Professional Documents
Culture Documents
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
Tobias Wittmann
Agent-Based
Models of Energy
Investment Decisions
Physica-Verlag
A Springer Company
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
Abstract
st
Contents
Contents
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
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
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
1.1 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).
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
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.
Disruptive technology
Digital photography
Mobile telephony
Distributed power generation
Internet protocols and Java applet
model
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.
10
11
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?
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.
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.
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.
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.
15
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.
17
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.
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
20
2 Model Design
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
Control Domains
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
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
25
Commercial actors
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.
27
22
28
2 Model Design
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.
3.1 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
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.
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
34
35
36
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
38
39
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:
(3.3)
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-
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.
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.
43
44
25
I would like to thank Martin Beckenkamp and Fritz Reusswig for helpful discussions.
45
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
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
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
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
49
constraints
decision strategy
analysis tool
search rule
traditionalist,
technology leader,
medium rational- medium rationality
ity
search rule
find_next
find_common
find_common
search domain
peer_group,
status_quo
topical
15%
7.50%
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
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
10%
5%
5%
search order
cost
discount rate
3%
6%
time horizon
15 years
10 years
environment
CO2-emissions
comfort
qualitative
qualitative
qualitative
strategy
LEX
LEX
LEX
goal ranking
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
3.5.1 Introductory Remarks
A prototype software of the decision model was developed, parameterized
and applied. This section discusses some exemplary decision outcomes. At
first, the general decision matrix is presented, followed by the single decision outcomes of the model for each type of agent for one representative
building. Finally, the aggregated outcome of the model, which was applied
to a prototype city, is given.
3.5.2 General Decision Matrix
The set of options consists of nine technology options and three efficiency
options, as described above. Each efficiency option can be combined with
each technology option, which leads to a set of 27 options in the general
decision matrix. The general goals of minimum cost, minimum environmental impact, and maximum comfort are calculated using three analysis
tools for the first two goals, and one for the last. Table 3.5 presents the
completed general decision matrix including the operational costs. Each
possible analysis tool was applied. The matrix was generated using the energy system optimization software tool deeco (Bruckner et al. 2003) coupled to a decision modeling software tool. The matrix is based on demand
and cost data for a typical single family building built between 1958 and
1968 and located in a quarter where gas and district heating grids are
available. The present state supply option is a conventional gas boiler
(Gas_BoiConv) with no insulation upgrades realized so far (maintenance). Note that an equivalent matrix can be calculated for each building type having access to different local infrastructures.
The general goal of minimum cost is determined using three different
analysis tools. The investment costs and operational costs can be understood intuitively by recognizing that additional insulation increases investment costs but saves energy and therefore operational costs. The calculation of the payback period and the net present value is done with
reference to the present state of the building. The payback period of the
present state is set to the aspiration level of the agent who carries out the
analysis. A negative net present value indicates that the regarded option is
not profitable in reference to the present state option; a positive value
represents the present cash-flow, when selecting that option in relation to
the status quo option.
52
3.5 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
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
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
3.5 Results
59
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)
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
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
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
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.
4.1 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
69
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
71
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
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.
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.
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
BU f BU .
(4.1)
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
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.
78
s fb
cs conventional
iconventional
T
= conventional
cs R & D
i R&D
T
R& D
(4.3)
S f = {s fb | b BU f } .
(4.4)
79
80
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.
81
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
83
38
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
86
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
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.
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
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:
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.
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
41
Appendix
100
Appendix
Appendix
101
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
z
102
Appendix
Fig. A.2. Actor specific diffusion curves for the technology leader as calculated by
the simulation in Chap. 3
Fig. A.3. Actor specific diffusion curves for the established actor as calculated by
the simulation in Chap. 3
Appendix
103
Fig. A.4. Actor specific diffusion curves for the traditional actor as calculated by
the simulation in Chap. 3
Fig. A.5. Actor specific diffusion curves for the real estate manager as calculated
by the simulation in Chap. 3
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
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)
References
108
References
References
109
110
References
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