Agent-Based Models of Energy Investment Decisions

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
Ren 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. Liner
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 Universitt Berlin
Institute for Energy Engineering
Marchstr.18
10587 Berlin
wittmann@iet.tu-berlin.de
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 & Vckler 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
3.2.2 Neoclassical Perspective.......................................................... 33
3.2.3 Behavioral Perspective ............................................................ 34
Contents
3.3 Bounded Rational Decision Models ............................................... 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
3.4.2 Aggregation of Technology and Infrastructure Information ... 42
3.4.3 Aggregation of Socio-economic Information.......................... 44
3.5 Results ............................................................................................ 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
4.2.2 Theoretical Background .......................................................... 68
4.2.3 Empirical Evidence ................................................................. 71
4.3 Aggregation of the Firm ................................................................. 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
p
RD
s
S
SR
SD
t
T
u
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 (Hffler 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.
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).
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 2050 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). Todays 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
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 1990s 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
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.
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.
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).
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 1.55MW off-shore. Consequentially, blade diameters
have risen to 125m. Costs have declined by 1218% with each doubling of
3
GE Energy bought Enron Wind in 2003 and Siemens bought Bonus in 2004.
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 2530% 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.
Worldwatch Institute (2005) estimates the energy cost for solar cells to be 0.16
0.32/kWh and for wind turbines to be 0.030.05/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.
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 20082012 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.
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
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
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
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

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.
Dependent on national norms, urban areas are defined as areas in which the
population density exceeds 200 inhabitants/km 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/km/a in 2001. e.dis supplied 3,257,000
customers. In contrast, the Neckarwerke Stuttgart AG faced a demand density of
10,938MWh/km/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.
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 1020 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.
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, 19011918, 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.
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
Wrzburg 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.
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 2550 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).
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-
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 actors 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
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

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
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
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
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 2050 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
3.3 Bounded Rational Decision Models

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
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 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 Tverskys (1972) decision strategy elimination by

aspects.
38
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
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 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:
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 =
Ct
(1 + i)
t =1
(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
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 =
g u og
(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.

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
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-
43
detached building, small residential building, large residential building,

multi-story building) and construction period (before 1900, 19011918,
19191948, 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
3.4.3 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.
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
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
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
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.
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
budget
125/m
125/m
250/m
debt
no
no
yes
comfort
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
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
AL environment
AL comfort
constraints
budget
250/m
250/m
500/m
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
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
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
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
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
99 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
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
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
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
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
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
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

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.
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 firms 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
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
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
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
firms history influenced the degree to which future technological options
were viewed as an opportunity or a threat34. 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
consumers 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.

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
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.
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
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.
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
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
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 tomorrows 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 firms 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
rates i and time horizons T are distinguished. A single strategy s fb of firm f

for the business unit b consists of six values:
s 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 = {s 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 fc reflecting the expected responses toward advertisements for a certain contract c,
fc reflecting the expected sensitivity toward price changes for contract c
and fc 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 firms 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 components
capacity, the energy demand within the control domain and the supply ob-
80
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 s 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
s fb. The profitability of options is calculated using the net present value
npv =
t =1
36
Ct
(1 + i)
(4.5)
The software tool deeco was developed as a PhD project at the University of
Wrzburg 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 s 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 + fc I c + fc p c + fc Ac p c
(4.6)
where nc is the number of customers who actually hold a contract c. fc, fc,
and fc are elasticities: fc 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. fc
is the assumed sensitivity of customer response if contract prices are
changed according to pc. Finally fc 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, fc is the expected share of buildings which will actually
connect.
82
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
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 IIs 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
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 IIs 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.
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
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.
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
type. In this example, after 2013 some established agents would select micro-cogeneration if they knew about it. Firm IIs 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
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
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
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
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
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
Todays 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 (Langni 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
Appendix
103
Fig. A.4. Actor specific diffusion curves for the traditional actor as calculated by
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 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 (cost = 0.33;
comfort = 0.33; environment = 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 f , f , and f. 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 fc, fc and fc are adjusted by
fc = f
ncto nc (t0 t )
(A.3)
,
t0
ct
t =t0 t
fc = f
fc = f
ncto nc (t0 t )
pc (t0 t ) pcto
ncto nc (t0 t )
t0
(A.4)
,
(A.5)
,
ct
t =t0 t
f + f + f = 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
, , and 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 20002001 (in German: Ttigkeitsbericht 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, 19001998. Energy The International Journal. 28:219273
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, Brger 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 29:9871005
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 Wrzburg
108
References
Bruckner T, Groscurth H-M, Kmmel R (1997) Competition and synergy between

energy technologies in municipal energy systems. Energy The International
Journal 22:10051014
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 121:151180
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 fr Elektrizitt, 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
34:669700
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- Gebudesanierungsprogramm 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, Flbach, 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, Kmmel R (1995) Modeling of energy-services supply systems. Energy The International Journal. 20:941958
Grozev GV (2004) Can we simulate Australias 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 Gebuden). Schriften des Forschungszentrums Jlich. 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 krftig). 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
fr 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.
Hffler F, Wittmann T (2007) Netting of capacity in interconnector auctions. The
Energy Journal 8:113-144
Hu X (2004) Understanding generators bidding strategies in Australias 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 33:133150
Pehnt M, Cames M, Fischer C, Praetorius B, Schneider L, Schumacher K, Vo 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 Wrmemarktes fr den Gebudesektor 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
Langni O, Aretz A, Bhnisch 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 Einzelmanahmen 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 41:612625
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 5966
Riahi K, Rubin ES, Taylor MR, Schrattenholzer L, Hounshell D. (2004) Technological learning for carbon capture and sequestration technologies. Energy
Economics 26:539564
Rogers E (1995) Diffusion of Innovations. The Free Press, New York
Roth U, Hubi F, Albrecht J (1980) Interaction between the district structure and
the energy supply system (in German: Wechselwirkungen zwischen der Siedlungsstruktur und dem Wrmeversorgungssystemen). Schriftreihe Raumordnung des Bundesministers fr Raumordnung, Bauwesen und Stdtebau. 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, Mnchen
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 7582
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. 9:114127
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, OMalley 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, OMalley (2004) The Economics of Energy Efficiency: Barriers to Cost-effective Investment. Edward Elgar, Cheltenham
Sperling T, Hnsch R, Reitter B, Bertram L (2004) Estimating the energy yield by
simulation Estimating the wind power feed-in in Europe (in German: Simulation ermglicht Energieertragsprognose - Operationelle Vorhersage der europaweiten Windstromeinspeisung) ew - Das Magazin fr 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 Fernwrmeverbundnetzes) Euroheat & Power Fernwrme 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 5:17180
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

Agent-Based Models of Energy Investment Decisions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Agent-Based Models of Energy Investment Decisions

Uploaded by

Copyright:

Available Formats

Sustainability and Innovation

Sustainability and Innovation

Dr.-Ing. Tobias Wittmann

Berlin, September 2007

At the start of the 21 century societies face the challenge of securing an

1 Drivers of Change and Energy Models................................................. 1

3.3 Bounded Rational Decision Models ............................................... 36

List of Abbreviations, Symbols and Indices

List of Abbreviations, Symbols and Indices

1 Drivers of Change and Energy Models

1 Drivers of Change and Energy Models

A survey of the major drivers of change in the energy industry suggests

1.2 Drivers of Change

1.2 Drivers of Change

1 Drivers of Change and Energy Models

energy is necessarily conserved. Therefore, technological change in the

1.2 Drivers of Change

Thesis II Technological Change: New distributed technologies have a

Corporate universities and in-house

1 Drivers of Change and Energy Models

1.2 Drivers of Change

1 Drivers of Change and Energy Models

1.2.5 Climate Policy

See http://www.unfccc.int for details on the UNFCCC.

1.3 Energy Models a Review of the State of the Art

Thesis IV Climate Policy: Efforts to stabilize and reduce greenhouse gas

1.3 Energy Models a Review of the State of the Art

1 Drivers of Change and Energy Models

In most cases, the operation of energy technologies is only simulated for

1.4 Motivation and Research Questions

1.4 Motivation and Research Questions

Technical high-resolution models, intertemporal optimization models

The subsequent chapters are structured as follows. Chap. 2 will describe

1 Drivers of Change and Energy Models

2.2 Users and Intentions

2.3 Geographical and Socio-economic Scope

2.3 Geographical and Socio-economic Scope

2.3.2 Geographical Scope

2.3.3 Socio-economic Scope

2.4 The Layer Concept

2.4 The Layer Concept

Fig. 2.1. Technical layer and agent layer

Assumption III Layers: A technical energy system and a socio-economic

Fig. 2.2. Hierarchy of timeframes

2.4 The Layer Concept

Assumption IV Modeling Timeframes: A one hour resolution covering

2.4.3 Technical Layer

To facilitate numerical modeling, the amount of data used to specify the

Fig. 2.3. Aggregation of infrastructure15

2.4 The Layer Concept

Assumption V Aggregation of Infrastructure and Technology: An urban

Control domains16 cluster network components and connections which are

2.4 The Layer Concept

conditions like temperature, wind speed, and insulation on the performance

A list of deeco features can be obtained from http://iet.tu-berlin.de/deeco.

Only 3.7% of private consumers switched electricity supply contracts between

2.4 The Layer Concept

Assumption VIII Private Actors: Private actors use heuristics to make

Commercial actors operate and/or invest into energy systems as one of

A range of software tools is commercially available. See e.g.

2.4.5 Energy Markets

2.4 The Layer Concept

gated demand of all commercial agents in one distribution system on an

Models such as MARKAL, TIMES and GEMS may be used.