (black line with dots, Fig 4.7) constantly decreases over the scenario. In contrast, Firm II can only show a positive cash-flow after 2018 (black line with dots, Fig 4.8). The revenues of Firm I from heat supply contracts increase after a short reduction over the first five years because of the increasing number of connections and an increasing price (2% per year) for district heating supply contracts. In addition, the revenues from electricity supply contracts increase because the demand reduction is overcompensated by a price increase of 1.75% per year. Incomes from sales on the electricity wholesale market increase, firstly, because Firm I becomes an exporter of electricity, and secondly, because wholesale market prices are assumed to increase at 1% per year. Finally, expenditures of Firm I are carried out in order to purchase natural gas to operate the cogeneration plant. The decrease in total cash-flow is mainly due to the increasing sales of electricity on the wholesale market. Gas-fired central cogeneration units have relatively high variable costs. Further, high voltage grid transmission charges do not have to be paid for electricity which is generated and consumed within an urban area within the same trading interval. Thus, the margin for electricity generated and sold locally is higher than that for electricity generated locally and sold on the wholesale market. To realize a constant increase in cashflow, Firm I could realize different options. Firstly, the demand of the district heating grid could be supplied by the central cogeneration plant, and additionally by a simple boiler. The boiler should be operated in order to avoid exports of electricity to the wholesale market when prices are below the marginal cost of the cogeneration unit. Secondly, Firm I could increase the competition on the local market by lowering the prices for supply contracts in order to maintain its market share. This strategy may altogether prevent the market entry of Firm II. The revenues of Firm II from electricity supply contracts benefit from a growing number of customers and a price increase of 1.75% per year. The 40
Note that only expenditures, revenues, and cash-flows from the electricity and heat market of the prototype city are shown. Results should serve to prove the concept of the model. If business models of real firms should be tested, a real world environment has to be modeled as well.
92
4 Commercial Actor Model
business model of Firm II tries to balance the feed-in from distributed generation with its supply obligations to local customers in order to save high voltage transmission charges. Therefore, the revenues from trades on the electricity wholesale market are approximately zero until 2025, when they start rising. Further, as the market share of distributed generation increases, expenditures due to the purchase of contracts for surplus electricity will increase as well. Finally, the total cash-flow of Firm II is negative until 2018, when it starts increasing constantly. Fig 4.2 has shown that the advertising campaign launched by Firm II can successfully stimulate the diffusion process in the second scenario after 2013, particularly among established agents. Thus, the results suggest that Firm II should not enter the market before 2013 or after 2018 to avoid false investments. Further, the results indicate that revenues from supply contracts are higher than revenues from wholesale market sales. Hence, Firm II should concentrate on advertising electricity supply contracts to local consumers after 2025 to balance demand and supply. Further, micro-cogeneration units are always operated in combination with a peak boiler and heat storage. Therefore, the operation of those distributed plants is flexible. Firm II could choose to generate more electricity when wholesale market prices are high and store the heat or operate a peak boiler if surplus electricity cannot be profitably sold. Neither Firm I nor Firm II operated the gas network or offered gas supply contracts to local consumers in the simulated scenario. Assuming that Firm I is a traditional utility company, which supplies gas, the increase in gas demand due to distributed generation will lead to additional revenues from gas contracts. Thus the market entrance of Firm II might lead to an increase in cash-flows of Firm II as well. In contrast, the business model of Firm II could be operated by a non-integrated local gas supplier, who also faces the challenge of competing on liberalized gas markets. This strategy might enable it to win over the customers and enter the electricity market. Finally, the diffusion of micro-cogeneration, which reached a market share of 12% by the end of the second scenario, affected the ownership of electricity generation in the urban area. By 2030, over 15% of the electricity demand is supplied by private distributed generation.
4.6 Discussion This chapter developed an agent-based model of heterogeneous energy firms competing in urban areas, suitably integrated with the model of private agent investment decisions given earlier. The chosen approach accounts for different initial resource endowments, different perspectives as to the development of future markets, and subsequently different commer-
4.6 Discussion
93
cial strategies. This enables us to explore how competitive advantages might be developed, sustained, or lost in urban markets. The conceptual approach was realized and tested in a prototype computer model. This proof of concept should be followed by an extended implementation. The model is built on the assumption that the different business units of firms can be distinguished, and that each option a firm might choose can be assigned to at least one of those units. It further assumes that each business unit has implemented a strategy and that these strategies can be sufficiently captured by assigning different shares of the firms’ capital to a particular unit, distinguishing between conventional and research and development capital, and using different time horizons and interest rates to evaluate the assigned options. Strategies should depend on both the initial resource endowment and the perspectives about the expected future requirements in order to be competitive. Finally, the model assumes that perspectives of firms are adaptive in fixed intervals with regard to real-world developments. The results given in the previous section show that the chosen approach might be used to test different business strategies in a competitive environment. The implemented simplified prototype did not enable firms to choose or alter options nor to adapt their perspectives. Therefore, two different scenarios were used to explore the impact of an advertising campaign on the diffusion of a technology and the resulting economic performance of firms. Further, possible options to address the changing market environment for both firms were described. Further development of the model will support these options endogenously and account for firms altering prices, adapting the advertising level or investing into technology directly. The coupling of the firm model with the private agent model was able to prove the concept of the integrated framework. Firms can influence private agents’ decisions by changing prices for commodities and by advertising technologies. Different types of private agents will respond differently to the advertising campaign of Firm II. This result yields that the chosen approach to reward feed-in electricity from micro-cogeneration could only benefit a selected target group, namely established agents and real estate managers who have a sufficient budget and can operate the technology economically in their buildings. To extend the market share beyond the one realized in the second scenario, the contract offer could allow agents to choose between a reward for the feed-in of surplus electricity or an initial investment support. This variability should enable traditional agents to install micro-cogeneration if they evaluated it to be economic, despite their limited budgets. Finally, the economic performance of firms was influenced by private agents who could choose supply technologies and efficiency standards de-
94
4 Commercial Actor Model
pending on the commonness of technologies and commodity and technology prices. As a result, the aggregated demand for different energy carriers changed and the related expenditures, revenues, and cash-flows also varied.
5 Conclusions
5.1 Introduction Today’s energy industry is undergoing fundamental institutional, commercial and technological changes. Four major drivers of change, the ongoing market deregulation, technological change, firms conduct, and climate policy have been identified and described in Chap. 1. This survey suggests that urban areas are the most likely to see essential changes in their energy infrastructure. To study these changes a highly resolved agent-based energy model was developed aiming to address the following questions:
•
Which distributed technologies percolate into energy systems, what de termines the diffusion rate, and how does the demand for and mix of energy change?
•
How does this diffusion alter the ownership structure of generation technologies and what is the likely impact on central generation?
•
To which degree do the status quo of infrastructure and ownership, corporate strategies, and public policy shape the future structure of an energy system and its related emissions, demands, and prices?
This final chapter discusses how the proposed framework and its models are suited to answer the identified questions, highlights potential applications and relevance for policy makers, and provides an outlook on bounded rational decision models.
5.2 Discussion of the Model Design This work provides a novel modeling approach to study the evolution of energy systems in urban areas. Two different classes of agent-based models for energy investment decisions, one for private actors and one for commercial actors, have been developed. Both models were designed to interact in an integrated framework consisting of two layers, the technical and the agent layer. The two layer design facilitate the combination of the
96
5 Conclusions
agent models with a highly resolved energy system model which then could be used to simulate the operation of the energy system and to inform agents’ decisions. The approach combines concepts from sociology, economics, and engineering science in an integrated framework that allows for the actual structure of energy supply systems, their current operation and future development to be modeled. It places agents that use local profit maximization routines simultaneously with agents that exhibit bounded rationality into a complex environment which is itself characterized by a multitude of interdependencies arising from the technical system and from market interactions. The decisions of the agents alter the operation and structure of the networks connecting them and thereby create a complex adaptive system. Simulation results from the private agent model as presented in Chaps. 3 and 4 show that different diffusion trajectories for supply technologies and efficiency upgrades could be obtained and interpreted with respect to building and technology characteristics and agent types. The integration of the highly resolved energy system model enables us to analyze the context dependent performance of options in different building prototypes. Further, a change in scenario data (e.g. the purchase price for surplus electricity from micro-cogeneration or the commonness of a technology) had an impact on diffusion curves. Additionally, the aggregated energy demands of a residential sector in an urban area for different energy carriers were calculated and might serve to estimate the future demands and the corresponding market size. Likewise, the introduction of commercial agents and their interaction with private agents can change diffusion trajectories for technologies and yield additional insight in the impact and performance of firms’ strategies. The presented concept to aggregate firms in order to derive commercial agent decision models is well suited to account for difference in the initial resource structure of companies. The results in Chap. 4 might be used by companies to evaluate their actual resource structure, and the threats and opportunities arising from changes in policy, technologies, and institutional settings. Further, new entrants could test their business model in the target market. In addition, the impact of the structure of companies active in an urban area on the overall development of the energy system can be assessed. Finally, the model outcomes are able to serve to assess the potential impact of policy on emissions, prices or technology diffusion.
5.3 Outlook
97
5.3 Outlook In order to obtain illustrative results, the model was applied to investigate a prototype city. The data needed to characterize the agent types and the present state of the infrastructure was based on general assumptions that must be replaced by a site specific investigation of the urban system considered, resting on empirical findings. Once private and commercial agents are grounded empirically, outcomes can be used to benchmark policy measures and strategies of firms. The infrastructure and technology data of an urban area can be obtained using sophisticated geographical information systems in combination with satellite photographs. Further, data from municipalities, energy firms and a recent census could provide sufficient information to reconstruct the present state building infrastructure, the technology endowment of each building, and the network infrastructure of the respective city. Finally, the private agent types have to be distributed among building prototypes using target group marketing data obtained by related surveys. Energy models which follow the paradigm of this work might contribute to the development of sophisticated public energy policy. The ongoing efforts of policy makers to reduce local and global emissions of harmful substances such as dust, NOx, SO2 and greenhouse gases led to the implementation of a range of policy measures targeting households, building owners, and firms. For example energy labels should rank domestic applicants by their energy consumption, solar thermal collectors are supported by a fixed investment support, loans at a reduced interest rate aim to encourage building owners to carry out efficiency upgrades, mandatory energy standards for buildings and heating system regulate the maximal energy consumption and emissions of buildings, fixed feed-in tariffs are scheduled in order to stimulate investment in renewable electricity generation, and CO2 certificates price the emission of carbon dioxide. Any policy should reach its goal efficiently. Models which are suited to better describe how actors might decide in reality could be used to test or evaluate policy measures to avoid overpay or free riding. Imagine that a government supports the installation of solar thermal collectors with a fixed amount of money per square meter. This support scheme aims to increase the profitability and thus should speed up diffusion. A test in the private agent model might yield a different picture. Note that some agents would choose environmentally benefiting technologies if they were available and affordable. Others have different requirements such as comfort or cost in the first place. Additionally, the actual market share has an impact
98
5 Conclusions
on the commonness and thereby the visibility of a given technology. The results from Chap. 3 have shown that the use of solar thermal collectors spreads among technology leaders first, and later, when the technology is visible to them, among established agents even without any financial support. Thus, a supporting scheme that does not win an additional target group is just extra money for the others41. Governments might concentrate on increasing the trust in a new technology or establishing manufacturers’ and retailers’ networks instead. The private agent model used bounded rational decision rules to determine decision outcomes. This kind of model recognizes that human decision makers adapt their decision rules to the structure of the environment, leading to a limited searching and processing of information and satisfactory decisions. To enable computational modeling a given decision problem first needs to be completely defined. Secondly, the modeler has to specify how the decision outcome is determined. The bounded rational decision models, as used in this work, introduce search rules, analysis tools, and decision strategies. Unfortunately, compared to rational choice approaches, the number of parameters to be specified considerably increased. This work proposes to overcome this problem by designing bounded rational decision models in a way that most of the parameters can be specified using structured interviews or population surveys. Therefore, two complementary approaches to cluster decision makers, one derived from social milieus and one derived from rationality categories, are suggested. This combination should enable matching questioners to model requirements and thereby develop consistent agent types for a specific problem domain. Future research, aiming to derive a consistent set of agents representing private building owners, could provide an elastic long-term demand model for the residential sector. Finally, the modeling framework, which was especially designed to analyze urban energy systems, could be transferred to other domains. For instance, research projects focusing on sustainable water resource management, i.e. the interaction between nature, water and humans, or on sociodemographic changes in city districts, might decide upon related settings.
41
An evaluation of the government support scheme for solar thermal collectors in Germany revealed that 59% would have invested in the same or a slightly different setting even without fixed financial support. Further, most investors stated that the support scheme has encouraged them to make the investment earlier or to choose a newer technology. Only 41% would not have invested in absence of the program. Additionally, two thirds of those whose applications where rejected invested anyway (Langniß et al. p. 119 ff).
Appendix
A.1 Private Actor Model This appendix provides the superstructure of all possible supply networks as used for the private actor decision model in Fig. A.1. Further, Table A.1 shows which processes of the superstructure form a single supply network. Diffusion curves for all actor types used in Chap. 3 are given in the Figs. A.2 to A.5. Finally, results from the private actor model assuming that a single actor uses the weighted adding decision strategy are given in Figs. A.6 and A.7. A.1.1 Supply Superstructure and Networks This section provides the superstructure of all possible supply networks as used for the private actor decision model in Fig. A.1. Further, Table A.1 shows which combination of processes from the superstructure forms the set of supply networks.
100
Appendix
Fig. A.1. Superstructure used for the private agent model
Appendix
101
Table A.1 Construction of residential supply networks from the superstructure
Demand_El_1
Heat_Grid
Pel_SolarBoiConv
Pel_BoiConv
Gas_SolarBoiConv
Gas_BoiCondSolar
Gas_CogenConv
Description
Gas_BoiCond
Gas_BoiConv
Networks
Oil_BoiConv
Process Name
zzzzzzzzz Electricity demand
Heat demand roomheating Heat demand water z Demand_H_2 heating Heat demand room + wazzzz zzzz Demand_H_3 ter-heating z Conventional gas boiler Boiler_Gas_1 zz Condensing gas boiler Boiler_Gas_2 Condensing gas boiler zz Boiler_Gas_3 additional z Condensing gas boiler Boiler_Gas_4 z Conventional oil boiler Boiler_Oil_1 z Conventional pellet boiler Boiler_Pellet_1 Conventional pellet boiler z Boiler_Pellet_2 additional z Gas-motor cogeneration Cogen_Gas_1 Enthalpy match solar zz Matching_Gas_1 gas boiler Enthalpy match solar Matching_Pellet_1 z pellets boiler Solar thermal collectors z Solar_H_1 small Solar thermal collectors z z Solar_H_2 large z Solar heat storage small Storage_H_1 z z Solar heat storage large Storage_H_2 zzzzzzzz Electricity import Import_El_1 z Electricity export Export_El_1 z Import_H_1 Heat import zzzzz Gas import Import_Gas_1 z Oil import Import_Oil_1 zz Pellet import Import_Pellet_1 Each point indicates which process belongs to a specific network. Demand_H_1
z
102
Appendix
A.1.2 Agent Specific Diffusion Curves Figs. A.2 to A.5 give the actor type specific results from the simulation presented in Chap. 3. The aggregated diffusion curves of each technology given there are obtained by summing up the actor specific diffusion curves given here.
Fig. A.2. Actor specific diffusion curves for the technology leader as calculated by the simulation in Chap. 3
Fig. A.3. Actor specific diffusion curves for the established actor as calculated by the simulation in Chap. 3
Appendix
103
Fig. A.4. Actor specific diffusion curves for the traditional actor as calculated by the simulation in Chap. 3
Fig. A.5. Actor specific diffusion curves for the real estate manager as calculated by the simulation in Chap. 3
A.1.3 Results from Weighted Adding Strategy The bounded rational decision model provided in Chap. 3 could also be used to model the decision of a single rational actor using the weighted adding strategy (WADD). Fig. A.6 and A.7 show results for this actor type
104
Appendix
using different weight factors Ȧ for the goals and assuming that goals are analyzed using the net present value for cost, the CO2 emissions for environmental impact, and the qualitative ranking for comfort. The discount factor was set to 4%, the time horizon to 10 years. All other parameters are equal to those used in Chap. 3.
Fig. A.6. Actor specific diffusion curves for an actor using WADD ( comfort = 0.25; environment = 0.25)
cost
= 0.5;
Fig. A.7. Actor specific diffusion curves for an actor using WADD (Ȧcost = 0.33; Ȧcomfort = 0.33; Ȧenvironment = 0.33)
Appendix
105
A.2 Commercial Actor Model An adaptation phase can be added to the commercial actor decision model. It should be entered after a fixed number of consecutive operational and structural timeframes. It compares actual energy prices, sensitivities, and the actual performance of each realized investment option to the expectations of each firm. The adaptation concerns the assumed price trajectories pfe(t) for energy carrier e, and the sensitivities Į f , ȕf , and Ȗf. Fig. A.8 shows how the adaptation phases fit in the modeling framework.
Fig. A.8. Integration of the selection and adaptation phase
Both the actual values and the slope of the expected price trajectories pfe(t) of firm f for the energy carrier e are adjusted by taking the actual scenario price trajectory into account. Therefore, the average slope of the scenario price trajectory and the slope of the expected price trajectory over the past five years are calculated. The new slope sfe of the expected price trajectory is then calculated following s fe =
(
)
1 ⎡ ( pe (t 0 ) − pe (t 0 − ∆t ) ) p fe (t 0 ) − p fe (t 0 − ∆t ) ⎤ + ⎢ ⎥. ∆t 2⎣ ∆t ⎦
(A.1)
Note that t0 is the index of the interval in which the strategy adaptation is undertaken and ¨t is the number of years between adaptation phases. Fi-
106
Appendix
nally, the expected price trajectory firm f uses in the following five consecutive structural timeframes for energy carrier e is obtained by
p fe (t ) = pe (t 0 ) + s fe t .
(A.2)
This approach adjusts price expectations taking the real price trajectories into account while also preserving optimistic or pessimistic views of firms regarding specific future prices. The sensitivity Įfc, ȕfc and Ȗfc are adjusted by
α fc = ω fα
ncto − nc (t0 −∆t )
(A.3) ,
t0
∑I
ct
t =t0 − ∆t
β fc = −ω fβ
γ fc = ω fγ
ncto − nc (t0 −∆t ) pc (t0 −∆t ) − pcto
ncto − nc (t0 −∆t ) t0
(A.4) ,
(A.5) ,
∑A
ct
t =t0 − ∆t
ω fα + ω fβ + ω fγ = 1 ,
(A.6)
where nc (t0 −∆t ) is the number of clients who held contract c, ¨t intervals before the actual interval t0 and pc (t0 −∆t ) is its price at that time. The factors ȦĮ, Ȧȕ, and ȦȖ are introduced to weight the contribution of advertising, contract price changes and infrastructure expansion to the total cash-flow.
References
Ajzen I (1991) The Theory of Planned Behavior. Organizational Behavior and Human Decision Processes. 50:179-211 de Almeida ELF (1998) Energy efficiency and the limits of market forces: The example of the electric motor market in France. Energy Policy 26:643-653. Andersen S (2001) EU Energy Policy: Interest Interaction and Supranational Authority. In: Andersen S, Eliassen KA (eds.) Making Policy in Europe. Sage Publications, London, pp 106-123 ARE (2002) Activity report regional energy supply 2000–2001 (in German: Tätigkeitsbericht regionale Energieversorgung 2000-2001). Available at: http://www.vre-online.de/vre/veroeffentlichungen/taetigkeitsberichte.shtml Augenstein E, Herbergs S, Kuperjans I, Lucas K (2005) Simulation of industrial Energy Supply Systems with integrated Cost optimization. In: Proceedings of 18th International Conference on Efficiency, Cost, Optimization, Simulation, and Environmental Impact of Energy Systems pp. 627-634 Ayres RU, Ayres LW, Warr B (2003) Exergy, power, and work in the US economy, 1900–1998. Energy – The International Journal. 28:219–273 Banfi S, Farsi M, Filippini M Jakob M. (2005). Willingness to Pay for EnergySaving Measures in Residential Buildings. CEPE Working Paper No. 45 Barney JB (1991). Firm Resources and Sustained Competitive Advantage. Journal of Management. 17:99-120 Bauknecht D, Bürger V (2003) Report in the Development of the Electricity Sector (in German: Report zur Entwicklung des Versorgungssektors Strom). Available at http://www.mikrosysteme.org/downloads.htm Bejan A, Tsatsaronis G, Moran M (1996) Thermal Design and Optimization. John Wiley & Sohns, New York Bettman JR, Luce MF, Payne JW (1998) Constructive Consumer Choice Processes. Journal of Consumer Research 25:187-217 Berger T (2001) Agent-based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agricultural Economics 25:245-260. Bourdieu P (1984) Distinction: A Social Critique of the Judgement of Taste. Harvard University Press, Cambridge Bower J, Bunn DW, Wattendrup C (2001) A model-based analysis of strategic consolidation in the German electricity industry. Energy Policy 29:987–1005 Bruckner T (1997) Dynamic energy and emissions optimization of regional energy systems (in German: Dynamische Energie- und Emissionsoptimierung regionaler Energiesysteme). Ph.D. thesis, University of Würzburg
108
References
Bruckner T, Groscurth H-M, Kümmel R (1997) Competition and synergy between energy technologies in municipal energy systems. Energy – The International Journal 22:1005–1014 Bruckner T, Morrison R, Handley C, Patterson M (2003) High-resolution modeling of energy-services supply systems using deeco: overview and application to policy development. Annals of Operations Research 121:151–180 Bruckner T, Morrison R, Wittmann T (2005) Public policy modeling of distributed energy technologies : strategies, attributes, and challenges Ecological Economics. 54:328-345 Bundesnetzagentur (2006) Monitoring report of Federal Network Agency for Electricity, Gas, Telecommunication and Post (in German: Monitoringbericht 2006 der Bundesnetzagentur für Elektrizität, Gas, Telekommunikation, Post und Eisenbahn). Available at http://www.bundesnetzagentur.de/media/archive/7263.pdf Casti JL (1997) Would-be worlds How simulation is changing the frontiers of science. John Wiley & Sons, New York Conlisk J (1996) Why bounded rationality? Journal of Economic Literature 34:669–700 Christensen CM (2000) The Innovator's Dilemma. When New Technologies Cause Great Firms to Fail. HarperBusiness, New York Davies S (1979) The Diffusion of Process Innovations. Cambridge University Press, Cambridge Davis F (1989) Perceived Usefulness, Perceived Ease of Use, and User acceptance of Information Technology. MIS Quarterly 13:318-340 Diefenbach N, Enseling A, Loga T, Hertle H, Jahn D, Duscha M (2005) Contribution of the Energy Saving Regulation (EnEV) and the Energy Saving Incentive Program “KfW-CO2- Gebäudesanierungsprogramm” to the National Climate Protection Program. Available at http://www.iwu.de/downloads/fachinfos/altbausanierung Eliasson B, Lee Y (2003) Integrated Assessment of Sustainable Energy Systems in China. Springer, Dordrecht Bosten London Epstein JM, Axtell R (1996) Growing Artificial Societies. The MIT Press, Cambridge Ernst A, Schulz C, Schwarz N, Janisch S. (2005) Shallow and deep modelling of water use in a large, spatially explicit, coupled simulation system. In: Troitzsch K. (ed) Representing Social Reality, Fölbach, Koblenz, pp 158-164 Eyre N. (1997) Barriers to energy efficiency: more than just market failure. Energy & Environment. 8:25-43 Fishbone LG, Abilock H (1981) Markal, a linear-programming model for energy systems analysis: technical description of the BNL version. International Journal of Energy Research 5:353-375 Freeman C, Soete L (1997) The Economics of Industrial Innovation. The MIT Press, Cambridge Geroski PA (2000) Models of technology diffusion. Research Policy 29:603-625 Gigerenzer G, Todd PM, ABC Research Group (1999) Simple heuristics that make us smart. Oxford University Press, Oxford New York
References
109
Gigerenzer G, Selten R (2001) Bounded rationality - the adaptive toolbox. The MIT Press, Cambridge Gigerenzer G, Engel C (2006) Heuristics and the law. The MIT Press, Cambridge Graham JR, Harvey CR (2001) The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics 60:187-243 Graham S, Marvin S (2001) Splintering Urbanism: Networked Infrastrictures, Technological Mobilities, and the Urban Condition: Networked Infrastructures, Technological Mobilites and the Urban Condition. Routledge, London New York Groscurth H-M, Bruckner T, Kümmel R (1995) Modeling of energy-services supply systems. Energy– The International Journal. 20:941–958 Grozev GV (2004) Can we simulate Australia’s National Electricity Market with the Agent-based Tool NEMSIM. In: Proceedings of the International Conference on Optimization : Techniques and Applications (ICOTA) Grubler A, Nakicenovic N, Victor DG (1999) Dynamics of energy technologies and global change. Energy Policy 27:247-280 Hake JF, Kleemann M, Kolb G (1999) Climate Protection and energy efficiency measures of buildings (in German: Klimaschutz durch energetische Sanierung von Gebäuden). Schriften des Forschungszentrums Jülich. Reihe Umwelt/Environment. Band 21 Hamacher T, Lako P, Ybema JR, Korhonen R, Aquilonius K, Cabal H, Hallberg B, Lechon Y, Lepicard S, Saez RM, Schneider T, Ward D (2001) Can fusion help to mitigate greenhouse gas emissions? Fusion Engineering and Design 58:1087-1090 Handelsblatt (2006) Strong growth of solar firms (in German: Solarfirmen wachsen kräftig). Handelsblatt Nr. 111 from June, 12st 2006, page b05 Hartman RS, Doane MJ, Woo C-K (1991) Consumer Rationality and the Status quo. The Quarterly Journal of Economics 106:141-162 Heuck K, Dettmann K-D (1999) Electricity supply: generation, transport and distribution of electricity for education and practice (in German: Elektrische Energieversorgung: Erzeugung, Transport und Verteilung elektrischer Energie für Studium und Praxis). Vieweg, Braunschweig Hewett MJ (1998) Achieving energy efficiency in a restructured electric utility industry. Report prepared for Minnesotans for an Energy Efficient Economy, Center for Energy & Environment. Available at http://www.mncee.org/pdf/util_restructt.pdf on December 17st 2006. Höffler F, Wittmann T (2007) Netting of capacity in interconnector auctions. The Energy Journal 8:113-144 Hu X (2004) Understanding generators’ bidding strategies in Australia’s National Electricity Market. In: Proceedings of the International Conference on Optimization: Techniques and Applications (ICOTA) IEA (2004) World Energy Outlook 2004. International Energy Agency, Paris IEA (2005) CO2 Emissions form fuel combustion. International Energy Agency, Paris Interlaboratory Working Group (2000) Scenarios for a Clean Energy Future. Available at: http://www.ornl.gov/sci/eere/cef
110
References
IPCC (2001) Climate Change 2001: Mitigation Cambridge University Press, Cambridge IPPC (2001 ) Climate Change 2001: Synthesis Report. Cambridge University Press, Cambridge Jaffe A, Stavins R (1994) The Energy Efficiency Gap What does it mean? Energy Policy 22:804-810 Jank R (2000) A Guidebook for Advanced Local Energy Planning (ALEP). International Energy Agency, Paris Jochem E (1999) Energy Efficiency - the Focus for Transition from an Energy Supply to an Energy Service Policy. Vierteljahreshefte des DIW, 4/99 Junginger M, Faaij A, Turkenburg WC (2005) Global experience curves for wind farms. Energy Policy 33:133–150 Pehnt M, Cames M, Fischer C, Praetorius B, Schneider L, Schumacher K, Voß J-P (2006) Micro Cogeneration. Towards Decentralized Energy Systems. Springer, Berlin Heidelberg New York Kahneman D, Knetch JL, Thaler RH (1991) Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias. Journal of Economic Perspectives. 5:193-206 Kammen DM (2006) The Rise of Renewable Energy. Scientific American, September 2006 Kleemann M, Heckler R, Kolb G, Hille M (2000) Development of the Heat Market in the residential Sector until 2050 (in German: Die Entwicklung des Wärmemarktes für den Gebäudesektor bis 2050). Schriften des FZJ, Reihe Umwelt, Band 23 Kramer N (2002) Modeling of prices in deregulated electricity markets (in German: Modellierung von Preisbildungsmechanismen im liberalisierten Strommarkt). Ph.D. thesis. Technical University of Freiberg Levy DL, Rothenberg S (2002) Heterogeneity and change in environmental strategy: technological and political responses to climate change in the automobile industry. In Hoffman A. Ventresca M. (eds.). Organizations, Policy and the Natural Environment: Institutional and Strategic Perspectives, Stanford University Press, Stanford, pp. 173-193 Langniß O, Aretz A, Böhnisch H, Steinborn F, Gruber E, Mannsbart W, Ragwitz M (2004) Evaluation of measures for using renewable energies from January 2002 to August 2004 (in German: Evaluierung von Einzelmaßnahmen zur Nutzung erneuerbarer Energien (Marktanreizprogramm) im Zeitraum Januar 2002 bis August 2004. Available at http://www.bmu.de/files/erneuerbare/energien/application/pdf/eva_map_gesa mt.pdf Lutzenhiser L (1993) Social and behavioral aspects of energy use. Annual Review of Energy and the Environment 18:247-289 Lutzenhiser L, Hackett B (1993) Social Stratification and Environmental Degradation: Understanding Households CO2 Production. Social Problems 40:50-73 Masuyama F (2001) History of Power Plants and Progress in Heat Resistant Steels. ISIJ International 41:612–625 Messner S, Strubegger M (1995) User's Guide for MESSAGE III. Working Paper 95-69. International Institute for Applied Systems Analysis, Laxenburg
References
111
Midttun A (2001) European Energy Industry Business Strategies. Elsevier, Amsterdam San Diego Oxford London North M, Macal C, Cirillo R, Conzelmann G, Koritarov V, Thimmapuram P, Veselka T (2002) Multi-agent modeling of electricity markets. In: Proceedings of Social Agents: Ecology, Exchange, and Evolution Conference, Chicago Oster S (1982) The diffusion of innovations among steel firms: the basic oxygen furnace. Bell Journal of Economics 13:45-56 Pfaffenberger W, Sioshansi FP (2006) Electricity Market Reform, An International Perspective. Elsevier, Amsterdam San Diego Oxford London Penrose E (1959) The Theory of the Growth of the Firm. John Wiley, New York Porter M (2004) Competitive Strategy. Techniques for Analyzing Industries and Competitors. Free Press, New York Power UK (2005) The Power UK Interview with Professor Stephen Littlechild. Power UK 141:42-46 Productivity Commission (2005) Energy Efficiency. Draft Report, Melbourne Reinganum JF (1989) The timing of innovation: research, development and diffusion. In: Schmalensee R, Willing R (eds). Handbook of Industrial Organization, vol. I. Elsevier, Amsterdam San Diego Oxford London Reisinger H, Dulle H, Pittermann B (2002) Distributed Generation versus Central Generation. In: VLEEM – Very Long Term Energy Environment Modelling. Verbundplan GmbH, Vienna Remme U, Goldstein GA, Schellmann U, Schlenzig C (2002) MESAP/TIMES – advanced decision support for energy and environmental planning. In: Operations Research Proceedings 2001, Duisburg, pp 59–66 Riahi K, Rubin ES, Taylor MR, Schrattenholzer L, Hounshell D. (2004) Technological learning for carbon capture and sequestration technologies. Energy Economics 26:539–564 Rogers E (1995) Diffusion of Innovations. The Free Press, New York Roth U, Häubi F, Albrecht J (1980) Interaction between the district structure and the energy supply system (in German: Wechselwirkungen zwischen der Siedlungsstruktur und dem Wärmeversorgungssystemen). Schriftreihe Raumordnung des Bundesministers für Raumordnung, Bauwesen und Städtebau. Bonn. Rumelt R. (1984) Towards a strategic theory of the firm. In: Lamb RB (ed) Competitive Strategic Management. Englewood Cliffs, New York, pp 556-570 Samuelson W, Zeckhauser R (1988) Status quo bias in decision making. Journal of Risk and Uncertainty 1:7-59 Sanstad AH, Howarth RB (1994) 'Normal' markets, market imperfections and energy efficiency. Energy Policy 22:811-818 Scheer H. (2007) Energy Autonomy. New Politics for Renewable Energy. Earthscan Publications: London Scheidt M, Jung T, Malinowski P (2004) Integrated Power Station Operation Optimization - BoFiT and Vattenfall Europe Case Study. In: Proceedings of the conference The European Electricity Market, Lodz. Schuler A, Weber C, Fahl U (2000) Energy consumption for space heating of West German households: Empirical evidence, scenario projections and policy implications. Energy Policy 28:877-894
112
References
Schulz W (2005) The Trend of Energy Markets up to the Year 2030 Reference Forecast for the Energy Sector. Oldenbourg – Industrieverlag, München Seebregts AJ, Goldstein GA, Smekens KE (2002) Energy/environmental modeling with the MARKAL family of models. In: Chamoni P, Leisten R, Martin A, Minnemann J, Stadtler H (eds) Operations Research Proceedings 2001 Duisburg, pp 75–82 Selten R (2001) What is bounded rationality? In: Gigerenzer G., Selten R. (eds) Bounded Rationality: The adaptive Toolbox, The MIT Press, Cambridge, pp. 13-36 Silberman S (2001) The energy web. Wired Magazine. 9:114–127 Simon HA (1955) A behavioural model of rational choice. Quarterly Journal of Economics 69:99-118 Simon HA (1956) Rational choice and the structure of environments. Psychological Review 63:129-138 Simon HA (1957) Models of Man. John Wiley , New York Simon HA (1990) Invariants of human behavior. Annual Review of Psychology 41:1-19 Sorrell S, Schleich J, Scott S, O’Malley E, Trace F, Boede U, Ostertag K, Radgen P (2000) Reducing Barriers to Energy Efficiency in Public and Private Organisations: Final Report, SPRU, University of Sussex, Brighton Sorrell S, Scott S, Schleich J, O’Malley (2004) The Economics of Energy Efficiency: Barriers to Cost-effective Investment. Edward Elgar, Cheltenham Sperling T, Hänsch R, Reitter B, Bertram L (2004) Estimating the energy yield by simulation – Estimating the wind power feed-in in Europe (in German: Simulation ermöglicht Energieertragsprognose - Operationelle Vorhersage der europaweiten Windstromeinspeisung) ew - Das Magazin für die Energiewirtschaft. 10:52-56 Stachowiak H (1973) General model theory (in German: Allgemeine Modelltheorie) Springer, Wien Stock G Mertsch R (1997) Operation optimization of a district heating grid using BoFiT (in German: Betriebsoptimierung mit BoFiT am Beispiel eines Fernwärmeverbundnetzes) Euroheat & Power Fernwärme international 26:565572 Stoneman P, Kwon M-J (1994) The diffusion of multiple process technologies. Economic Journal. 104:420-431 Tirole J (1988) The Theory of Industrial Organization. The MIT Press, Cambridge Tversky A (1972) Elimination by aspects: A theory of choice. Psychological Review 79:281-299 US DOE (2001) Residential Energy Consumption Survey: Housing Characteristics 2001. EIA, Energy Information Administration. Available at http://www.eia.doe.gov/emeu/recs on august 2nd, 2006. Valente TW (1996) Social network thresholds in the diffusion of innovations. Social Networks 18:69-89 Van der Voort E et al. 1985. Energy supply modeling package EFOM 12C MARK I. Vol. II (user's guide) EUR 8896 EN, Vol III EUR 8896 EN (CEC). Venkatesh V, Morris M, Davis G, Davis F (2003) User Acceptance of Information Technology: Toward a unified View. MIS Quarterly 27:425-478
References
113
Veselka T, Boyd G, Conzelmann G, Koritarov V, Macal C, North M, Schoepfle B, Thimmapuram P (2002), Simulating the Behavior of Electricity Markets with an Agent-Based Methodology: The Electricity Market Complex Adaptive System (EMCAS) Model. In Proceedings of the 22nd International Association for Energy Economics International Conference, Vancouver Weber L, Brunel M, Chaput F, Vinogradov SA, Campagne B, Canva M, Boilot JP, Brun A, (1997) Some reflections on barriers to the efficient use of energy. Energy Policy 25:833-835 Weber C, Perrels A (2000) Modelling lifestyle effects on energy demand and related emissions. Energy Policy. 28:549-566 Wenzel T, Koomey J, Rosenquist G, Sanchez M, Hanford J (1997) Energy Data Sourcebook for the U.S. Residential Sector. Lawrence Berkeley National Laboratory Wernerfelt B (1984) A resource-based view of the firm. Strategic Management Journal 5:171–80 Worldwatch Institute (2005) Renewables 2005: Global Status Report. Renewable Energy Policy Network for the 21st Century. Available at http://www.worldwatch.org/node/3984 Zilker SJ, Haarer D, Neij L (1997) Use of experience curves to analyse the prospects for diffusion and adoption of renewable energy technology. Energy Policy 25:1099-1107