Supply and Demand Dynamics in Energy Systems Modeling
Transcription
Supply and Demand Dynamics in Energy Systems Modeling
UNIVERSIDADE TÉCNICA DE LISBOA INSTITUTO SUPERIOR TÉCNICO Supply and Demand Dynamics in Energy Systems Modeling André Alves Pina Supervisor: Doctor Paulo Manuel Cadete Ferrão Co-supervisor: Doctor Carlos Augusto Santos Silva Thesis approved in public session to obtain the PhD Degree in Sustainable Energy Systems Jury final classification: Pass with Distinction Chairperson: Chairman of the IST Scientific Board Members of the Committee: Doctor António Manuel Barros Gomes de Vallêra Doctor António Manuel de Oliveira Gomes Martins Doctor Paulo Manuel Cadete Ferrão Doctor Luı́s Manuel de Carvalho Gato Doctor Carlos Augusto Santos Silva Doctor Stephen Robert Connors 2012 UNIVERSIDADE TÉCNICA DE LISBOA INSTITUTO SUPERIOR TÉCNICO Supply and Demand Dynamics in Energy Systems Modeling André Alves Pina Supervisor: Doctor Paulo Manuel Cadete Ferrão Co-supervisor: Doctor Carlos Augusto Santos Silva Thesis approved in public session to obtain the PhD Degree in Sustainable Energy Systems Jury final classification: Pass with Distinction Chairperson: Chairman of the IST Scientific Board Members of the Committee: Doctor António Manuel Barros Gomes de Vallêra, Full Professor (Retired) of Faculdade de Ciências, Universidade de Lisboa Doctor António Manuel de Oliveira Gomes Martins, Full Professor of Faculdade de Ciências e Tecnologia, Universidade de Coimbra Doctor Paulo Manuel Cadete Ferrão, Full Professor of Instituto Superior Técnico, Universidade Técnica de Lisboa Doctor Luı́s Manuel de Carvalho Gato, Associate Professor of Instituto Superior Técnico, Universidade Técnica de Lisboa Doctor Carlos Augusto Santos Silva, Invited Assistant Professor do Instituto Superior Técnico, Universidade Técnica de Lisboa Doctor Stephen Robert Connors, Researcher, Massachusetts Institute of Technology 2012 Resumo Actualmente, os maiores desafios dos sistemas de energia são a mudança para sistemas com elevadas penetrações de energias renováveis e o aumento da eficiência energética, tanto na oferta como na procura. O desenvolvimento de planos de transição para sistemas sustentáveis de energia que sejam económica e tecnicamente possı́veis requer o uso de modelos energéticos detalhados. No entanto, as metodologias mais utilizadas não estão preparadas para realizar análises a longo-prazo e simultâneamente levar em conta as dinâmicas de oferta e procura. O trabalho apresentado nesta tese analisa a importância de considerar as dinâmicas das fontes de energia renováveis e da introdução de mudanças nos padrões de consumo de electricidade, através da implementação de medidas de redução de consumo ou da introdução de veı́culos eléctricos, na optimização de investmentos em nova capacidade geradora. Uma nova metodologia com resolução temporal elevada que permite modelar a evolução no longo-prazo de sistemas de energia é proposta e testada para desenvolver planos de investimentos para sistemas com altas penetrações de fontes de energias renováveis. Os resultados mostram que considerar as dinâmicas é crucial para evitar o sobreinvestimento em nova capacidade geradora e aumentar a eficiência do sistema. Palavras-chave: modelação de sistemas energéticos; fontes de energia renová-veis; TIMES; sistemas de armazenamento de energia; veı́culos eléctricos; eficiência energética. i ii Abstract The shift towards high penetrations of renewable energy sources and the increase in energy efficiency, on both the supply and demand sides, are currently the main challenges that energy systems are facing. To design economically and technically feasible transition pathways to more sustainable energy systems, detailed energy models are required. However, the most commonly used modeling methodologies are not prepared to perform long-term analysis of energy systems while taking into account the inherent hourly dynamics of supply and demand. The work presented in this thesis analyzes the importance of accounting for the dynamics of renewable energy sources availability and of the introduction of changes in the patterns of electricity consumption, through the implementation of demand side management policies or the deployment of electric vehicles, in the optimization of investments in new generation capacity. A new framework that allows modeling the long-term evolution of energy systems while having a high temporal resolution is proposed and tested to develop investment plans for systems with high penetrations of renewable energy sources. The results show that accounting for the dynamics is crucial for avoiding the overinvestment in new generation capacity and increase the cost-effectiveness of the systems. Keywords: energy systems modeling; renewable energy resources; TIMES; energy storage systems; electric vehicles; energy efficiency. iii iv Acknowledgments I am grateful to many people, more than I can enumerate by name, for the guidance, help, encouragement and friendship during the development of this work. To my Supervisor Professor Paulo Ferrão, for all the opportunities he provided and the challenges he put me through which allowed me to learn more than I could ever expect from this whole experience. To my Co-Supervisor Carlos Silva, for his continuous teaching, support and, even more important, his friendship. His optimism and enthusiasm are definitely contagious. To Stephen Connors, from MIT, who provided valuable feedback throughout my work and enabled me to work and contact with incredible people from MIT. To Miguel Carvalho and Edward Spang, whom I had the pleasure to work closely with for the MIT Portugal Program and the Green Islands Project, as well as Ana Quaresma, João Fumega and everyone at the MIT Portugal Program I contacted with. All of them provided an incredibly friendly and supportive working atmosphere. To Gustavo Souza, Vı́tor Leal, Gonçalo Pereira and Alexandra Moutinho and all others who I was able to collaborate with for all their comments, suggestions and discussions. To my parents, my brother and my family for always encouraging me to keep improving and for providing a loving environment for me to do it. To all my friends, for the many funny and crazy times we had together. To Patrı́cia, whom I admire deeply for being an example of excellence, courage, dedication and determination. Thank you for all the patience and v support. Finally, this work would not have been possible without the financial support of the MIT Portugal Program and FCT Portuguese National Science Foundation though scholarship funding SFRH/BD/35334/2007. vi Acronyms CO2 Carbon dioxide EU27 European Union 27 countries EV Electric Vehicle DER-CAM Distributed Energy Resources Customer Adoption Model DSM Demand Side Management GHG Greenhouse Gases IPAC Integrated Policy Assessment model for China MARKAL MARket ALlocation NEMS National Energy Modeling System LEAP Long-range Energy Alternatives Planning System OECD Organisation for Economic Co-operation and Development RES Renewable Energy Sources TIMES The Integrated Markal-Efom System V2G Vehicle to Grid vii viii Contents 1 Introduction 1.1 1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 High penetration of renewable energy sources . . . . . 3 1.1.2 Energy storage systems . . . . . . . . . . . . . . . . . 5 1.1.3 Demand side management . . . . . . . . . . . . . . . . 6 1.1.4 Electric vehicles . . . . . . . . . . . . . . . . . . . . . 7 1.2 Research questions . . . . . . . . . . . . . . . . . . . . . . . . 8 1.3 Research strategy . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.3.1 Extended TIMES modeling methodology . . . . . . . 9 1.3.2 Hybrid modeling framework . . . . . . . . . . . . . . . 11 1.4 Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.5 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.6 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . 13 2 State of the art in energy systems modeling 2.1 15 Existing modeling methodologies . . . . . . . . . . . . . . . . 17 2.1.1 LEAP . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.1.2 NEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.1.3 MARKAL . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.1.4 TIMES - The Integrated Markal-Efom System . . . . 25 2.1.5 HOMER . . . . . . . . . . . . . . . . . . . . . . . . . . 27 ix 2.2 2.1.6 DER-CAM . . . . . . . . . . . . . . . . . . . . . . . . 28 2.1.7 EnergyPLAN . . . . . . . . . . . . . . . . . . . . . . . 29 Evolution of energy models . . . . . . . . . . . . . . . . . . . 29 2.2.1 Before 1980 - The energy crisis and the push towards nuclear . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 1981-1990 - The diversification of alternative energy sources 2.2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 2.4 3.2 34 2001-2010 - The emergence of sustainable energy systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Hybrid methodologies . . . . . . . . . . . . . . . . . . . . . . 36 2.3.1 PERSEUS and AEOLIUS . . . . . . . . . . . . . . . . 37 2.3.2 IPAC . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.3.3 TIMES and EMPS . . . . . . . . . . . . . . . . . . . . 41 2.3.4 GEMINI-E3 and TIAM . . . . . . . . . . . . . . . . . 42 2.3.5 The Altos Integrated Market Model Suite . . . . . . . 43 Key findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3 Case studies 3.1 33 1991-2000 - Renewable energy sources as a response to environmental concerns . . . . . . . . . . . . . . . . 2.2.4 32 47 The Azores islands . . . . . . . . . . . . . . . . . . . . . . . . 47 3.1.1 São Miguel . . . . . . . . . . . . . . . . . . . . . . . . 48 3.1.2 Flores . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Mainland Portugal . . . . . . . . . . . . . . . . . . . . . . . . 51 4 High-temporal resolution modeling of the dynamics of energy systems 4.1 55 Modeling hourly electricity dynamics for po-licy making in long-term scenarios . . . . . . . . . . . . . . . . . . . . . . . . 55 4.1.1 Aim of the study . . . . . . . . . . . . . . . . . . . . . 56 4.1.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . 57 x 4.2 4.3 The relevance of the energy resource dynamics in the mid/longterm energy planning models . . . . . . . . . . . . . . . . . . 60 4.2.1 Aim of the study . . . . . . . . . . . . . . . . . . . . . 60 4.2.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . 62 Key findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5 Modeling the dynamics of electricity demand 5.1 5.2 5.3 65 The impact of demand side management strategies in the penetration of renewable electricity . . . . . 65 5.1.1 Aim of the study . . . . . . . . . . . . . . . . . . . . . 66 5.1.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . 67 Modeling the introduction of electric vehicles in an island . . 69 5.2.1 Modeling methodology . . . . . . . . . . . . . . . . . . 69 5.2.2 Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . 71 5.2.3 Main results . . . . . . . . . . . . . . . . . . . . . . . . 72 Key findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 6 Hybrid modeling framework for high penetrations of renewable energy sources 6.1 6.2 77 Integrated Modeling Framework for Energy Systems Planning 77 6.1.1 Aim of the study . . . . . . . . . . . . . . . . . . . . . 78 6.1.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . 79 High-resolution modeling framework for the planning of electricity systems with high 6.3 penetration of renewables . . . . . . . . . . . . . . . . . . . . 81 6.2.1 Aim of the study . . . . . . . . . . . . . . . . . . . . . 81 6.2.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . 82 Key findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 xi 7 Conclusions and future work 7.1 State of the art in energy systems modeling . . . . . . . . . . 7.2 High-temporal resolution modeling of the dynamics of energy 87 87 systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 7.3 Modeling the dynamics of electricity demand . . . . . . . . . 90 7.4 Hybrid modeling framework for high penetrations of renew- 7.5 able energy sources . . . . . . . . . . . . . . . . . . . . . . . . 91 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 Bibliography 93 Papers 111 Contributions in each paper . . . . . . . . . . . . . . . . . . . . . . 111 Paper I - Modeling hourly electricity dynamics for policy making in long-term scenarios . . . . . . . . . . . . . . . . . . . . . . 113 Paper II - The relevance of the energy resource dynamics in the mid/long-term energy planning models . . . . . . . . . . . . . 127 Paper III - The impact of demand side management strategies in the penetration of renewable electricity . . . . . . . . . . . . . 137 Paper IV - Integrated Modeling Framework for Energy Systems Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Paper V - High-resolution modeling framework for the planning of electricity systems with high penetration of renewables . . . . 159 xii List of Figures 1.1 State of development of technologies for Sustainable Energy Systems in terms of development stages and challenge for implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 2 Expected costs of different electricity generation technologies using renewable and non-renewable energy sources for the years 2007, 2020 and 2030 . . . . . . . . . . . . . . . . . . . . 1.3 4 Characterization of different energy storage systems in terms of rated power (MW) and discharge time (hours) . . . . . . . 5 1.4 Types of Demand Side Management strategies . . . . . . . . 7 2.1 Types of models typically used for energy analysis . . . . . . 16 2.2 Development of new modeling tools since 1970 . . . . . . . . 29 2.3 Overview of existing modeling tools based on their time horizon and the number of time periods per year they consider . 2.4 Evolution of the number of energy models, temporal resolution and time horizon from before 1980 until 2010 . . . . . . 2.5 31 Evolution of the electricity generation mix from 1970 until 2007 for a combination of countries . . . . . . . . . . . . . . . 2.6 30 32 Development of hybrid modeling frameworks and their general purpose depending on the types of energy modeling approaches that are combined . . . . . . . . . . . . . . . . . . . 2.7 Framework developed to combine the models PERSEUS and AEOLIUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.8 37 39 Framework of the Integrated Policy Assessment model for China (IPAC) . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii 40 2.9 Modeling framework using EMPS and TIMES . . . . . . . . . 42 2.10 Coupling framework using GEMINI-E3 and TIAM . . . . . . 43 2.11 Application of the Altos Integrated Market Model Suite . . . 45 3.1 Electricity production in São Miguel, by generation type, from 1994 until 2009 . . . . . . . . . . . . . . . . . . . . . . . 3.2 Electricity consumption in São Miguel, by sector, from 1994 until 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 51 Main inputs and outputs of a typical TIMES model for electricity systems . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 50 Electricity consumption in Flores, by sector, from 1994 until 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 49 Electricity production in Flores, by generation type, from 1994 until 2009 . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 48 56 Comparison of electricity production from wind energy from models with different time resolution, for the time periods of 2010-2015, 2010-2020 and 2010-2025 . . . . . . . . . . . . . . 58 4.3 Wind energy generation capacity for the different scenarios . 59 4.4 Hourly load curve for 13 days for January 2020 for Flores, supply scenario D, dynamic balance method in EnergyPLAN 63 5.1 Scenarios description according to the options for DSM . . . 67 5.2 Utilization rates of renewable electricity generators for each hour of the average day in 2020 for scenarios a) without standby power elimination b) with standby power elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 67 Fraction of possible shiftable loads that is shifted in each year for the different scenarios . . . . . . . . . . . . . . . . . . . . 68 5.4 Framework considered for modeling the introduction of EVs . 70 5.5 Electricity consumption from EVs in the scenarios considered 71 5.6 Share of the daily demand of electricity by the EVs assumed for each hour in the fixed charging strategy scenario . . . . . xiv 72 5.7 Share of RES on the electricity that was produced for EVs 5.8 Share of RES on the monthly electricity produced for EVs in 5.9 . 73 the year 2025 . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Optimal average share of the daily demand of electricity by the EVs in each hour for the scenarios with flexible charging in the year 2025 . . . . . . . . . . . . . . . . . . . . . . . . . . 74 6.1 Proposed integrated modeling framework . . . . . . . . . . . 79 6.2 Electricity production by source from TIMES initialization . 80 6.3 Electricity production by source after the last iteration . . . . 80 6.4 Modeling framework proposed . . . . . . . . . . . . . . . . . . 81 6.5 Installed capacity from renewable energy sources a) before the first iteration b) after the last iteration . . . . . . . . . . 6.6 83 Electricity mix in the different scenarios a) in the 1036 MW scenario, iteration 1 b) in the 1036 MW scenario, last iteration 84 6.7 Percentage of excess electricity from renewable energy sources in the different scenarios a) for all renewable energy sources and b) considering only solar, wind offshore and wave energies 85 xv xvi List of Tables 1.1 Studies performed and corresponding section of thesis, paper where it is presented, case study used and summary of the work 14 2.1 Modeling tools analyzed . . . . . . . . . . . . . . . . . . . . . 18 2.2 Overview of the MARKAL family of models . . . . . . . . . . 24 3.1 Installed generation capacity in São Miguel . . . . . . . . . . 49 3.2 Installed generation capacity in Flores . . . . . . . . . . . . . 51 3.3 Generation capacity (MW) in the Portuguese electricity system 3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Total installed capacity (MW) of renewable energy sources proposed by the Portuguese Government . . . . . . . . . . . . 4.1 53 Range of capacity factors in each temporal level, for each resource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 52 57 Qualitative characterization of renewable resource dynamics: + (important), O (mild), - (not important) . . . . . . . . . . 60 4.3 Generation capacity (MW) considered in each scenario . . . . 61 4.4 RES results using the integral, semi-dynamic and dynamic balance methods for all scenarios . . . . . . . . . . . . . . . . 5.1 62 Scenarios for the penetration of EVs and time of charging strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii 71 xviii Chapter 1 Introduction Worldwide, energy systems are characterized by an intensive use of fossil fuels to respond to an ever increasing demand. In 2007, coal, peat, oil and gas accounted for around 81.4% of the primary energy supplied[1]. For most countries, this implies depending heavily on imports to meet their domestic needs, which has significant economic and security of supply risks. In 2008, the EU27 had an estimated average dependence of 53.8%, with Denmark being the only net exporter of energy [2]. Furthermore, the combustion of fossil fuels results in the emission of greenhouse gases (GHG), which have been considered as one of the most important factor influencing climate change. One of the most important energy vectors is electricity. In 2007, around 31% of the primary energy consumption worldwide was used for electricity production, with an average efficiency of 39%, and electricity accounted for around 17% of total final energy consumption (21% for OECD countries) [1]. To address this, policy makers have been developing policies for reducing final energy demand, increase energy efficiency and use endogenous resources with the goal of promoting more sustainable energy systems. 1.1 Problem statement The main policies that are being discussed by policy makers include the investment in renewable energy sources (RES) for electricity and heat supply, installation of energy storage systems to improve the efficiency of electricity 1 systems, promotion of demand side management (DSM) policies to increase energy efficiency and support of technological changes, such as the introduction of electric vehicles (EVs), to shift the consumption of traditional fossil fuels to more efficient and sustainable alternatives. The wide range of options that can be adopted results from the fact that no silver bullet exists to address the sustainability issues of an energy system (national, regional or local), with different technologies having different degrees of development. While some are already mature technologies or ready to be deployed, others are still in demonstration, development or even research stage. The multitude of technologies to develop cleaner energy systems that exist or are currently being developed is shown in Figure 1.1, based on [3, 4]. This large number of options requires policy makers to assess and analyze what different solutions exist for their system in order to identify what are the benefits that can be achieved and develop transition pathways for the implementation of the most suitable set of options. Figure 1.1: State of development of technologies for Sustainable Energy Systems in terms of development stages and challenge for implementation To support the necessary studies of how technological development can 2 help increase the sustainability of a region, a large of number of modeling tools have been developed with different objectives and scope. Generally, the scope of energy system modeling tools can be categorized as at least one of the following: • to understand the relation between changes in the energy system and economy; • to analyze possible evolutions of energy systems; • to analyze the balance between energy sources and demand; • to study how different components of the electricity supply sector, such as individual plants or the electricity grid, work on a specific energy system. While all these issues are important for assessing how well a policy will work, most studies performed focus only on either the multi-year analysis that allows assessing the long-term implications of a certain policy or the high temporal resolution modeling that allows to accurately calculate the true impacts of a policy. The hypothesis of this thesis it that the combination of different approaches would allow policy makers to design more efficient policies and understand the impacts each policy can have on others. The work presented in this thesis focuses on combining the multi-year modeling approach with the balance modeling approach for the analysis of electricity systems by using TIMES [5], an existing energy modeling tool, with a high temporal resolution. This allows accounting for supply and demand dynamics to study the topics of high penetrations of RES with and without energy storage and the impact of DSM strategies and EVs, which are technologies on the demonstration or deployment phase, as shown in Figure 1.1. 1.1.1 High penetration of renewable energy sources The incorporation of RES (such as wind, hydro and solar) in the electricity mix of a system allows reducing the use of fossil fuels, which can increase the security of supply and have significant environmental benefits. However, the introduction of RES must be analyzed in detail due to their higher capital 3 investment costs, higher levelized cost and variability. The higher costs when compared to fossil fuels result, generally, from three main factors: early stage development of technologies, lack of economies of scale and low capacity factors. The increasing investment in electricity generation from RES is expected to enable a lowering of the costs of these technologies, as shown in Figure 1.2 based on [6], due to reductions in capital and maintenance costs and increases in efficiency and lifetime. With the costs of technologies using fossil fuels expected to increase until 2030, some of the technologies using RES can achieve grid parity during this period. The problem of RES variability is particularly important for the introduction of non-dispatchable sources such as small hydro systems, wind and solar. To assess the large scale investment in RES in energy systems, two approaches are typically used: long-term energy models with low temporal resolution or short-term energy models with high temporal resolution. The first approach is able to consider long-term changes in costs such as technology development and variability in fuel prices, while the second approach is capable of accounting for the hourly fluctuations of electricity production from RES. Figure 1.2: Expected costs of different electricity generation technologies using renewable and non-renewable energy sources for the years 2007, 2020 and 2030 For an efficient energy planning, however, all these factors need to be considered as the costs determine how much new generation capacity can be built, while the amount of electricity that is effectively introduced in the grid will ultimately influence the cost-effectiveness of the investment. 4 1.1.2 Energy storage systems Energy storage systems can be used to address different issues in electricity systems such as guaranteeing power quality, providing bridging power and allowing a better energy management. Power quality applications is when energy storage systems are charged and discharged in a time scale of seconds or less, only to help guarantee the quality of power in every instance. Bridging power is when energy storage systems are used to help make the transition while a new power plant comes online, which can take from seconds to minutes. Energy management is considered to be when a storage unit is used to store energy for large time periods, hours or higher, in order to utilize electricity that is generated when it is not needed in another period of time where it is needed but would be not be readily available. Examples of energy storage systems and their characteristics are shown in Figure 1.3 [7]. Figure 1.3: Characterization of different energy storage systems in terms of rated power (MW) and discharge time (hours) The use of energy storage systems is particularly important when plan- 5 ning systems with high penetrations of RES, since their variability and intermittency can lead to mismatches between electricity supply and demand as well as grid operation issues. The planning of systems with high penetrations of RES must therefore consider the option of using energy storage systems. 1.1.3 Demand side management Demand side management strategies have been used for many decades in order to promote a more efficient energy consumption. Generally, DSM strategies allow a better management of the available energy resources and can help reduce the cost of a system, as well as avoid interruption of service. There are six different objectives for DSM strategies [8], as shown in Figure 1.4: • Peak clipping consists on the reduction of electricity demand during high consumption periods, thus avoiding high electricity costs; • Valley filling consists on the increase in demand during low consumption hours, which may enable the production of electricity at a low cost; • Load shifting consists on changing the consumption of electricity from hours with high consumption to hours with low consumption, which generally results in no savings in consumption but can help decrease the overall costs; • Strategic conservation is the overall reduction of demand through technological advances such as appliances with higher efficiencies; • Strategic growth is the overall increase of electricity demand, which can result from the introduction of new applications; • Flexible load shape consists on controlling appliances from the endusers so that they can respond to signals from the supply side and change their consumption according to the suppliers needs (demand response). 6 Figure 1.4: Types of Demand Side Management strategies While on the one hand, the planning of electricity systems must take into consideration the DSM strategies that are being undertaken in order to avoid over or under investment in new generation capacity, on the other hand, the design of new DSM policies must be analyzed in detail as they can have an impact on the cost-benefit of the existing generation capacity. Furthermore, the development of demand response programs can benefit the investment in new generation capacity from RES by enabling a better match between electricity supply and demand. 1.1.4 Electric vehicles Electric vehicles are currently considered as one of the best alternatives for the use of conventional fossil fuels in the transportation sector, which has led major vehicle manufacturers to develop their own vehicles. The switch from conventional internal combustion engines to EVs enables the reduction of local GHG and particule emissions, as these are made during electricity production. The full benefits of the introduction of EVs are determined by the energy source that is used to charge the vehicles, which depends on two crucial factors: the electricity supply system and the charging hours of the vehicles. In a system with a low penetration of RES or nuclear, the environmental benefits of EVs are less significant since the traditional transportation fuels are being substituted by other fossil fuels such as natural gas, coal or fuel oil. However, in systems with a high penetration of RES, the amount of GHG emissions relative to the transportation sector can reduce significantly. 7 In what concerns the time of charging, peak consumption hours have, in general, higher marginal GHG emission factors due to the use of fossil fuel plants whereas low consumption periods can have excess of electricity from RES that can be used to charge EVs. Furthermore, the large scale introduction of EVs can pose a real problem for electricity systems if the additional electricity demand occurs during peak hours and more investment is necessary in peak power plants. However, if an intelligent charging strategy is adopted, EVs can be an important asset in the management of electricity systems as the charging of the batteries can be done in periods with low consumption or in periods in which there is excess of electricity production from RES. This would allow running the necessary power plants more efficiently as well as increasing the use of RES. 1.2 Research questions The research questions addressed in this thesis are: • Are the current modeling methodologies able to consider the necessary dynamics of electricity supply and demand and what is the impact of using different temporal resolutions? • What impact different DSM strategies have in the planning of future electricity systems? • How can the shift to electric mobility help reduce CO2 emissions and what is the impact on the electricity system? • How can we account for the seasonal and daily variability of supply and demand in order to develop expansion plans of electricity systems with high penetration of RES, and taking into account the existence of energy storage systems? 1.3 Research strategy The research questions expressed in the previous section were supported by a comprehensive literature review on the existing modeling methodologies. It 8 was mainly intended to identify the shortcomings in the modeling strategies of systems with high penetrations of RES. The challenges for energy modeling tools were analyzed and the main gap identified was associated with the lack of optimization approaches for the investment in additional renewable energy generation capacity that may take into account the seasonal, daily and hourly dynamics of energy supply and demand. This gap was addressed with the development of two new modeling methodologies: • an ”Extended TIMES modeling methodology” that is able to analyze a period of 40 years with 288 periods per year; • a ”Hybrid modeling framework” that couples the extended TIMES model with a short-term model to enable an hourly temporal resolution. Both methodologies minimize the total costs of energy systems from a holistic system approach, while assuming that all agents (consumers, policy makers, utilities and all other relevant companies) work together to enable the best possible outcome for each scenario. 1.3.1 Extended TIMES modeling methodology The extended TIMES modeling methodology consists on an innovative approach designed in TIMES, a bottom-up optimization model for multi-year analysis of energy systems. This new modeling methodology diverged from the previous applications of TIMES as it provides a higher temporal resolution by dividing each year into 4 seasons, 3 typical days per season (Saturday, Sunday and weekday) and 24 hours per day. This increase in temporal resolution allows for the consideration of seasonal, daily and hourly dynamics of energy demand and supply. The modeling of electricity supply and demand dynamics with the extended TIMES model is explained and validated in terms of its accuracy in Paper I - ”Modeling hourly electricity dynamics for policy making in longterm scenarios”. The results showed that the increase in temporal resolution is crucial to achieve better estimates for the potential of RES and allow the design of more robust investment plans for electricity systems. 9 The extended TIMES modeling methodology was compared with two other methods with different temporal resolutions concerning the estimation of electricity production from RES in Paper II - ”The relevance of the energy resource dynamics in the mid/long-term energy planning models”. This comparison showed that while the extended TIMES modeling methodology can help improve the analysis of the investment in RES, the modeling of very high penetrations of RES requires modeling frameworks with higher temporal resolutions. This new methodology was used to perform two innovative analysis on the use of DSM strategies and different time of charging strategies for EVs as means of increasing the penetration of RES. The impact of promoting different DSM strategies on the planning of electricity systems was tested in the extended TIMES modeling methodology in Paper III - ”The impact of demand side management strategies in the penetration of renewable electricity”. In particular, this methodology was used to study the introduction of demand response in the residential sector in Flores island, which has hourly dynamics and impacts on the investment decisions in new generation capacity. The introduction of dynamic demand technologies was shown to help delay the investment in new generation capacity by reducing peak demand and enabling a better management of the already existing capacity. The use of different time of charging strategies for EVs to increase the share of RES in the electricity consumed by the vehicles was studied by using the extended TIMES modeling methodology and a short-term model with hourly resolution. The extended TIMES modeling methodology was used to optimize the investment in new generation capacity over a large time horizon and the short-term model to calculate the effective electricity production from RES for each year, while considering the installed capacities as decided by TIMES. The results showed that the use of flexible time of charging strategies can help increase the share of RES in the electricity produced for the charging of the EVs by providing a better match between supply and demand. 10 1.3.2 Hybrid modeling framework To improve the accuracy of the modeling of systems with very high penetrations of RES, a new hybrid modeling framework to perform multi-year optimization of energy systems was developed. This framework consists of using an iterative algorithm in which the results obtained from the shortterm model are used to introduce constraints in the TIMES model whenever an operation condition is not met for one year. Two applications of this framework were performed in order to develop transition pathways to systems with high penetrations of RES. One application concerned the development of an investment plan for the island of São Miguel that took into account the dynamics of Geothermal and Wind energy as well as the variability of demand. This work, presented in Paper IV - ”Integrated Modeling Framework for Energy Systems Planning”, showed that the increase in temporal resolution allows avoiding the overinvestment in new generation capacity by delaying the investments which would lead to excesses in electricity production from RES that cannot be used by the system. The second application of the hybrid modeling framework was developed for mainland Portugal to study the importance of considering RES variability and complementarity for different cases of energy storage systems capacities. The results, presented in Paper V - ”High-resolution modeling framework for the planning of electricity systems with high penetration of renewables”, showed that for systems with lower energy storage capacities a detailed model of energy systems is crucial to develop investment plans that consider accurate estimations of the electricity production from each RES. 1.4 Case studies The case studies used in these works were the islands of São Miguel and Flores in the Azores archipelago and mainland Portugal. The main purpose of working with the islands of the Azores is that they represent excellent case studies due to having well defined boundaries, facing severe security of supply issues and having a large potential for the use of RES. Furthermore, the research developed using the islands of the Azores was 11 a part of the Green Islands Project, a research project on sustainable energy systems planning that is being developed as a collaboration between the MIT Portugal Program and the Government of Azores. This project has the purpose of developing sustainable pathways for the energy systems of the Azores islands, which requires the use of comprehensive modeling methodologies to understand which challenges and opportunities can arise in each island. As such, São Miguel was used as a case study for the development of the proposed modeling methodologies due to its particular constraints regarding the use of geothermal energy, which is considered as a non-dispatchable energy source by the local electricity company. The island of Flores was used to test the impact of the introduction of DSM strategies and of EVs since it presents already a considerably large penetration of RES, with several hours during the year having 100% renewable electricity. This creates an opportunity to use dynamic demand technologies as a way of increasing the penetration of RES. The mainland Portugal case study was used to test the proposed modeling framework in a system with a diversified electricity mix and that can invest in several different energy sources. This case study also allowed understanding the need for hybrid modeling methodologies with high temporal resolution depending on the storage capacities that are considered to exist. 1.5 Publications The work developed in this thesis resulted in several conference and journal papers, some of which included in this thesis: • Paper I - ”Modeling hourly electricity dynamics for policy making in long-term scenarios” by André Pina, Carlos Silva and Paulo Ferrão published in the journal Energy Policy [9], 2011. • Paper II - ”The relevance of the energy resource dynamics in the mid/long-term energy planning models” by Gustavo Haydt, Vı́tor Leal, André Pina and Carlos A. Silva published in the journal Renewable Energy [10], 2011. 12 • Paper III - ”The impact of demand side management strategies in the pe-netration of renewable electricity” by André Pina, Carlos Silva and Paulo Ferrão published in the journal Energy [11], 2011. • Paper IV - ”Integrated Modeling Framework for Energy Systems Planning” by Carlos Silva, André Pina, Gonçalo Pereira and Alexandra Moutinho published in Volume 3 of the Proceedings of the 23rd International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems [12], 2011. • Paper V - ”High-resolution modeling framework for the planning of electricity systems with high penetration of renewables” by André Pina, Carlos Silva and Paulo Ferrão to be submitted to the journal Applied Energy. 1.6 Outline of thesis The thesis is organized in seven chapters. In Chapter 2, the existing energy systems modeling tools are compared in order to identify the gaps for effective policy making regarding electricity systems and different tools and hybrid methodologies are described. The energy systems used as case studies are presented in Chapter 3. Chapter 4 analyzes how different modeling methodologies are capable of accounting for the dynamics of electricity systems and the impact of considering the seasonal and daily variability of electricity supply and demand for designing optimal expansion plans in systems with high penetrations of RES. In Chapter 5, the impact of changes in electricity demand patterns through the promotion of DSM strategies or the introduction of EVs is analyzed. The development of a hybrid modeling framework for analyzing systems with high penetrations of RES that combines, iteratively, two types of models is presented in Chapter 6. Finally, Chapter 7 presents some conclusions of the work that was performed and presents the basis for future work. The studies included in this thesis are identified in Table 1.1 in terms of the section of the thesis and paper in which they are presented, the case study to which they were applied to and their main purpose. 13 Table 1.1: Studies performed and corresponding section of thesis, paper where it is presented, case study used and summary of the work Section 4.1 Paper I Case study São Miguel 4.2 II Flores 5.1 III Flores 5.2 - Flores 6.1 IV São Miguel 6.2 V Mainland Portugal 14 Main purpose Development of the extended TIMES modeling methodology and analysis of using different number of time steps within a day Comparison of the accuracy of using different temporal resolutions for modeling energy systems using the extended TIMES methodology, EnergyPlan and LEAP Use of the extended TIMES methodology to assess the impact of energy efficiency and demand response on the planning of electricity systems Use of the extended TIMES methodology to estimate the use of renewable electricity for electric vehicles under different time of charging strategies Development of the hybrid modeling framework and analysis of investment in new renewable generation capacity for electricity production Use of the hybrid modeling framework to analyse the investment in new generation capacity for electricity production under a CO2 restriction and considering different sizes of energy storage systems Chapter 2 State of the art in energy systems modeling The development of tools for the modeling of energy systems has been seen as crucial for a few decades now [13, 14]. With these tools, governments, energy agencies and private investors can analyze different scenarios and make better and more informed decisions regarding investments on renewable energies, energy efficiency plans, subsidies, etc [15, 16, 17]. There are four major types of energy analysis models, as shown in Figure 2.1, with very different focus and temporal and spatial resolutions: • Macro-economic models, such as GEMINI-E3 [18], are models used to analyze the relations between energy systems and the rest of the economy; • Energy systems planning models, such as LEAP [19] and TIMES [5], are models used to analyze the possible evolution of energy systems; • Energy systems balancing models, such as HOMER [20] and EnergyPLAN [21], are models used to analyze the balance between energy sources and demand and calculate how different energy sources can contribute in the supply side of an energy system; • Grid operation and dispatch models, such as DYMONDS [22], are models used to study how the electricity grid is able to cope with 15 different energy mixes and how specific power plants can be used in an energy system; In this work, the focus will be on energy systems planning and balancing models. Generally, energy modeling tools of these types are prepared to be applied to either the micro- or the macro- scales [23]. The micro-scale models are usually detailed technology models for short-term simulation/optimization of components, technologies, buildings or facilities, with high temporal resolution [24, 25]. The macro-scale models are aggregated models for long-term planning of large energy systems or making economic/policy recommendations for countries or groups of countries [26]. Due to the size of the regions being studied in the macro models, they usually have yearly time steps, which remove many of the dynamics that enable the matching of supply and demand [27]. This development of energy modeling tools has been pushed by the continuous development of new energy technologies, which needed to be studied before being implemented. Figure 2.1: Types of models typically used for energy analysis However, the technologies that are currently being developed have created the need to incorporate high detail into long-term modeling, which has pushed for the development of new modeling methodologies to narrow the 16 gap between the micro- and macro- scales [28, 29]. The convergence to this meso-scale has been done, mainly, by stretching the boundaries of existing tools from the two types of models to include more dynamics and new disaggregation levels or by combining different methodologies and tools. In this Chapter, a total of 90 modeling methodologies are compared based on different characteristics and their development is analyzed based on the evolution of energy systems through the years. Furthermore, a description is provided for the most used energy models, as well as some of the hybrid methodologies that have been developed in the last years. 2.1 Existing modeling methodologies The number of modeling tools for electricity systems has increased significantly in the last few decades, with each model being developed for a specific focus. While several studies have been made regarding the classification of energy models [30, 31, 32, 33], a different analysis is made here focusing on the time horizon, temporal resolution and the mode of operation. A total of 90 modeling tools, as shown in Table 2.1, were analyzed and categorized based on different characteristics: • Year of development; • Time horizon, which is the number of years considered by the tool: 2 years or less, between 2 and 30 years, more than 30 years; • Temporal resolution, which is the number of time periods per year: 8760 or more (hourly or higher), between 8760 and 1 (between hourly and yearly), 1 or less (yearly or less); • Optimization goal of the tool: which can be optimization of the operation of the system, optimization of investment in the system, optimization of both operation and investment or simulation of the operation of pre-determined energy systems only. 17 Table 2.1: Modeling tools analyzed Model Date Optimization goal Time horizon Time step (years) (per year) AEOLIUS [34] 2005 Simulation ≤2 ≥8760 AIM [35] 1996 Investment >30 ≤1 BALMOREL [36] 2000 Operation+investment >1 and ≤30 ≥8760 BCHP Screening Tool [37] 2003 Operation ≤2 ≥8760 CGEN [38] 2007 Operation ≤2 ≥8760 CHP capacity optimizer [39] 2005 Operation+investment >2 and ≤30 ≥8760 CHPSizer [40] 2001 Operation >30 >1 and <8760 COMPETES [41] 2004 Operation ≤2 >1 and <8760 2008 Operation+investment >30 ≥8760 COSMEE [43] 1998 Operation ≤2 ≥8760 DEARS [44] 2005 Investment >30 ≤1 ≤2 ≥8760 COMPOSE [42] DER-CAM[45] 2000 Operation+investment DIME [46] 2007 Operation+investment >2 and ≤30 >1 and <8760 DNE21 [47] 1999 Investment DTI [48] 1999 Investment E2M2s [49] >30 ≤1 >2 and ≤30 >1 and <8760 2002 Operation+investment >2 and ≤30 >1 and <8760 E3ME / E3MG / MDM [50] 1996 Investment >30 ≤1 Investment >2 and ≤30 ≤1 E4cast [51] 2000 EGEAS [52] 1979 Operation+investment >2 and ≤30 ≥8760 Elfin [53] 1991 Operation ≤2 ≥8760 EMCAS [54] 2002 Investment >30 ≥8760 EMELIE [55] 2003 Simulation >30 >1 and <8760 EMINENT [56] 2005 Simulation ≤2 ≥8760 EMPS [57] 1975 Operation >2 and ≤30 >1 and <8760 Energy 2020 [58] 1972 Simulation >2 and ≤30 ≤1 EnergyPLAN [21] 1999 Operation+investment ≤2 ≥8760 energyPRO [59] 1987 Operation+investment >30 ≥8760 ENPEP [60] 1985 Simulation >30 ≤1 ERIS [61] 1996 Investment >30 ≤1 ESPAUT [62] 2005 Investment ≤2 ≥8760 Continued on Next Page. . . 18 Table 2.1 – Continued Model Date Optimization goal Time horizon Time step (years) (per year) FOSSIL2 [63] 1972 Investment >30 >1 and <8760 genEris [64] 2005 Investment >30 ≤1 GTMax [65] 1995 Operation ≤2 ≥8760 2003 Operation+investment ≤2 ≥8760 2000 >30 ≥8760 ≤2 ≥8760 Simulation ≤2 >1 and <8760 H2A analysis [66] H2RES [67] HOMER [20] Operation 1992 Operation+investment HUD CHP Screening Tool [68] 2002 HYDROGEMS [69] 1995 Simulation ≤2 ≥8760 HYPRO [70] 2005 Simulation >2 and ≤30 ≤1 IKARUS [71] 1991 Investment >30 ≤1 1997 Operation+investment >30 >1 and <8760 INFORSE [73] 2002 Simulation >30 ≤1 Invert [74] 2003 Investment >2 and ≤30 ≤1 IMAGE-TIMER [72] IPM [75] LEAP [19] MARKAL/TIMES [76, 5] 1980 Operation+investment >2 and ≤30 1980 Simulation 1978 Operation+investment ≥8760 >30 ≤1 >30 >1 and <8760 MELP [77] 1997 Investment >2 and ≤30 ≤1 MERGE [78] 1994 Simulation >30 ≤1 Mesap PlaNet [79] 1997 Simulation >30 ≥8760 MESSAGE [80] 1978 Operation+investment >30 ≤1 MiniCAM [81] 1982 Simulation >30 ≤1 MTSIM [82] 2009 Operation ≤2 ≥8760 NARE [83] 1973 Operation+investment >2 and ≤30 >1 and <8760 NEMS [84] 1993 Simulation >2 and ≤30 NEWAVE [85] 1992 Operation >2 and ≤30 >1 and <8760 ORCED [86] 1996 Operation+investment ≤1 ≤2 ≥8760 OVER/UNDER [87] 1977 Investment >2 and ≤30 ≤1 PERSEUS [34] 1996 Investment >30 >1 and <8760 PLEXOS [88] 2003 Operation+investment >2 and ≤30 ≥8760 POLES [89] 1993 Simulation >30 ≤1 POWERS [90] 2001 Simulation >30 >1 and <8760 Continued on Next Page. . . 19 Table 2.1 – Continued Model Date Optimization goal Time step (years) (per year) >30 ≤1 PRIMES [91] 1994 ProdRisk [92] 1994 Operation+investment >2 and ≤30 PROMETHEUS [93] PROVIEW [94] 2007 Simulation Time horizon Simulation >30 ≥8760 ≤1 1980 Operation+investment >2 and ≤30 ≥8760 ≥8760 RAMSES [95] 1988 Operation >2 and ≤30 Ready Reckoner [96] 1997 Simulation >2 and ≤30 >1 and <8760 REMARK [62] 2006 Operation ≤2 ≥8760 RETScreen [97] 1996 Investment >30 >1 and <8760 ROM [98] 2007 Operation ≤2 ≥8760 RSTEM [99] 2005 Operation SAFIRE [100] 1997 Investment >30 ≤1 SAGE [101] 2002 Investment >2 and ≤30 ≤1 SAMLAST [102] 1993 Operation ≤2 >1 and <8760 SESAM [103] 1987 Simulation SimREN [104] 1999 Simulation ≤2 ≥8760 SIVAEL [105] 1990 Simulation ≤2 ≥8760 SMART [106] 2009 Operation+investment >30 ≥8760 SPSEK [107] 1984 Operation+investment >2 and ≤30 STREAM [108] 2004 Simulation System optimizer [109] 2004 Investment >2 and ≤30 >1 and <8760 >2 and ≤30 >1 and <8760 ≤2 ≤1 ≥8760 >2 and ≤30 >1 and <8760 TRNSYS16 [110] 1975 Operation+investment >30 ≥8760 UniSyD3.0 [111] 2003 Simulation >30 >1 and <8760 UPLAN [112] 1988 Operation >2 and ≤30 ≥8760 UREM [113] 2007 Investment >30 ≤1 VTT-EMM [114] 2007 Operation >2 and ≤30 >1 and <8760 WASP [112] 1972 Investment >2 and ≤30 >1 and <8760 WEM [115] 1993 Investment >2 and ≤30 ≤1 WILMAR Planning Tool [116] 2006 Operation ≤2 ≥8760 Investment >30 ≤1 WITCH [117] 2005 20 A description of some of the most used energy models [32] is given below. 2.1.1 LEAP The development of Long-range Energy Alternatives Planning System (LEAP) was initiated in 1997 by SEI-Boston (Boston Center of the Stockholm Environment Institute) [19] and other five institutions: EDRC (South Africa), ENDA (West Africa), ETC(Europe), FAO-RWEDP (Asia), IDEE (Latin America). Since then it has been used for many different purposes by organizations from more than 150 countries. LEAP is designed as an energy-environment modeling tool based on long-range scenario analysis. These scenarios are alternatives of how the energy system will evolve throughout the years and are developed using a comprehensive accounting of how energy is acquired, converted and used in a given region or economy under a range of alternative assumptions on population, economic development, technology, price and so on. LEAP, unlike macroeconomic models, does not attempt to estimate the impact of energy policies on employment or GDP. It does not automatically generate optimum or market-equilibrium scenarios, but it can be used to identify least-cost scenarios. Since LEAP is a tool intended for medium- to long-term planning, most of its calculations occur on an annual time-step. However, some calculations can be done with higher temporal resolution. The electric demand can be modeled using three alternative methods for describing electric system load. The first one is using a load duration curve (LDC) for the entire system, which is exogenously determined. The second is using seasonal/time of day load shape for each electric demand technology separately, with the overall system load shape being calculated internally by summing across the various load shapes. The third option is to use seasonal/time of day load shape for the system as a whole. Regarding the supply side, the technologies can be specified by the user and the installed capacity can either be inserted exogenously, where values are explicitly entered by the user reflecting existing capacity, or endogenously, where values are calculated internally by LEAP in order to maintain a minimum planning reserve margin. Besides the capacity, the capacity fac- 21 tor or the maximum availability is also introduced by the user. The balance between demand and supply is done using a set of dispatch rules. Some examples of uses of LEAP are: in China, the Chinese Energy Research Institute (ERI) has used LEAP to explore how China could achieve its development goals whilst also reducing its carbon intensity; in the U.S., a prominent Non-Governmental Organization, the Natural Resources Defense Council (NRDC) uses LEAP to analyze national fuel economy standards and advocate for policies that encourage clean vehicles and fuels; in Rhode Island, LEAP has been the main organizational tool for analyzing and monitoring the States award-winning GHG mitigation process, in which multiple stakeholders are guiding the States efforts to meet its GHG emission reduction goals; in the Philippines, LEAP is used by the Department of Energy to help develop its National Energy Plans. 2.1.2 NEMS The National Energy Modeling System (NEMS) was developed by Energy Information Administration of the Department of Energy of the USA [84]. It models the energy system of the United States and is mainly used to produce the Annual Energy Outlook. NEMS is an energy-economic model of US energy markets for the midterm period through 2025. It projects the production, imports, conversion, consumption, and prices of energy, subject to assumptions on macroeconomic and financial factors, world energy markets, resource availability and costs, behavioral and technological choice criteria, cost and performance characteristics of energy technologies, and demographics. As an annual model, NEMS can also provide the impacts of transitions to new energy programs and policies. NEMS is divided into several modules: four supply modules (oil and gas, natural gas transmission and distribution, coal, and renewable fuels); two conversion modules (electricity and petroleum refineries); four end-use demand modules (residential, commercial, transportation, and industrial); one module to simulate energy/economy interactions (macroeconomic activity); one module to simulate world oil markets (international energy activity); and one module that provides the mechanism to achieve a general market 22 equilibrium among all the other modules (integrating module). Since energy resources and prices, the demand for specific energy services, and other characteristics of energy markets vary widely across the United States, NEMS is a regional model. Depending on data availability and other factors, different modules can have different regional disaggregation. For example, the demand modules (e.g., residential, commercial, industrial and transportation) use the nine Census divisions, the Electricity Market Module uses 15 supply regions based on the North American Electric Reliability Council (NERC) regions, the Oil and Gas Supply Module uses 7 onshore and 3 offshore supply regions based on geologic breakdowns, and the Petroleum Market Module uses 3 regions based on combinations of the five Petroleum Administration for Defense Districts. 2.1.3 MARKAL The development of the MARKet ALlocation (MARKAL) started at the Brook-haven National Laboratory in the 70s [76]. Later on, the coordination and further development of the model was transferred to ETSAP (Energy Technology System Analysis Program). ETSAP has individual national teams (in over 35 countries) with a common, comparable and combinable methodology, mainly based on the MARKAL model, permitting in-depth national and multi-country evaluations. MARKAL is used to do long-term energy studies at the world, national, regional and state/province level. It is used with time-periods that are usually 5 or 10 years long, but allows some resolution as it recognizes three seasons (Winter, Summer and Intermediate) and two diurnal divisions (Day and Night). The model is built using as inputs data regarding end-use energy services demand for a base year, a set of existing and future technologies for energy production and consumption, and trends for future energy demands. The model then optimizes the investment and operation of existing and new energy production facilities, by minimizing the total energy system cost. MARKAL represents the energy system throughout all energy conversions, from different sources to end-use. Therefore, there are 5 different classes in which energy can flow: resources, stocks, processes (from energy 23 carrier to energy carrier), generation (from energy carrier to electricity and heat) and end-use devices. The demand of energy by the end-use devices is calculated using efficiencies as MARKAL is driven by the demand of energy services (instead of energy itself). Several different models have been created using MARKAL as a basis, as exemplified in Table 2.2 [23], using different modeling methods such as linear programming (LP), nonlinear programming (NLP), stochastic programming (SP) or mixed integer programming (MIP). Table 2.2: Overview of the MARKAL family of models Member/version Type of model Short description MARKAL LP Standard model. Exogenous energy demand. MARKAL-MACRO NLP Coupling to macro-economic mode, energy demand endogenous. MARKAL-MICRO NLP Coupling to micro-economic model, energy demand endogenous, responsive to price changes. MARKAL-ED (MED) LP As MARKAL-MICRO but with step-wise linear representation of demand function. MARKAL with NLP multiple regions Linkage of multiple country specific MARKAL-ED and MARKAL-MACRO, including trade of emissions permit. MARKAL with LP material flows Besides energy flow (electricity, heat), material flows and recycling of materials can be modelled in the RES. MARKAL with SP uncertainties MARKAL-ETL Stochastic Programming. Only with standard model. MIP Endogenous technology learning based on learning by doing curve. Specific cost decreases as function of cumulative experience. Some of the examples of the use of MARKAL [118] are the global models maintained by International Energy Agency (ETP - Energy Technology Perspectives MARKAL Model) and by the Energy Information Administration of the Department of Energy of the USA (SAGE - System for the Analysis of Global Energy markets), and the national model of China maintained by the Tsinghua University in Beijing (China Tsinghua). 24 2.1.4 TIMES - The Integrated Markal-Efom System The Integrated MARKAL-EFOM System (TIMES), introduced in 1999, is the latest development of the MARKAL framework maintained by the IEA’s Energy Technology System Analysis Program [5, 119]. TIMES is a tool used to estimate energy dynamics in local, national or multi-regional energy systems (Germany, SADC, World) [120, 121, 122] over a long-term, multi-period time horizon. The model is built through a detailed description of technologies and commodities that characterize the energy system. The main advantage that TIMES has regarding its predecessors MARKAL and EFOM is its flexibility. With TIMES, it is possible to sub-divide the year in several time periods with different lengths (defined by the user), which is useful for more detailed studies that need to incorporate more dynamics in either supply or demand. However, given that the model uses an optimization algorithm and has a very large technology base, the number of time periods will have a large impact on the computational complexity of the model. Other advantages are the possibility to have different levels of disaggregation for different sectors and the option of making investment in blocks. TIMES can be described as a dynamic partial equilibrium optimization model as it finds the minimum cost solution to provide the modeled energy demands through several time periods by making decisions on equipment investment and operation, primary energy supply and energy trades. To build the model, the user must provide information regarding end-use energy service demands for a reference case, estimates of existing stock of energy related equipment in all sectors, the characteristics of available future technologies and also present and future sources of primary energy supply and their potentials. Using an optimization algorithm, the model determines the best solution such that the quantities and prices in each time period are such that the suppliers produce exactly the quantities demanded by the consumers, which means that the total surplus is maximized. The dispatch of electricity and any other commodities is done by an ascending order of cost subject to any constraint the user defines (for example a limitation on renewable energy shares). The model can also incorporate environmental 25 issues, making it suitable to analyze the impact of energy and environmental policies. TIMES has been applied in a large number of projects, with different scopes. These range from the modeling of the energy choices in a typical (non-electrified) rural village up in South Africa up to 2017 [123], to the modeling of the World to study the impact of carbon taxes on the energy sector of each region and the emissions associated with it [124, 125]. The African village application [123] presents a model of energy system dynamics for a low-income rural community in South Africa. The authors analyze future consumption of energy services in the village and the impacts of gaining access to grid electricity. One of the main points of interest of this work is the introduction of load curves in TIMES. The authors divide the year in 4 seasons, and for each season they introduce six 4-hours time-slices. This means that for each season they have different consumption according to the hour of the day. One example of a national model is the TIMES PT model [126, 127]. This TIMES model was developed within the EU FP7 research project NEEDS and was calibrated and validated to model the Portuguese system from 2000 to 2050. The yearly time steps are divided into 4 seasons, which allows for some resolution on medium-term dynamics (such as hydro availability). TIMES PT has been used to study the evolution of the Portuguese energy system under climate change scenarios and the European Emissions Trading Schemes. The World-TIMES model [124, 125] divides the world into 15 large regions, with the energy demand being characterized in 42 segments, and determines energy dynamics for 100 years. The authors analyze the penetration level of nuclear energy under contrasted sets of assumptions on technology parameters and exogenous constraints on nuclear energy development to reflect some negative social perceptions of nuclear power. Recently, TIMES has been used in combination with other models such as EMPS [128] and GEMINI-E3 [129] to develop more accurate estimations of the evolution of energy systems. This is due to the increased flexibility of TIMES in defining time periods and processes, which makes it easy to adapt it and combine it with other models. 26 2.1.5 HOMER HOMER started being developed in 1993 by the National Renewable Energy Laboratory. It is a micro-power optimization model, indented to support design of off-grid and grid-connected power systems [20, 130]. The main goal of HOMER is to help design power systems by answering questions such as: What components does it make sense to include in the system design? How many and what size of each component should you use? HOMER uses simulation and sensitivity analysis algorithms that allow to evaluate the economic and technical feasibility of a large number of technology options and to account for uncertainty in technology costs, energy resource availability, and other variables. To do this, Homer simulates the operation of the power system at a given site by making energy balance calculations for each of the 8760 hours in a year. For each hour, HOMER compares the electric and thermal load (space heating, crop drying) to the energy that the system can supply in that hour and computes the flows of energy between each component of the system. HOMER performs these energy balance calculations for each system configuration that the user wants to test. It then determines whether a configuration is feasible, i.e., whether it can meet the electric demand under the specified conditions, and estimates the cost of installing and operating the system over the lifetime of the project. The system cost calculations account for costs such as capital, replacement, operation and maintenance, fuel, and interest. Although being commonly referred to as an optimization tool, HOMER is more like a simulation tool. After the user defines the set of configurations that he wants to test, HOMER simulates them all and then sorts them by net present cost, allowing all the system design options to be compared. Furthermore, HOMER allows the user to define sensitivity variables as inputs. For each sensitivity variable, HOMER analyzes the different system configurations. Some examples of applications of HOMER [131] are the Very Large-Scale Deployment of Grid-Connected Solar Photovoltaics in the United States: Challenges and Opportunities, and Advancing Clean Energy Use in Mexico. 27 2.1.6 DER-CAM The Distributed Energy Resources Customer Adoption Model (DER-CAM) was developed in 2000 at Lawrence Berkeley National Laboratory [45]. It is an economic model of customer Distributed Energy Resources (DER) adoption implemented in the General Algebraic Modeling System (GAMS) optimization software and CPLEX solver. The main goal of the model is to minimize the cost of supplying electric and heat loads of a specific customer site by optimizing the installation and operation of distributed generation, combined heat and power, and thermally activated cooling equipment. To do this, the model must receive as inputs the costumer’s end-use hourly load profiles (for space heat, hot water, gas, cooling and electricity), the costumer’s default electricity tariff, fuel prices and other relevant price data, the capital, operating and maintenance (O&M), the interest rate on customer investment and the basic physical characteristics of alternative generating, heat recovery and cooling technologies, including the thermal-electric ratio that determines how much residual heat is available as a function of generator electric output. The outputs of the model are the capacities of distributed generation and combined heat and power technology or combination of technologies to be installed (if any), when and how much of the capacity installed will be running and the total cost of supplying the electric and heat loads. DER-CAM can be used for many different applications. It can be used to guide choices of equipment at specific sites, or provide general solutions for example sites and propose good choices for sites with similar circumstances. It can additionally provide the basis for the operations of installed on-site generation. Also, it can be used to assess the market potential of technologies by anticipating which kinds of customers might find various technologies attractive. Some examples of applications of DER-CAM [132] are the Assessment of Distributed Energy Adoption in Commercial Buildings, The Value of Distributed Generation under Different Tariff Structures and Distributed Generation Investment by a Microgrid Under Uncertainty. 28 2.1.7 EnergyPLAN EnergyPLAN was developed in 1999 at the Aalborg University, Denmark, to simulate energy systems and provide information for the design of national or regional energy planning strategies [21]. It optimizes the operation of an energy system for a year using hourly resolution, including heat and electricity supplies and the transport and industrial sectors, by following a cost minimization strategy. It can be used for different analysis of energy systems: technical analysis, market exchange analysis and feasibility studies. The use of hourly resolution enables the model to consider the temporal dynamics of energy demand, the variability of RES, minimum technical constraints on electricity production for grid stabilization and optimization of the use of energy storage units. EnergyPLAN has been used to develop several studies that include optimizing the combination of RES [133], management of surplus electricity [134], the effect of energy storage [135]. 2.2 Evolution of energy models Due to the increased awareness regarding the importance of security of supply, economically competitive electricity systems and the potential impacts of GHG emissions to the environment, the number of tools developed has been increasing significantly. Figure 2.2 shows the distribution of the modeling tools that were analyzed since 1970 until 2010, in periods of 5 years. Figure 2.2: Development of new modeling tools since 1970 29 As shown by the figure, the number of energy modeling tools has increased significantly in the last 20 years. It should be noted that the number of models developed in the period 2006-2010 might be significantly higher, with their applications not having yet been published. The tools mentioned above were divided based on their time horizon and temporal resolution, as well as their mode of operation, as shown in Figure 2.3. From the analysis it is possible to see that there are two main types of models: • Long-term energy systems planning tools with temporal resolution of a year or less that are used to study the evolution of a system over a time period of over 30 years (tools for simulation and optimization of investment); • Short-term tools of a year or less with hourly or higher temporal resolution that are capable of analyzing with detail the operation of a system (tools for optimization of operation and optimization of both operation and investment). Figure 2.3: Overview of existing modeling tools based on their time horizon and the number of time periods per year they consider On the one hand, long-term modeling tools are characterized as being capable of optimizing the investment in new generation capacity over a large number of years or simulating pre-determined energy systems to see how well they are able to address the yearly growth of electricity demand. This is due to their low temporal resolution, which does not allow them to capture the 30 necessary dynamics for performing the optimization of operating the system. On the other hand, short-term tools are generally capable of optimizing the operation of a system, and sometimes are even able to optimize the investment in new plants taking into account only one year. This reflects their ability to model hourly dynamics of electricity supply and demand, which makes them suitable for analyzing how a system should be operated. The analysis of the models developed during different time periods shows that not only has the number of models increased significantly in recent times, but the scope of the models has also been changing, as shown in Figure 2.4 for the time periods of: before 1980, from 1981 to 1990, from 1991 to 2000 and from 2001 to 2010. Figure 2.4: Evolution of the number of energy models, temporal resolution and time horizon from before 1980 until 2010 This evolution throughout the years is a response to the evolution of electricity systems, which have been conditioned by several events such as the oil crisis of the 1970’s, the explosion of the nuclear power plant in Chernobyl in 1986 and the signing of the Kyoto Protocol in 1997. These events mark clear changes in the mix of electricity systems, as shown in Figure 2.5 for a combination of countries 1 1 [136]. The countries accounted for were Canada, United States, Japan, Australia, New 31 Figure 2.5: Evolution of the electricity generation mix from 1970 until 2007 for a combination of countries The following sections establish a relation between the evolution of the electricity systems and the development of new modeling methodologies. 2.2.1 Before 1980 - The energy crisis and the push towards nuclear The decade of 1970’s was characterized by the first worldwide energy crisis due to the shortage of petroleum in the major industrial countries. The rising oil prices, combined with geopolitical factors such as the Arab Oil Embargo of OAPEC and the Iranian Revolution, raised for the first time the concerns of energy dependency on oil and security of supply. To mitigate the potential impacts of future increases of oil prices, other solutions started to be analyzed with greater emphasis, such as the substitution of oil by other thermal energy sources, the large scale investment in nuclear energy and the impact of energy efficiency on reducing energy demand. These three topics were the driving forces for the models that were developed during this period. Zealand, Austria, Belgium, Denmark, Finland, France, Germany, Greece, Hungary, Iceland Ireland, Italy, Luxembourg, Netherlands, Norway, Poland, Portugal, Spain, Sweden, Switzerland, Turkey and United Kingdom. The limitation on the amount of countries considered was due to data availability. 32 Models such as EMPS, IPM, LEAP and MARKAL were developed with the goal of long-term planning of energy systems by comparing the traditional electricity production from oil with other thermal energy sources such as coal and natural gas, as well as the already commonly used hydro power. The main purpose of WASP, on the other hand, was analyzing the potential and competitiveness of Nuclear energy and Hydro power, while the objective behind the development of Fossil 2 and Energy 2020 (its successor) was to understand the impact of energy efficiency on investment decisions in future generation capacity. Given the long-term impacts of the topics under study, the modeling platforms that were developed during this period had the common characteristic of performing multi-year analysis. Most of these models have been further developed in order to take into account other issues that appeared in following decades, such as MARKAL/TIMES and LEAP. To this date, these two modeling tools are still continuously updated and are widely used by the modeling community, whether it is at the academic or the policy making level. Examples of the studies performed using these tools include the modeling of oil and gas supply in long-range forecasting using LEAP [137], a comparison of the energy systems of the United Kingdom, the federal Republic of Germany and Belgium using MARKAL [138] and a survey of the market for nuclear power in developing countries using WASP [139]. 2.2.2 1981-1990 - The diversification of alternative energy sources Following the oil crisis in the 1970’s, the consumption of petroleum products decreased considerably during the 1980’s which led to the decrease in oil prices. This volatility resulted in an increased interest in the diversification of energy sources, mainly nuclear energy. However, in April 1986, an explosion at the nuclear power plant in Chernobyl, Ukraine, created a plume of highly radioactive smoke fallout which spread over large parts of the Western Soviet Union and Europe. This disaster raised the issue of safety concerning nuclear power plants and public fear forced several countries to refrain from building new nuclear plants. To this 33 day, this disaster still influences the public’s perception of nuclear energy. The desire to reduce the dependency on oil and the disinvestment in nuclear energy, which was considered to be a strong contender for the replacement of conventional thermal generation, opened the opportunity for the use of alternative energy sources such as wind and solar. This led to the development of new modeling methodologies that had the objective of comparing different energy sources for electricity generation under competitive market structures while taking into account the impacts on the environment, such as EnergyPRO, ENPEP, SPSEK and others. It is particularly interesting to note that models such as RAMSES, SIVAEL and SESAM were also used to design district heating systems. Generally, the tools developed during the 1980’s were designed to look into the long-term evolution of energy systems, following the trend of the past decade. The only identified tool that had a different focus was SIVAEL, which is able to model energy systems with hourly temporal resolution but only for 1 year. Some of the studies performed during these years include an evaluation of the environmental impacts of the energy system using ENPEP [140] and the planning of energy systems while taking into account air quality using UPLAN [141]. 2.2.3 1991-2000 - Renewable energy sources as a response to environmental concerns One of the major global discussion points in the 1990’s was the potential environmental impacts of the high increase of energy demand. Following the Vienna Convention for the Protection of the Ozone Layer, and its successor Montreal Protocol on Substances That Deplete the Ozone Layer, an international agreement for the ”stabilization of greenhouse gas concentrations in the atmosphere at a level that would prevent dangerous anthropogenic interference with the climate system” was reached in Kyoto, Japan, in 1997. In what concerns energy systems, this increased awareness of the impact of human activities on the environment provided a decisive boost for the large-scale investment in RES, as their emission coefficients of GHG is much lower than that of conventional fuels. 34 In the electricity sector, two important key questions needed to be answered: were the existing renewable energy technologies economically viable (and if not, when would they be), and what were the economic and technological impacts on the system of large-scale deployment of these technologies. While the first question regards the evolution of energy systems through time, the second question concerns the day-to-day and hour-to-hour operation of the system. To address these issues, two main trends for the development of new energy models stood out, as shown in Figure 2.4: • Long-term energy models capable of simulating and optimizing the investment in new generation capacity with low temporal resolution, such as ERIS, MERGE and NEMS; • Short-term energy models that calculate the energy balances of supply and demand and the electricity spot prices with hourly resolution (or higher) for a one or two years time horizon, such as EnergyPLAN, HOMER and DER-CAM. Long-term analysis performed during this period include an analysis on emission reduction strategies using MERGE [142], the modeling of energy storage systems using batteries with NEMS [143], estimating the environmental and economic impacts of the large-scale introduction of PV in the residential sector using NEMS [144] and a post-Kyoto analysis at a global scale using ERIS [145]. Examples of analysis performed using short-term energy models include hybrid systems for a village in Chile using HOMER [146], the value of wind power using ELFIN [147], a simulation of the California power market using ELFIN [148] and the impact of carbon taxes or allowances on the electricity generation market in Ohio using ORCED [149]. 2.2.4 2001-2010 - The emergence of sustainable energy systems The 2000’s saw a significant shift on how the evolution of energy systems was perceived. Energy efficiency, smart grids, dynamic demand and EVs became the most common discussion topics to address a continuous increase of energy demand in the electricity sector and others and a new steep increase 35 in energy costs, mainly due to geopolitical factors such as the instability in the Middle East. Energy efficiency has been identified as a cost-effective solution to reduce greenhouse gases emissions and fossil fuel usage, as well as increase the economic competitiveness of companies. Smart grids and dynamic demand provided a way to better manage the necessary balancing between electricity supply and demand, which can help increase the penetration of RES and delay the investment in new generation capacity by lowering peak electricity demand. EVs are considered an opportunity for reducing GHG emissions in the road transportation sector, by using electricity produced from RES, and for performing supply and demand balancing for grid stability. The possibility of distributed energy storage using the batteries of the vehicles is also something that is being studied in detail. Due to the impacts of these emergent technologies on the operation of the energy systems, the models developed during this decade are generally characterized by having a higher temporal resolution, with most of them being able to model every hour of the year or even more. Examples of these types of models include AEOLIUS, CGEN, COMPOSE and PLEXOS. Some of the studies that were performed using these tools include analyses of the gas and electricity network of GB using CGEN [38], of different options for dealing with the intermittency of wind power in a large-scale deployment on the West Danish energy system using COMPOSE [150], of the benefits of energy storage systems and relocation options in renewable energy systems using COMPOSE [151] and of the impact of wind power on the operation costs in Germany and Scandinavia using Willmar [152]. 2.3 Hybrid methodologies While the more common approaches to energy systems modeling consist on using just one modeling tool, frameworks that combine the capabilities of two or more models have been developed in recent years to more accurately represent the complexity of the existing energy systems, as shown in Figure 2.6. The combination of different models in one framework allows taking advantage of the strengths of those models without having the large computa36 tional requirements that a single large modeling tool would require. These frameworks are developed according to the desired goal and can combine large macro-economic models with long-term energy systems planning models for economically robust analysis [153, 154], long-term energy systems planning models with short-term energy balance models for optimizing the evolution of energy systems [128, 34] or even combine short-term energy balance models with grid operation models to guarantee the technical feasibility of a given system. Furthermore, more than two types of models can be combined in complex frameworks to achieve more than one objective [155, 156, 157]. Figure 2.6: Development of hybrid modeling frameworks and their general purpose depending on the types of energy modeling approaches that are combined The next sections describe some examples of hybrid methodologies. 2.3.1 PERSEUS and AEOLIUS With the purpose of analyzing the effects of having large-scale wind power production in a system, the French-German Institute for Environmental Research of the University of Karlsruhe in Germany, developed a framework that couples the long- and short-term modeling approaches [34]. The models 37 considered in this framework are PERSEUS for the long-term modeling of the system and AEOLIUS for the short-term modeling. The PERSEUS model is a multi-period bottom-up model that is used to determine the evolution of the electricity system of a region for a time horizon of around 20 years, taking into account the potential and cost data for different renewable energy technologies. Due to the large size of the model in terms of regions and time frame, each year is only divided in 36 time slots, which results from considering 24 hours in 8 typical days. While this temporal resolution is sufficient for systems with low shares of wind energy (or other intermittent RES), a higher one is necessary to analyze systems with high shares [34]. To account for the high variability of the wind resource, PERSEUS was coupled with the AEOLIUS model, which is a power plant scheduling simulation model with high temporal resolution. AEOLIUS is able to simulate the hourly operation of an electricity system for a period of 1 year. The scheduling of power plants is done using a day-ahead approach, as occurs with the current market regulations in Germany and other countries, with the model also allowing the introduction of short-term, intra-day forecasts. As such, the model is capable of covering in detail the provision of stand-by capacities and control power from renewable and conventional capacities, as well as intermediate storage [158]. The PERSEUS and AEOLIUS models are used in an iterative process in which the investment decisions made by PERSEUS are used in AEOLIUS as inputs, as shown in Figure 2.7 [158]. AEOLIUS is then used to test the feasibility of the system and produces a number of restrictions (regarding capacities, reserve requirements and efficiency losses) that are then included in PERSEUS for a new iteration. With this iterative process it is possible to design the expansion of an existing energy system for a long-term period while ensuring an economically and ecologically optimized integration of fluctuating renewable power generation. This hybrid methodology has already been applied to analyze the largescale introduction of wind energy in the energy systems of Germany and Spain [34, 158] and the interactions between RES and emissions trading in the European electricity system [159]. 38 Figure 2.7: Framework developed to combine the models PERSEUS and AEOLIUS 2.3.2 IPAC The Integrated Policy Assessment model for China (IPAC) was based on work developed by the Energy Research Institute (ERI) since 1992 [160]. The main goal of IPAC is to develop scenarios for the evolution of the energy mix of China, as well as for GHG emissions. IPAC is a complex framework that consists of 12 different models (IPAC-SGM, IPAC-AIM/CGE, IPACTIMES, IPAC-Emission, IPAC/Tech, IPAC/SE, IPAC/EAlarm, IPAC/AIMLocal, IPAC/Gains-Asia, IPAC-health, AIM-air and IPAC-AIM/tech), as shown in Figure 2.8 [155]. The development of IPAC had benefited from several international collaboration using individual modeling tools such as AIM (developed with the National Institute for Environment Studies), SGM (developed with the Pacific Northwest National Laboratory) and TIMER (developed with RIVM). To be able to provide support to several different audiences, IPAC includes bottom-up and top-down models. Each model can be used separately and studies can be done using combinations of only some of the models. Some of the main models are: • The IPAC-AIM/technology model is a bottom-up technology assess- 39 ment tool that analyzes detailed policy options on energy and environment. It is a single region model for China and has three modules: energy service demand projection, energy efficiency estimation and technology selection. • The IPAC-SGM model is a computational general equilibrium topdown model that links economic activities with energy and environment. This model is an extension of the original SGM model and includes data for China. It considers 9 production sectors and eleven consuming sectors. • The IPAC-Emission model is a global model, for analyzing scenarios on energy and GHG emissions. It divides the World in 9 regions and considers the major emission sources, such as energy activities, industries, land use, agriculture, and forests. • The IPAC-AIM/Local model is used to perform regional analysis covering various sectors and taking into account detailed technologies. Figure 2.8: Framework of the Integrated Policy Assessment model for China (IPAC) 40 IPAC has been used in a number of studies such as the design of a low carbon scenario up to 2050 for China [155], analyzing scenarios and developing policies for energy demand and emissions in buildings in China [161], analyzing industry growth and energy use [162], assessing fiscal and financing mechanisms for energy systems in China [162] and assessing clean coal as an option for China [162]. 2.3.3 TIMES and EMPS The Institute for Energy Technology (IFE) in Norway has developed a modeling framework that uses TIMES to model the evolution of energy systems and EMPS to model the electricity market [128]. A TIMES model with higher than usual temporal resolution was developed to simulate electricity production in Norway. The model was developed using several previous MARKAL models (National and some regional models) and is capable of simulating end use options more accurately. This model divides the country in 7 regions, with each year being divided into 260 periods (every week, 5 time periods per week), and demand being characterized by 25 segments. The EMPS model is a stochastic model that optimizes the operation of hydro-thermal systems and has been widely used by stakeholders in the Nordic market for the last two decades [57]. The model is typically used for simulating the operation for a 3-5 years time horizon. The main goal of the work done by IFE is the interaction between TIMES and the EMPS model to model the power market of Norway while taking into account the evolution of the energy system and the operational and market dynamics of running the system. TIMES is used to provide the electricity demand, which is then used as input in EMPS. EMPS then calculates the electricity price for each time period. After the electricity price is computed in EMPS, this is introduced in TIMES in order to obtain the new energy system and demand. Both models are used in an iterative process to obtain a more realistic solution, as shown in Figure 2.9 [163]. While this methodology is still under development and has only been applied to single regions, the initial results showed a good agreement with the expected load curves and price development during the year. Furthermore, 41 the methodology allowed improving the modeling of hydropower storage, which is important in systems with high penetration of RES and water storage systems. The authors expect to improve the methodology and apply it to all seven regions. Open research questions include the linking of regions to allow grid exchanges and the use of Mixed Integer Programming due to the small size of the regions. Figure 2.9: Modeling framework using EMPS and TIMES 2.3.4 GEMINI-E3 and TIAM To produce better descriptions of the world energy system, a new modeling framework combining the ETSAP-TIAM energy model and the GEMINIE3 general equilibrium model was developed [154]. This methodology enables using the detailed technological representation of ETSAP-TIAM for the computation of energy flows and prices and the general equilibrium effects of GEMINI-E3 to include the main economy factors and their impacts on energy demand. GEMINI-E3 is a general equilibrium model that considers 24 regions in the World. It is able to analyze 14 different sectors and account for CO2 and GHG emissions, with a time period of 2001-2050. The model uses recursive dynamics, the expectations of agents are based on adaptive rules and the model does not presume perfect foresight. The ETSAP-TIAM is a model developed by members from ETSAP using TIMES [119]. TIAM (TIMES Integrated Assessment Model) is a global model with 15 regions covering the entire planet and a time horizon until 42 2050. TIAM is a technology rich, dynamic inter-temporal partial equilibrium model, which maximizes total surplus using linear programming. The model is driven by end-use demands and considers price elasticity. It divides the World in 15 regions and considers 9 energy commodity trades, as well as CO2 and GHG emissions. The algorithm used in the framework developed is the Gauss-Seidel Method, where the method searches for a fixed point for the useful demand by using an iterative procedure, as shown in Figure 2.10 [154]. TIAM is run to produce a fuel mix, a technological progress for each sector and the energy prices. This is then used as input in GEMINI-E3 to calculate the GDP and the industrial outputs, which are then used to build a new demand function to be used in TIAM. This procedure continues until the demand stabilizes. Figure 2.10: Coupling framework using GEMINI-E3 and TIAM 2.3.5 The Altos Integrated Market Model Suite To help Government agencies, corporations and national energy companies deal with the complexities of energy systems, the Altos Management Partners Inc have developed a fully integrated suite of models [156]. This can be used by companies engaged in asset, operation, trading, fuel acquisition and other businesses as well as by agencies that engage in regulatory, policy and other activities. The Altos Integrated Market Model Suite is an agent-based, interconnected suite of models that considers all fuels and regions of North America. 43 For natural gas and oil, the World market is considered. An example of the application of the Altos Integrated Market Model Suite is shown in Figure 2.11 [156]. In total, the Altos Integrated Market Model Suite consists of 6 different models: • The Altos North American Regional Gas (NARG) Model is a simulation model that considers the regional interactions of supply, transportation and demand of natural gas in North America. • The Altos World Gas Trade Model (WGTM) is a global model for analyzing the supply, transportation, shipping, liquefaction, regasification, infrastructure and demand of natural gas. • The Altos North American Regional Electric (NARE) Model is used to analyze fuel substition, generation investment, operation and retirement, market price competition, net energy for load and ancillary services, outbound/inbound transmissions and transmission capacity expansion in the North America electric market. • The Altos North American Regional Coal (NARC) Model simulates the mining, supply, transformation and consumption of coal in North America. It determines market clearing prices, coal reserve additions, flowing volumes, inventory additions and withdrawals and new transportation entry and exit. • The Altos ”Cap and Trade” Tradable Emissions Model analyzes the effects of cap and trade environmental markets on energy markets, and vice versa. The model is applied to the North American market and considers a 30 years time horizon. • The Altos World Oil Model (WOM) is fundamental economic supplydemand model of World oil markets. The interface developed by Altos integrates all models in a single analysis model. The models are run together and interact with each other to obtain more accurate analysis of the North America energy systems. 44 Figure 2.11: Application of the Altos Integrated Market Model Suite 2.4 Key findings The key findings after analyzing the current state-of-the-art of energy systems modeling are: • The development of new modeling methodologies, and improvement of previously existing ones, is pushed by the need to provide answers to the challenges energy systems face. • Currently, there are two main types of modeling tools: long-term energy systems planning tool with temporal resolution of a year or less that are used to study the evolution of a system over a time period of over 30 years (tools for simulation and optimization of investment) and short-term tools of a year or less with hourly or higher temporal resolution that are capable of analyzing with detail the operation of a system (tools for optimization of operation and optimization of both operation and investment). • There is a lack of modeling tools capable of addressing the expected large-scale deployment of RES, the development of new technologies 45 such as EVs and demand response and the wide spread introduction of DSM policies, which require energy models to be able to analyze the long-term evolution of energy systems while taking into account the short-term dynamics of electricity supply and demand. • A new set of energy modeling methodologies that link two or more specific energy modeling tools and are capable of producing more detailed representations of energy systems have started to be developed in the last years. However, the application of these hybrid modeling methodologies to develop systems with high penetrations of RES while considering technologies such as energy storage systems still need to be further researched. 46 Chapter 3 Case studies For the work developed in this thesis, 3 different case studies were used: 2 islands of the Azores archipelago (São Miguel and Flores) and mainland Portugal. 3.1 The Azores islands The energy systems of the nine islands of Azores are naturally isolated autonomous energy systems with limited prospects of inter-connection due to the depth of the sea and the large distance between each island. At the present moment, most of the energy systems are largely dependent on imported fuels, diesel and heavy oil, for electricity production and transports, exposing the region to the economic burden of fluctuating global oil prices and to weather enforced limitations on the normal operation of the logistic chains. In 2007 in the Azores, fossil fuels accounted for around 88% of all primary energy consumption, with the electricity production (around 38%) and transportation (around 34%) sectors being the main fossil fuels consumers [164]. The Government of Azores has an ambitious energy strategy that aims to achieve 50% of renewable electricity production in 2013 and 75% in 2018. This is in accordance with Electricity of the Azores (EDA) strategy to increase renewable penetration, which includes several investments in geothermal power plants in the major islands, like São Miguel, and several wind 47 farms and hydro stations in the smaller islands, such as Flores. Another policy goal of the Government is to increase the renewable energy penetration in the primary energy up to 40%, which is an extremely difficult target to achieve as all sectors of energy consumption must be addressed (electricity, transportation, domestic water heating and cooking, among others). 3.1.1 São Miguel Electricity supply in São Miguel consists mainly of geothermal energy, hydro power and thermal engines using heavy oil, with renewable electricity generation experiencing a large increase in the period 1994 to 2009, as illustrated in Figure 3.1 [164], as a result of a major policy of the regional government to increase the use of endogenous energy resources. The main increases in renewable electricity are due to investments in new geothermal generation capacity. Concerning the further investment in RES, the main sources available are geothermal and wind energy with a 9 MW wind park currently under construction. Figure 3.1: Electricity production in São Miguel, by generation type, from 1994 until 2009 Currently, the system is composed of 8 fuel oil thermal generation units totaling 98 MW, 2 geothermal units with a total installed capacity of approximately 27 MW and 7 run-of-river hydro systems with a total capacity of 5 MW [165], as shown in Table 3.1. 48 Table 3.1: Installed generation capacity in São Miguel Plant Energy source Year Capacity (MW) Caldeirão Fuel oil 1987 98.064 Tambores Hydro 1909 0.094 Fábrica Nova Hydro 1927 0.608 Canrio Hydro 1990 0.400 Foz da Ribeira Hydro 1990 0.800 Ribeira da Praia Hydro 1991 0.800 Salto do Cabrito Hydro 1997 0.670 Túneis Hydro 2000 1.658 Ribeira Grande Geothermal 1994 14.800 Pico Vermelho Geothermal 2007 13.000 Electricity consumption has grown considerably between 1994 and 2009, and the main consumers are the Domestic and the Commerce and Services sectors which represent around 70% of total electricity consumed in the island. Figure 3.2 shows the evolution of electricity consumption in the 5 main sectors: Domestic, Commerce and Services, Public Services, Industry and Public Lighting. The evolution presented in Figure 3.2 shows that after a high growth period, the last 3 years have shown a much smaller increase of electricity consumption. Figure 3.2: Electricity consumption in São Miguel, by sector, from 1994 until 2009 49 3.1.2 Flores Flores is one of the most isolated islands of the Azores, with an area of around 141 km2 , and around 4117 inhabitants [166]. It is the island with the largest penetration of renewable energy in electricity production in the archipelago, as around 54% of all electricity produced in 2009 was based in either hydro or wind power. In fact, renewable energies have provided a large part of the electricity needs of the island for many years now, as shown in Figure 3.3 [164]. Figure 3.3: Electricity production in Flores, by generation type, from 1994 until 2009 Currently, the electricity system in Flores is a combination of wind turbines, hydropower and diesel engines, aided by a flywheel energy storage system [167]. The system is composed of 4 hydropower generators, 2 wind turbines and 4 diesel engines. Table 3.2 shows the year of installation and the capacity of each generator [168]. In order to be able to use the total installed power of wind and hydro, a flywheel energy storage system was installed in the island. This system helps maintain frequency and voltage stability, which were the main problems for the wind-hydro-diesel system, and enables the island to have periods of 100% renewable based electricity, if the conditions are favorable. Being a small, services driven economy, the main electricity consumption sectors in Flores are Domestic and Commerce and services. Together 50 Table 3.2: Installed generation capacity in Flores Plant Energy source Year Capacity (KW) Além-Fazenda Hydro 1966 1400 Além-Fazenda Diesel 1991 2310 Boca da Vereda Wind 2002 600 they are responsible for more than 75% of all electricity consumed in the island, with the Domestic sector representing around 40%. However, the relative weight of each sector has been changing throughout the years, as electricity consumption has been growing at a higher rate in the Commerce and services sector than in the Domestic sector. Figure 3.4 [164] shows the evolution of electricity consumption in Flores, divided in 5 different sectors: Domestic, Commerce and services (private), Public services, Industries and Public lighting. Figure 3.4: Electricity consumption in Flores, by sector, from 1994 until 2009 3.2 Mainland Portugal In the last decade, Portugal has followed an ambitious energy policy with the goals of reducing the dependence on imports from fossil fuels and the sub51 stitution of oil for cleaner energy sources such as renewable energy sources and natural gas. To this end, the investment in renewable energy sources have been promoted through feed-in tariffs while several agreements have been established in order to decrease the import costs of natural gas as well as to diversify the origin of the imports. In this work, the Portuguese electricity system was considered as isolated and is analyzed considering a diverse portfolio for investment in generation capacity from renewable and non-renewable energy sources. The possible investment in nuclear energy was not considered as the focus of the work was to develop modeling methodologies for systems with high penetrations of RES. Current system The electricity generation system in Portugal has changed significantly in the last five years with large increases of the installed capacities of natural gas and wind energy and the introduction of solar and wave energy, as shown in Table 3.3 [169]. The term status producers reflects the special status that some energy sources, such as cogeneration and biomass (thermal status producers), smaller hydro, wind, solar and wave, benefit from and that enables them to not be a part of the electricity market and have priority in dispatch. The large hydro system also include a 1036 MW storage capacity which is used to maximize the electricity generation from renewable energy sources and take advantage of the low market prices of electricity during the night period. Coal Oil Natural Gas Large Hydro Thermal Status Producers Hydro Status Producers Wind Status Producers Solar Status Producers Wave Status Producers 2005 1776 1909 2166 4582 1166 333 891 0 0 2006 1776 1909 2166 4582 1295 365 1515 0 0 2007 1776 1877 2166 4578 1365 374 2048 13 0 2008 1776 1877 2166 4578 1424 385 2662 53 2 2009 1756 1878 2992 4578 1610 395 3357 95 2 2010 1756 1822 3829 4578 1698 410 3705 122 2 Table 3.3: Generation capacity (MW) in the Portuguese electricity system 52 Expected system Following the European Directive 2009/28/EC of the European Parliament and of the Council, Portugal designed a National Action Plan for Renewable Energies [170] that establishes the targets for the use of renewable energy sources in the country for the time period of 2011-2020. In what concerns the electricity system, a summary of the targets for investment in generation capacity from renewable energy sources is shown in Table 3.4 [170]. The main investments will be made in large hydro systems and onshore wind energy, with the investment in other energy sources being less significant. The storage capacity is expected to increase from the current 1036 MW to 4302 MW in this period. 2011 2012 2013 2014 2015 2016 2017 2018 2019 Large Hydro 4524 5231 5231 5476 6467 7489 8394 8712 8798 Small Hydro 457 503 503 550 550 600 650 650 700 Wind onshore 4928 5600 5600 5600 6100 6100 6100 6600 6800 Wind offshore 0 0 0 0 25 25 25 25 25 Solar 258 340 465 590 720 860 1005 1160 1325 Wave 5 5 10 35 60 75 100 125 175 2020 8798 750 6800 75 1500 250 Table 3.4: Total installed capacity (MW) of renewable energy sources proposed by the Portuguese Government While offshore wind, solar and wave energy are still considered expensive technologies, the decision to invest in these systems reflects the importance of having a diversified energy system and takes into account potential benefits of the development of local businesses and competences. 53 54 Chapter 4 High-temporal resolution modeling of the dynamics of energy systems The capability that different modeling tools have of incorporating electricity supply and demand dynamics depends largely on the temporal resolution of the model and has a clear impact on the model results. In this section, a TIMES model with 288 time periods per year that is able to consider some of the dynamics of RES and electricity demand and making investment decisions for a multi-year period is presented. The model is applied to São Miguel Island, Azores, and different temporal resolutions are compared. The proposed methodology is also compared with 2 other modeling methodologies with different temporal resolution, LEAP with 9 time periods per year and EnergyPLAN which considers all hours of the year, using Flores Island, Azores, as case study. 4.1 Modeling hourly electricity dynamics for policy making in long-term scenarios This work is presented in detail in Paper I - ”Modeling hourly electricity dynamics for policy making in long-term scenarios” by André Pina, Carlos Silva and Paulo Ferrão published in the journal Energy Policy [9]. 55 4.1.1 Aim of the study The paper presents an extension of the TIMES energy planning tool for investment decisions in electricity production that considers seasonal, daily and hourly supply and demand dynamics by dividing each year in 4 seasons, 3 days per season (weekday, Saturday and Sunday) and 24 hours per day (totaling 288 periods per year). The inclusion of these dynamics enables the model to produce more accurate results in what concerns the impact of introducing energy efficiency policies and the increased use of renewable energies. A major feature of TIMES consists on its flexibility in what concerns the time resolution, which can be defined by the user with as much or as little as desired. A typical TIMES model for analyzing the evolution of the electricity system of a region can be schematically represented as shown in Figure 4.1, with the inputs and outputs that are different due to the increase in time resolution being represented in dashed lines. This increase in time resolution needs electricity consumption profiles for each sector and profiles for the availability of renewable energies to be assumed for the different days and seasons. Figure 4.1: Main inputs and outputs of a typical TIMES model for electricity systems The model was validated in São Miguel (Azores, Portugal) for the years 2006-2009, which has already a considerable penetration of RES. In 2007, RES in São Miguel accounted for 18.8% of the primary energy supply and 56 46.7% of the electricity production, where geothermal was the main renewable energy source, representing 41.4% of the electricity produced, while hydro represented 5.3%. In the same year, 40% of primary energy was used for electricity production and 36% for transportation [164]. To increase the share of RES in the electricity sector, Electricidade dos Açores (EDA) is also planning to install wind turbines in the island, totaling an installed capacity of 9 MW by 2014. For the model used in this work, the range of capacity factors for the RES considered are presented in Table 4.1, having been defined based on historical information provided by EDA [171]. Hourly Min Max Geothermal Hydro Wind 0% 100% Resource Seasonal Min Max 36% 62% 14% 39% Yearly Min Max 55% 73% 40% 53% 25% 28% Table 4.1: Range of capacity factors in each temporal level, for each resource The benefits of using a model with increased time resolution were analyzed for two different situations. First, TIMES models using the same data but with different time resolutions of 1, 2, 3, 4, 6 and 8 hours per day (which corresponds to having 24, 12, 8, 6, 4 and 3 time periods of equal length per day) were run for the same time period (2005 - 2025) and analyzed in terms of the production of electricity from wind energy in order to quantify the impact of using models with lower temporal resolution. Second, the more accurate modeling of energy efficiency policies is studied in terms of the impact it can have on decisions concerning the investment in new generation capacity. 4.1.2 Main results Comparison of different temporal resolutions The amount of electricity produced from wind energy from the models with different time resolution was analyzed for 3 different time periods (20102015, 2010-2020 and 2010-2025). Assuming that the values obtained by the 1 hour (24 periods) time resolution model are closer to reality due to the higher level of detail, the values obtained from the different time resolution 57 models are normalized to the values obtained from the 1 hour time resolution model. The results, represented in Figure 4.2, demonstrate that the use of lower time resolutions can lead to overestimations of the renewable electricity absorbed by the system. This effect is particularly significant for the first years being analyzed (2010-2015), when the installed capacity is lower and the amount of electricity produced is also smaller, with an overestimation of around 3 times the amount of electricity from wind energy that can be absorbed by the models with low time resolution. This overestimation was due to the models with lower resolution installing more wind generation capacity in the first years than the 1 hour resolution model, which might not be economically viable as the system might not be able to absorb as much electricity as expected and therefore generate lower revenues. Figure 4.2: Comparison of electricity production from wind energy from models with different time resolution, for the time periods of 2010-2015, 2010-2020 and 2010-2025 Importance of correctly modeling electricity demand The elimination of standby power can have a significant impact on the planning of electricity systems as it will lower the consumption during the night period, which is typically characterized by low consumption. The impact of correctly modeling the elimination of standby power on the investment in new generation capacity from RES is analyzed for the São Miguel case study for the time period of 2005 - 2025. Three different situations were tested: 58 • No efficiency scenario in which no standby power is eliminated; • Low detail scenario in which the consumption from the residential sector was reduced by 5.2% every hour, which corresponds to the standby power [172]; • High detail scenario in which the overall reduction is the same but the reductions are considered to take place mostly in the night period (1:00 to 7:00) and during the working period (7:00 to 17:00), with no significant gains occurring during peak hours. The amount of wind energy generation capacity that is economically feasible in each scenario is significantly different, as shown in Figure 4.3, with the amount of wind capacity suggested by the model to be installed in the High detail scenario being much smaller than the one in the Low detail scenario for most of the time period considered (around 50% in 2020). Furthermore, the installation of 5 MW of wind energy occurs in 2015 for the No efficiency scenario, in 2016 for the Low detail scenario and only in 2018 for the High detail scenario, while, by 2025, there is not much difference between the installed capacity in all the scenarios. This is due to two effects: increase in overall electricity demand and increase in fossil fuel prices through the years, which benefits the installation of wind energy even if they cannot be used at the maximum possible capacity factor. Figure 4.3: Wind energy generation capacity for the different scenarios 59 4.2 The relevance of the energy resource dynamics in the mid/long-term energy planning models This work is presented in detail in Paper II - ”The relevance of the energy resource dynamics in the mid/long-term energy planning models” by Gustavo Haydt, Vı́tor Leal, André Pina and Carlos A. Silva published in the journal Renewable Energy [10]. 4.2.1 Aim of the study The paper presents an analysis of different modeling methodologies for balancing electricity supply sources and demand in systems with high penetration of intermittent renewable energy sources, such as wind and run-of-river hydro and explores the reasons and circumstances where common balance approaches used by mid- and long-term energy models show significant differences in the amount of electricity from RES that is dispatched. Renewable energy sources vary in different timescales, that range from seconds to decades and with different relevance depending on the resource, as shown in Table 4.2. Resource/ Timescale Hour Day Seasonal Year Decades Hydro (Dam) O + + + + Hydro (Run-of-river) + + + + + Solar O + + + O Wind + + + + O Geothermal O Biomass O O O Tidal + + + O O Wave + + + O O Table 4.2: Qualitative characterization of renewable resource dynamics: + (important), O (mild), - (not important) Due to the time-dynamics of demand and supply, one cannot simply assume that the production from RES can keep growing indefinitely with all the electricity that they could potentially produce being used by the energy system, as there will be some periods in which there is excess of available electricity and others in which there is not enough. This is particularly important as the wind, solar, run-of-river hydro, waves and other non-controllable RES become more representative in power generation. 60 In this work, three different balance methods are compared to analyze how different temporal resolutions in each year influence the amount of RES that the model assumes that can be used. The three methods are: • Integral, which consists in representing the electricity demand in a load duration curve (LDC) and matching the needs represented by the curve with the available supply options, following a dispatch rule, with each year divided in a small number of time-slices (varying, in the majority of the cases, from 5 to 10 time-slices); • Semi-dynamic, which tries to capture some of the electricity supply and demand dynamics by using only selected typical days to represent the power variations along the year instead of using all days in a year (as used for the dynamic method); • Dynamic, which consists in representing the electricity demand in an hourly load curve for a period, which can range from a day to a few years, where all the hours are represented in the same sequence as they happen. The integral, semi-dynamic and dynamic balancing methods are modeled here in LEAP, TIMES and EnergyPLAN respectively. LEAP was used with 9 time-slices for the construction of the LDC. In TIMES, each year was divided into 4 seasons, 3 days per season (Wednesday, Saturday and Sunday), and 24h in each day (total of 288 time periods per year). In EnergyPLAN, every hour of the year was modeled. To test the results of different balancing methods with different renewable energy levels, four alternative scenarios for the installed capacity of RES were considered for Flores Island, Azores, for the year 2020, as seen in Table 4.3. Resource/Technology Existing Scenario A Scenario B Scenario C Scenario D Installed 2008 2020 2020 2020 2020 Diesel 2700 2700 2700 2700 2700 Hydro 1480 1480 2080 2680 2680 Wind 600 1100 1600 2100 2800 Total 4780 5280 6380 7480 8180 Table 4.3: Generation capacity (MW) considered in each scenario 61 4.2.2 Main results The outputs of each model consisted on the amount of electricity produced from each source (hydro, wind and thermal) in each of the scenarios that were considered. The penetration of RES achieved in each scenario by each method is shown in Table 4.4, as well as indicators comparing the generation capacity from RES and peak and valley electricity demands. Scenario Renewables installed/ Renewables installed/ Demand peak Demand valley A 0.79 1.86 B 1.12 2.66 C 1.46 3.45 D 1.67 3.96 Share of renewable electricity Integral Semi-dynamic Dynamic 33.9% 34.0% 33.9% 48.6% 48.6% 48.0% 63.0% 62.6% 58.9% 72.9% 71.1% 63.9% Table 4.4: RES results using the integral, semi-dynamic and dynamic balance methods for all scenarios Comparing the results from the integral balance method in LEAP, the semi-dynamic balance method in TIMES and the dynamic balance method in EnergyPLAN, it is possible to observe that in situations where the installed capacity from RES is similar to the demand peak, or in other perspective is below to three times the demand valley (scenarios A and B), there is no relevant difference between the results of the different balance methods. However, as the generation capacity from RES increases, the share of renewable electricity estimated by the semi-dynamic and the dynamic methods becomes significantly lower than the one estimated by the integral method with the difference between the semi-dynamic and the integral methods being of 1.8 decimal points and between the dynamic and integral methods of 9 decimal points. This difference can have a significant impact on achieving (or not) a preset target. The lower penetrations obtained by the higher temporal resolution models are because they are more prepared to identify the mismatches that sometimes exist between the amount of electricity from RES that can be generated and the amount of electricity that the system needs. An example of this mismatch using the dynamic balance method in scenario D is shown in Figure 4.4 where there are a large number of hours in which the electricity that can be generated from RES far exceeds the demand. 62 Figure 4.4: Hourly load curve for 13 days for January 2020 for Flores, supply scenario D, dynamic balance method in EnergyPLAN 4.3 Key findings The key findings from the high-temporal resolution modeling of the dynamics of energy systems are: • The increase in time resolution of the models built in TIMES is crucial to achieve better estimates for the potential of RES in electricity systems and to allow designing more robust investment plans for the evolution of electricity systems. • While the proposed temporal resolution for TIMES (4 seasons, 3 days per season and 24 hours per day for each year) can help improve the analysis of the investment in RES, its results still have a significant difference when compared to the modeling with hourly resolution. • The use of low temporal resolution models can lead to the overestimation of the amount of electricity from RES that a system can use and the overinvestment in new generation capacity, which results in a reduced cost-effectiveness of the system and an underestimation of the CO2 emissions to be expected from the system. • The modeling of electricity demand dynamics must be taken into account as they can have a significant impact on the optimization of the new generation capacity from RES, depending on the penetration rate of RES. 63 64 Chapter 5 Modeling the dynamics of electricity demand The introduction of new technologies and policies in an energy system can have a clear impact on the dynamics of electricity demand by leading to the decrease or increase of consumption. Without proper planning, these changes in the electricity demand sector can lower the economic competitiveness of the electricity supply structure by making already existing capacities unnecessary or by requiring the investment in additional capacity that will be used only in short periods of time. In this chapter, the TIMES model previously presented is used to analyze the impact of different DSM strategies on the penetration of RES in the electricity sector. A study concerning EVs is also performed in order to identify the potential benefits of adopting an intelligent time of charging strategy. 5.1 The impact of demand side management strategies in the penetration of renewable electricity This work is presented in detail in Paper III - ”The impact of demand side management strategies in the penetration of renewable electricity” by André Pina, Carlos Silva and Paulo Ferrão published in the journal Energy [11]. 65 5.1.1 Aim of the study The paper presents an analysis on the impact of DSM strategies in the evolution of the electricity mix of Flores Island, Azores, which is characterized by high shares of RES in the electricity sector. The introduction of DSM strategies in the island needs to be well planned in order to enable a better management of the existing resources by reducing investment needs and increasing the use of RES. The TIMES methodology presented previously is used to model the electricity system of Flores and optimize the investment and operation of wind and hydro plants until 2020 based on scenarios for demand growth (General efficiency scenarios), promotion of behavioral changes to eliminate standby power (No standby power) and deployment of demand response technologies in the domestic sector (Dynamic demand). The different DSM strategies were considered in the model as follows: • ”General efficiency” is the option to invest in energy efficiency programs across all sectors, starting in 2011, which have the effect of reducing demand growth per year to only 50% of the one estimated with the linear regressions for each sector. • ”No standby power” is the option to gradually eliminate the standby power consumption of the Domestic sector by educating the population or the introduction of new technologies that do not have this. Standby power is estimated as being responsible for around 5% of the electricity consumed by the sector in Portugal [172]. If this option is chosen, the model gradually eliminates standby power using a linear decrease in each year, starting in 2011 and disappearing completely in 2015. This scenario is primarily intended to test the evolution of the energy infrastructure for lower demand in low consumption hours. • Finally, ”Dynamic demand” is the option to gradually enable the washing, dryer and dish washing machines to be operated remotely by the grid operator when it is more convenient. If chosen, the automation of these machines begins in 2013, with all of them having this capability in 2018. 66 Based on whether these options are considered or not, 8 different scenarios were designed, as shown in Figure 5.1. Figure 5.1: Scenarios description according to the options for DSM 5.1.2 Main results The results showed that the scenarios with higher demand growth (scenarios 1 to 4) installed all the potential generation capacity from RES and achieved higher shares of electricity from those sources than the other scenarios. Furthermore, the scenarios that considered the introduction of dynamic demand capabilities showed a delay of 3 to 4 years in the installation of new generation capacity. This was due to the dynamic demand allowing a better management of the generation capacity that was installed, as shown in Figure 5.2 for the hours of the average day of 2020 in which the scenarios 2, 4, 6 and 8 have higher utilization rates (electricity produced divided by the potential to produce it) during the night and early morning period. (a) (b) Figure 5.2: Utilization rates of renewable electricity generators for each hour of the average day in 2020 for scenarios a) without standby power elimination b) with standby power elimination 67 The higher utilization rates will improve the cost-effectiveness of the installed generation capacity by increasing the potential revenues from selling electricity. Due to having lower installed generation capacities, the scenarios with ”General Efficiency” are able to achieve higher utilization rates throughout the day. In the model that was developed, the optimal load pattern for the domestic machines is calculated in order to obtain the best economic performance of the electricity generation system. Figure 5.3 shows the fraction of loads that are shifted in each of the scenarios that have this option enabled, as well as the maximum fraction of load that can be shifted according to the assumed technology deployment (starting in 2013 and growing gradually until all machines have this possibility in 2018). Figure 5.3: Fraction of possible shiftable loads that is shifted in each year for the different scenarios The results show that all scenarios have a similar behavior, in which there is first an increase of the fraction of loads that are shifted, followed by a stabilization of this value. The use of this technology seems to stabilize at around 40% for scenarios 4, 6 and 8, and at around 35% for scenario 2. This means that even if all machines have this technology, the loads will be shifted around 40% of the time on a per year basis or that, if the penetration of this technology is lower, the machines that do have it will have their loads shifted a larger fraction of the time. 68 5.2 Modeling the introduction of electric vehicles in an island The introduction of EVs has been identified as one of the key issues for the next years due to the potential gains that can be achieved, such as the reduction of the consumption of fossil fuels in the transportation sector, the reduction of imports of fossil fuels and the maximization of the utilization of RES. On the long-run, the possibility of having EVs exchanging electricity with the grid from the vehicle to the grid (V2G) can provide significant benefits for grid stabilization and by serving as distributed storage systems. However, the impact on the life cycle of the bateries would also have to be accounted for. Electric vehicles can also present some challenges on the electricity system as the expansion of the systems needs to be addressed and, if not properly managed, the vehicles could be charged using electricity produced from fossil fuels if no other alternative is available. In this case, EVs would only be shifting the consumption of fossil fuels from the roads to the power plants. Therefore, it is important to understand how the benefits of EVs can be maximized. This work analyses the introduction of EVs in the island of Flores, Azores, and how different penetrations of EVs influence the investment in new RES generation capacity and the share of RES in the electricity produced for the EVs. The possibility of optimizing the time of charging of EVs in order to increase the use of RES is also studied. For the modeling performed it was considered that the EVs system was not able to be used for energy storage. 5.2.1 Modeling methodology To improve the modeling of electricity supply and demand dynamics, as well as the dynamics of the charging of EVs, a two step modeling approach was used, as shown in Figure 5.4. The framework used in this work consists on using first a medium-term energy model (energy systems planning model) to optimize the investment in new generation capacity from RES by taking into account the evolution of electricity demand and fuel prices through the 69 years. The output from this model is the installed capacity in each year, which is then used as input for a short-term. The short-term model (energy systems balancing model) is a model with one year time horizon and hourly temporal resolution that is able to optimize the balancing between electricity production from the different energy sources and electricity demand. The medium-term model used in this work is the same as the one used in Section 5.1 and PAPER III. The model is capable of analyzing the Flores electricity system in the time horizon of 2010-2030 and optimizes the investment in new hydro generation capacity taking into account the projected growth for the island and the penetration of EVs. The possibilities for the installation of new hydro plants were one plant of 1600 kW and another of 1040 kW that can be built from 2011 onward. The short-term model is a MATLAB model that had been developed for another work, Section 6.1 and PAPER IV, which was expanded in order to include the additional electricity demand from EVs and the capability to optimize the time of charging of the vehicles. The main constraint that was added to model concerns the charging necessities of the EVs, which were assumed to be the same every day of the year. As such, a new constraint that requires that the same amount of electricity is produced in each day for EVs was introduced in the model. Figure 5.4: Framework considered for modeling the introduction of EVs 70 5.2.2 Scenarios To study the potential benefits of the introduction of EVs, four different scenarios were designed taking into account different penetration rates of EVs and different time of charging strategies, as shown in Table 5.1. These are two decisive factors on how the introduction of EVs will impact the electricity production system of a region and which energy sources can be used to charge the vehicles. Scenario Scenario Scenario Scenario 1 2 3 4 Penetration rate of EVs Time of charging Low Fixed Low Flexible High Fixed High Flexible Table 5.1: Scenarios for the penetration of EVs and time of charging strategies Regarding the penetration rates of EVs, two different situations were considered, as shown in Figure 5.5. The Low scenario considers the introduction of EVs from 2011 and a slow growth of the electricity consumed by EVs until it reaches around 305 MWh per year by 2030. The High scenario considers a faster growth but also a delay in the introduction of EVs, starting only by 2015. In this scenario, the electricity consumption of EVs is around 1188 MWh per year by 2030. Figure 5.5: Electricity consumption from EVs in the scenarios considered Two different time of charging strategies were considered: fixed and flexible. The fixed time of charging considers the situation in which people 71 connect their vehicles to the grid and charge them right after they arrive home. The profile of the charging of EVs, shown in Figure 5.6, was based on a study concerning the mobility patterns in Portugal and the hour at which the vehicles arrive home from their last daily journey [173]. Figure 5.6: Share of the daily demand of electricity by the EVs assumed for each hour in the fixed charging strategy scenario 5.2.3 Main results For all scenarios, the medium-term model decided to invest in the new hydro plants as soon as it was possible. Due to the penetration of EVs being low in both scenarios when compared to the total electricity demand (1.7% in the low penetration rate scenario and 6.2% in the high penetration rate scenario by 2030), the share of RES in the overall electricity production is very similar in all scenarios reaching a share of 70% in 2011 and decreasing to around 60% by 2030. Regarding the share of RES on the additional electricity that is produced due to the demand by EVs, the situation is very different for the four scenarios as shown in Figure 5.7. In all scenarios, the share of RES decreases with the years due to the increasing demand and because no new generation capacity from RES is installed after 2011. For scenarios 1 and 3, the share of RES is the same until 2023 due to the low penetration of EVs. After 2023, the share of RES in scenario 3 decreases faster in accordance with the higher penetration rate of EVs. In the scenarios with flexible time of charging, scenarios 2 and 4, the share of RES is around 2.5 times higher 72 than the share of scenarios 1 and 3, respectively, due to some of the charging of EVs being shifted to periods in which there was excess of electricity from RES. The share of RES decreases faster also in scenario 4 due to the high penetration rate of EVs and the total amount of electricity that can be shifted being limited by the excess of electricity from RES in each day. Figure 5.7: Share of RES on the electricity that was produced for EVs The charging of EVs with RES occurs only when there is excess of electricity from these resources when compared to the business as usual demand. In this case, the RES used for electricity production in Flores are Hydro and Wind which have lower capacity factors during the summer months (May through August). Therefore, the share of RES in the electricity produced for the charging of EVs is lower during these months, as shown in Figure 5.8. Figure 5.8: Share of RES on the monthly electricity produced for EVs in the year 2025 73 As such, the goal of switching to EVs in order to reduce the consumption of fossil fuels would not be accomplished during the summer months as the electricity would have to be produced from fossil fuels. The use of flexible time of charging strategies is not able to help increase the share of RES during these months due to the low availability of the resources. In order to maximize the benefits of the introduction of EVs, the electricity supply sector needs to be planned accordingly to guarantee the existence of sufficient generation capacity for the additional demand that will exist. While the charging of EVs in the fixed scenario occurred primarily during the evening period (18:00 to 24:00) in the fixed time of strategy scenarios, Figure 5.6, the optimal time of charging distribution, Figure 5.9, showed that to maximize the use of RES the charging should occur during the night period (1:00 to 8:00). This is due to the lower electricity consumption during the night hours, which increases the possibility of having excess of electricity from RES during these hours. Figure 5.9: Optimal average share of the daily demand of electricity by the EVs in each hour for the scenarios with flexible charging in the year 2025 However, the charging patterns are not the same for all days as they depend on the hours in which there is excess of electricity from RES. For the days in which there was no excess of RES, the charging of EVs could be performed as soon as people reached their homes. This explains why, in the average day, the difference between electricity consumption during the night period (1:00 to 8:00) would be only slightly higher than the consumption verified during typical consumption hours (19:00 to 23:00). 74 Therefore, in order to obtain the maximum benefits by optimizing the time of charging of EVs, the system would require the automation of the charging structure as to allow the EVs to charge whenever the system operator decided it was the optimal time for the charging. The implementation of such a system would only be possible with the development of policies that could convince the people to allow the time of charging of their vehicles to be decided by the system operator. 5.3 Key findings The key findings from the modeling of the dynamics of electricity demand are: • The modeling of electricity systems needs to include the modeling of the DSM strategies that are being considered for the region, and their dynamics, as they can change significantly the optimal investment plan due to changes in the electricity demand patterns. • The introduction of dynamic demand technologies can help delay the investment in new generation capacity by reducing peak demand and enabling a better management of the already existing capacity. • The use of flexible time of charging strategies can help increase the share of RES in the electricity produced for the charging of the EVs by providing a better match between supply and demand. • The environmental benefits from switching from conventional transportation technologies to EVs can change drastically from one month to the other based on the RES that are used for the production of electricity. • Generally, the charging of EVs during the night period can increase the cost-effectiveness of the EVs and increase the environmental benefits since that is where it is most likely to have excess of electricity production from RES. 75 76 Chapter 6 Hybrid modeling framework for high penetrations of renewable energy sources The development of investment plans in new generation capacity using RES requires the use of modeling tools capable of accounting for the long-term evolutions of technology costs, fuel costs and demand and the short-term dynamics of balancing electricity supply and demand in order to avoid over investment and large periods of excess of RES. In this section, a hybrid modeling framework that uses an energy system planning model and an energy system balancing model is presented. The modeling framework developed is applied to two different case studies: to the island of São Miguel, Azores, that can increase the penetration of RES by using geothermal energy and wind energy and to mainland Portugal which can invest in onshore and offshore wind, large hydro and run-of-river, solar energy and wave energy. 6.1 Integrated Modeling Framework for Energy Systems Planning This work is presented in detail in Paper IV - ”Integrated Modeling Framework for Energy Systems Planning” by Carlos Silva, André Pina, Gonçalo 77 Pereira and Alexandra Moutinho published in Volume 3 of the Proceedings of the 23rd International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems [12]. 6.1.1 Aim of the study The paper proposes an integrated modeling framework for energy systems planning, where a TIMES model is used in combination with a half-anhour resolution model for the system operation developed in MATLAB, in which the technical and economical feasibility of the long-term scenarios considering the installation of renewable resources plants are tested and validated at the operational level. These results are then feed back into the TIMES model as parameters for resource availability and plant efficiency in order to calibrate scenario design. The method proposed consists of the use of both models in an iterative cycle, in which the medium-term model is used to optimize the investment in new generation capacity and the short-term model is used to calculate the energy balances over one year with higher time resolution (hourly or less) by optimizing the operational costs. The medium-term model provides to the short-term model the installed capacities of each energy source that it has to consider for the simulation of each year; the short-term model updates the medium-term model constraints parameters regarding the operation of the electricity generation facilities, such as the impossibility to start the operation on a certain year or the limitation on the amount of electricity that can be absorbed by the grid (effectively lowering the availability factor in that year). Generally, the framework works as shown in Figure 6.1. At the beginning of each iteration, a run of the medium-term model provides the inputs for the short-term model. Then, the short-term model runs for the first year being considered. After this run, the outputs are analyzed and three things may happen: (i) if the outputs show that some criteria are not met, then there are restrictions that need to be introduced in the medium-term model for that year; (ii) if the outputs show that the necessary criteria are met and the end year has not been reached, then the short-term model is run for the following year; (iii) if the outputs show that the necessary criteria are met 78 and the year being considered is the end year of the optimization, then the process comes to an end. The proposed methodology is tested in the design of the electric system of the island of São Miguel in Azores - Portugal to determine when should the investment in new geothermal capacity occur (one plant of 10 MW is expected to start operations in 2013 and another plant of 10 MW can be installed from that year onward) and what is the capacity of the wind-farms that should be installed. Figure 6.1: Proposed integrated modeling framework 6.1.2 Main results When using only TIMES, the model decides to invest on the second 10 MW geothermal unit comes into operation in the year 2013 together with the already planned 10 MW plant which results in a large increase of the production from geothermal in that year, as shown in Figure 6.2. This is due to the low cost of producing electricity from geothermal energy, which makes it economically viable for the model to install even if generation had to be stopped during the night periods, with the early investment in geothermal delaying the investments in wind energy to 2017 and after. 79 Figure 6.2: Electricity production by source from TIMES initialization The results obtained after the last iteration contrast heavily with when only TIMES was used, where the early introduction of the geothermal plant postponed the installation of wind power. After the final iteration (iteration 10), the results from the model showed that the second 10 MW geothermal plant should be installed only in 2019 and that investment in wind energy could begin by 2011. However, the maximum capacity that the model decided to install was only 9.9 MW whereas when using only TIMES the wind power installed capacity reached 23 MW by 2020. The evolution of the electricity production after the last iteration is shown in Figure 6.3. Figure 6.3: Electricity production by source after the last iteration The reduction of the installed capacity of wind power was due to the use of the short-term model to evaluate the energy balance between supply and demand and identify the excesses of electricity from wind energy that cannot be used by the system. This allowed determining the real cost-effectiveness 80 of the wind turbines and, in this way, produce a more feasible solution for the expansion of the electricity supply system. 6.2 High-resolution modeling framework for the planning of electricity systems with high penetration of renewables This work is presented in detail in Paper V - ”High-resolution modeling framework for the planning of electricity systems with high penetration of renewables” by André Pina, Carlos Silva and Paulo Ferrão to be submitted to the journal Applied Energy. 6.2.1 Aim of the study The paper presents a modeling framework that is able to optimize the investment in new renewable generation capacity on the long-term while taking into account the hourly dynamics of electricity supply and demand, by combining two models are used in an iterative cycle, with each providing input for the other, as shown in Figure 6.4. Figure 6.4: Modeling framework proposed The framework uses two of the most used energy planning tools, EnergyPLAN which is able to model the hourly dynamics of RES and TIMES which is able to optimize the investment in new generation capacity over 81 the long-term. While the investment optimization model (TIMES) provides to the operation optimization model (EnergyPLAN) the installed capacities of each energy source that it has to consider for the simulation of each year, the results from the operation optimization model are used to update the investment optimization model constraints parameters regarding the amount of new installed capacity that the system can handle for that year. The algorithm is the same as the one presented in Section 6.1 and Figure 6.1. The criteria adopted to determine if the model needs to update the constraints of the investment model consists on determining if all the installed capacity from RES is able to produce at least 90% of the expected yearly output (for example, if the expected capacity factor of wind is 25%, the criteria are met if the real capacity factor is at least 22.5%). In the case the criteria are not met, the constraints to be introduced on the installed capacity from that iteration that maximize the production from renewable energy sources and fulfils the criteria are calculated and added to the TIMES model for the year being analyzed. The framework was applied to continental Portugal for the time period of 2005-2050, in order to identify optimal investment plans in new renewable and fossil generation capacity with the goal of achieving significant CO2 emissions reduction, under different scenarios. Two different scenarios were considered in this work based on the capacity to pump water to reservoirs. The expected capacity for storage is 4302 MW by 2020 due to the installation of several reversible turbines in future investments of large hydro systems. This situation is compared to the current 1036 MW of installed capacity, with no new reversible system being installed in the future. 6.2.2 Main results When using the TIMES model only, the investment decisions of the model were the same for the two different pumped storage capacities (4302 MW and 1036 MW), with the operation decisions being slightly different due to the higher storage capacity of the 4302 MW scenario. However, when applying the complete algorithm to both scenarios the results were very different. While in the case of the 4302 MW the first result obtained from TIMES verified the criteria in every year, the same did not happen for the 1036 MW case which needed 135 iterations. This is because the model did 82 not have any information on the amount of energy that can be stored at any given period, which greatly benefits the scenario with high storage installed capacity. As such, in spite of having a higher than usual time resolution, the TIMES model is not able to fully account for the variability of RES. The results obtained for before and after the iterative process are very different in what concerns the investment in generation capacity using RES, as shown in Figure 6.5, with the model deciding to install less generation capacity using RES after the iterative process (in 2040 there was a decrease of 14.5% from 27600 MW to 23600 MW and by 2050 the decrease was of 18.8% from 32000 MW to 26000 MW). (a) (b) Figure 6.5: Installed capacity from renewable energy sources a) before the first iteration b) after the last iteration Furthermore, after the iterative process, the model decides to invest 5 years sooner in Solar energy and 13 years sooner in Wave energy while reducing the investment in Onshore and Offshore Wind energy. This is due 83 to the model being able to take into account that using different resources with complementary variability guarantees that all installed capacity from renewable energy sources is used efficiently, thereby reducing the investment in generation capacity that would have low capacity factors. The yearly penetration of RES is higher in the 4302 MW scenario (around 87% by 2050) when compared to the 1036 MW scenarios. Despite having the same installed capacity, the 1036 MW scenario before the application of the proposed methodology has a lower share of RES than the 4302 MW (84% by 2050) because of the lower storage capacity which reduces the amount of renewable energy that can be effectively stored in periods of excess. Due to the lower installed generation capacity after the application of the proposed methodology to the 1036 MW of storage capacity, the share of RES in the electricity mix is only of around 72% by 2050, as shown in Figure 6.6. (a) (b) Figure 6.6: Electricity mix in the different scenarios a) in the 1036 MW scenario, iteration 1 b) in the 1036 MW scenario, last iteration 84 The large scale deployment of electricity generation capacity from renewable energy sources can lead to excess production if the amount of electricity that can be produced from these sources is higher than demand. This excess can create significant problems in the electricity grid as well as economic problems due to low prices of electricity in the spot markets which reduces the economic competitiveness of the installed generation capacity. The excess of electricity from renewable energy sources in the different scenarios is shown in Figure 6.7 a) when considering all renewable energy sources and in Figure 6.7 b) when considering only solar, wind offshore and wave, which are less mature technologies and generally more expensive when compared to wind onshore and hydro. (a) (b) Figure 6.7: Percentage of excess electricity from renewable energy sources in the different scenarios a) for all renewable energy sources and b) considering only solar, wind offshore and wave energies 85 6.3 Key findings The key findings from the development of a hybrid modeling framework for systems with high penetrations of renewable energy sources are: • The use of a hybrid modeling methodology that combines the longterm and short-term energy modeling approaches enables the optimization of investment in new generation capacity of RES while taking into account the hourly dynamics of RES and of electricity demand. • The increase in temporal resolution allows avoiding the overinvestment in new generation capacity by identifying which investments will lead to excesses in electricity production from RES that cannot be used by the system. • The modeling framework is able to optimize investments based on the complementarities of the availability of RES which can lead to the delay of investments for some RES and an anticipation for others. • For systems with lower energy storage capacities there is a necessity to model in detail the energy systems in order to develop investment plans that consider accurate estimation of the electricity production from each RES. 86 Chapter 7 Conclusions and future work The development of new modeling methodologies that combine the resolution of very detailed models with the long-term optimization capabilities of broader models is crucial to the efficient planning of sustainable energy systems with large penetrations of RES and to support the design of efficient energy policies. 7.1 State of the art in energy systems modeling The development of new modeling methodologies, and improvement of previously existing ones, has occurred in order to build up the capability to provide answers to the challenges energy systems face. This has been seen since the beginning of the development of energy modeling tools in the 1970’s with the scope of the modeling tools following the changes that occurred in energy systems due to events with global impacts such as the energy crisis in the 1970’s, the disaster in the nuclear plant of Chernobyl in 1986 and the signing of the Kyoto Protocol in 1997. Currently, there are two main types of modeling tools: Long-term energy systems planning tools with temporal resolution of a year or less and Short-term tools of a year or less with hourly or higher temporal resolution. Long-term modeling tools are characterized as being capable of optimizing the investment in new generation capacity over a large number of years or simulating pre-determined energy systems to see how well they are able to address the yearly growth of electricity demand. However, they have low 87 temporal resolution, which does not allow them to capture the necessary dynamics for performing the optimization of operating the system. Shortterm tools are generally capable of optimizing the operation of a system, and sometimes are even able to optimize the investment in new plants, but only taking into account one year. This reflects their ability to model hourly dynamics of electricity supply and demand, which makes them suitable for analyzing how a system should be operated. The expected large-scale deployment of RES, the development of new technologies such as EVs and demand response and the introduction of large sets of DSM policies requires energy models to be able to analyze the longterm evolution of energy systems while taking into account the short-term dynamics of electricity supply and demand. Energy models need to be able to account for the changes of demand and of technology and fuel costs as well as how the hourly match between electricity demand and electricity production from RES to optimize the generation capacity mix and avoid the investment in capacity that will lead to large periods of excess of electricity from RES. With this purpose, a new set of energy modeling methodologies that link two or more specific energy modeling tools in order to produce a more detailed representation of energy systems has started to be developed in the last years. These new modeling methodologies can have different purposes according to the modeling tools they include and can be used to perform studies that combined different technical and economical aspects. 7.2 High-temporal resolution modeling of the dynamics of energy systems The increase in time resolution of the models built in TIMES is crucial to achieve better estimates for the potential of RES in electricity systems and to allow designing more robust investment plans for the evolution of electricity systems. The constraints and dynamics introduced by having higher time resolution have a significant impact on the expected electricity production from new RES generation units and also on the amount of new generation capacity that can be installed, as lower time resolutions were shown to lead 88 to overestimations of the amount of new renewable energy that could be installed and absorbed by the system. While the proposed temporal resolution for TIMES (4 seasons, 3 days per season and 24 hours per day for each year) can help improve the analysis of the investment in RES, its results still have a significant difference when compared to the modeling with hourly resolution. Analyzing the results from three different electricity demand and supply balance methods (i.e., integral, semi-dynamic and dynamic methods) with different temporal resolutions it becomes clear that the approach used in TIMES can help to identify the existence of excesses of electricity from RES but not the total amount of electricity that will not be used. The use of low temporal resolution models can then lead to the overestimation of the amount of electricity from RES that a system can use and the overinvestment in new generation capacity, which results in a reduced costeffectiveness of the system and an underestimation of the CO2 emissions to be expected from the system. The results show that, if the proper hourly dynamics are not included when modeling electricity supply and demand, the information passed on to decision makers can lead to suboptimal energy systems with reduced cost-effectiveness as some investments could be made earlier than necessary. This could also lead the country or region to miss out on a set target such as use of renewable energies or CO2 emissions caps. The relevance of using energy models with high temporal resolution increases with the amount of generation capacity from RES that is installed in the system. From the results obtained, differences in the dispatch of electricity from RES could appear when the installed capacities from RES is slightly higher than the average power of the low consumption hours. Since the variability of the resource plays an important role in the electricity available for use, differences can be hidden by when average availability factors are used, which compensate situations where the resource is not available to fulfill the demand with others when more resource is available than what is needed by the demand. When the installed capacity from RES starts to be significantly high when compared to demand, the possibility of leveling the energy surplus with the energy shortfall from the resources ceases even for energy models with lower temporal resolutions and differences start to appear. 89 The modeling of electricity demand dynamics must also be taken into account as they can have a significant impact on the optimization of the new generation capacity from RES. The use of higher temporal resolution energy system models is therefore crucial to assess the real impact of specific energy policies and changes in demand patterns in the complete energy system as it allows calculating the potential benefits and drawbacks of introducing them. 7.3 Modeling the dynamics of electricity demand The introduction of demand management strategies, be it based on technological energy efficiency, consumer behavior changes or the introduction of dynamic demand-side management technologies is crucial for the longterm sustainability of any region. These options will play a large role in the transition to sustainable energy systems by keeping demand at levels in which renewable energies can be used effectively to meet that demand. Therefore, the modeling of electricity systems needs to include the modeling of the DSM strategies that are being considered for the region, and their dynamics, as they can change significantly the optimal investment plan due to changes in the electricity demand patterns. In particular, the introduction of dynamic demand technologies can help delay the need for investment in new generation capacity by reducing peak electricity demand and enabling a better management of the already existing capacity by shifting load to periods in which there would be excess of production from RES. This benefits the system by enabling a phasing of the investment and increasing the cost-effectiveness of the installed capacities. Regarding EVs, the modeling of electricity demand dynamics enables analyzing the impact on the electricity production system of using different charging strategies. The results show that the use of flexible time of charging strategies can help increase the share of RES in the electricity produced for the charging of the EVs by providing a better match between supply and demand. This can help double the share of RES used for charging the EVs, which has a direct impact on the reductions in CO2 emissions of the transportation sector. Generally, the results indicate that charging the EVs during the night period can increase the cost-effectiveness of the EVs and increase the environmental benefits since that is where it is most likely to 90 have excess of electricity production from RES. The environmental benefits of switching from conventional transportation technologies to EVs can also change drastically from one month to the other based on the RES that are used for the production of electricity. A high temporal resolution analysis allows identifying in which periods the EVs will be charged with electricity produced from fossil fuels. This information can then be used to develop alternative strategies for these periods that enable the increase of the environmental benefits of EVs. 7.4 Hybrid modeling framework for high penetrations of renewable energy sources The development of hybrid modeling framework that uses already existing energy modeling tools allows combining the strengths of different modeling approaches and model with more detail the true complexity of electricity systems. In the particular case analyzed in this thesis, the use of a hybrid modeling methodology that combines the long-term and short-term energy modeling approaches enables the optimization of investment in new generation capacity of RES while taking into account the hourly dynamics of RES and of electricity demand. The increase in temporal resolution to include every hour of the year allows identifying which investments will lead to excesses in electricity production from RES that cannot be used by the system. With this information, the model can then optimize investments in order to avoid the installation of new generation capacity that will have low effective capacity factors. Using an iterative process, the modeling framework is also able to optimize the generation capacity mix with the goal of reducing the amount of excess electricity from RES by delaying the investment in some RES and anticipating it for others, based on the complementarities of the availability of RES. This combination of available RES allows reducing drastically the amount of electricity from RES that would not be used while lowering only slightly the penetration of RES in the electricity system. The analysis of the complementaritities of resources and the matching of supply and demand also helps the model to avoid the trivial solution of investing only in 91 the cheapest technology available, as the investment in large quantities of the same RES technology can lead to continuous decreases in the effective capacity factor of the marginal installed MW that reduce their cost-effectiveness and make them more expensive than the other technologies. Furthermore, the impact of having different energy storage capacities on the planning of electricity systems can only be modeled in detail with the use of very detailed models that allow the effective calculation of the amount of energy that needs to be stored. The results show that the application of the proposed modeling framework does not have a significant impact in case a high storage capacity exists as the amount of excess electricity from renewable energy sources is small precisely because of the possibility to store a large part of it. However, in systems with low storage capacity the application of the proposed modeling framework resulted in a significantly different generation capacity mix. 7.5 Future work The proposed methodology can also be applied to study the introduction of different energy efficiency measures, electric vehicles or demand response programs as it enables the model to take into account the simultaneous evolution of the electricity supply and demand sectors. Regarding the inclusion of demand dynamics on the proposed modeling framework, it is possible to include changes in the hourly load curves throughout the years from the evolution of different activity sectors due to economic objectives and different growth rates as well as the short-term changes in electricity demand profiles from the shifting of consumption from one hour to the other using EVs or demand response. However, the adaptation of the proposed methodology to account for the shifting of loads requires the development of yearly models with hourly resolution that are able to model electricity dispatch systems while considering EVs and demand response technologies. From the analysis performed in this thesis, these models are currently lacking. Addressing these issues will allow the design of comprehensive action plans for achieving an economically and technically feasible pathway for achieving sustainable energy systems. 92 Bibliography [1] International Energy Agency, Key World Energy Statistics, Tech. rep. (2009). [2] Europe’s Energy Portal, Energy Dependency, www.energy.eu (accessed July, 2011). [3] S. Peteves, European Strategic Energy Technology Plan - Towards a new European Energy Technology Policy, in: IEA/ETO - SET-PLAN, Paris, May 15, 2008, 2008. [4] G. Booras, Australian Electricity Generation Technology Costs Reference Case 2010, Tech. rep. (2010). [5] ETSAP - Energy Technology Systems Analysis Program - Documentation, http://www.etsap.org/documentation.asp (2009). [6] Commission of the European Communities, Second Strategic Energy Review: An EU Energy Security and Solidarity Action Plan - Energy Sources, Production Costs and Performance of Technologies for Power Generation, Heating and Transport, Tech. rep. (2008). [7] Electricity Storage Association, Technology Comparison, www.electricitystorage.org (accessed August, 2011). [8] C. W. Gellings, W. M. Smith, Integrating demand-side management into utility planning, Proceedings of the IEEE 77(6) (1989) 908–918. [9] A. Pina, C. Silva, P. Ferrao, Modeling hourly electricity dynamics for policy making in long-term scenarios, Energy Policy 39(9) (2011) 4692–4702. 93 [10] G. Haydt, V. Leal, A. Pina, C. A. Silva, The relevance of the energy resource dynamics in the mid/long-term energy planning models, Renewable Energy 36(11) (2011) 3068–3074. [11] A. Pina, C. Silva, P. Ferrao, The impact of demand side management strategies in the penetration of renewable electricity, Energy (2011),doi:10.1016/j.energy.2011.06.013. [12] C. Silva, A. Pina, G. Pereira, A. Moutinho, Integrated Modeling Framework for Energy Systems Planning, in: Volume 3 of the Proceedings of the 23rd International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, 2010, pp. 323–330. [13] F. Urban, R. M. J. Benders, H. C. Moll, Modelling energy systems for developing countries, Energy Policy 35 (2007) 3473–3482. [14] T. Nakata, Energy-economic models and the environment, Progress in Energy and Combustion Science 30 (2004) 417–475. [15] T. F. Schulz, S. Kypreos, L. Barreto, A. Wokaun, Intermediate steps towards the 2000 w society in Switzerland: An energy-economic scenario analysis, Energy Policy 36 (2008) 1303–1317. [16] F. Pietrapertosa, C. Cosmi, M. Macchiato, G. Marmo, M. Salvia, Comprehensive modelling for approaching the Kyoto targets on a local scale, Renewable and Sustainable Energy Reviews 7 (2003) 249–270. [17] A. Busuttil, G. Krajacic, N. Duic, Energy scenarios for Malta, International Journal of Hydrogen Energy 33 (2008) 4235–4246. [18] A. Bernard, M. Vielle, GEMINI-E3 , a general equilibrium model of internationalnational interactions between economy, energy and the environment, COMPUTATIONAL MANAGEMENT SCIENCE 5(3) (2007) 173–206. [19] LEAP - Long range Energy Alternatives Planning system, http://www.energycommunity.org/default.asp?action=47 (2009). [20] HOMER - Analysis of micropower https://analysis.nrel.gov/homer/ (2009). 94 systems options, [21] Sustainable Energy Planning Research Group, Aalborg University, EnergyPlan, http://energy.plan.aau.dk/ (1999). [22] M. D. Ilic, J. Jhi-Young, X. Le, M. Prica, N. Rotering, A DecisionMaking Framework and Simulator for Sustainable Electric Energy Systems, IEEE Transactions on Sustainable Energy 2(1) (2011) 37–49. [23] A. Z. Morch, Software tools for energy planning: Overview and comparison, Tech. rep., SINTEF Energy Research (2005). [24] J. B. Greenblatt, S. Succar, D. C. Denkenberger, R. H. Williams, R. H. Socolow, Baseload wind energy: Modeling the competition between gas turbines and compressed air energy storage for supplemental generation, Energy Policy 35 (2007) 1474–1492. [25] J. M. Carrasco, L. G. Franquelo, J. T. Bialasiewicz, E. Galvn, R. C. P. Guisado, M. A. M. Prats, J. I. Len, N. Moreno-Alfonso, Powerelectronic systems for the grid integration of renewable energy sources: A survey, IEEE Transactions on Industrial Electronics 53 (4) (2006) 1002–1016. [26] H. Bozic, Energy system planning analysis using the integrated energy and macroeconomy model, Interdisciplinary Description of Complex Systems 5 (1) (2007) 39–47. [27] A. Lehtila, I. Savolainen, S. Syri, The role of technology development in greenhouse gas emissions reduction: The case of Finland, Energy 30 (2005) 2738–2758. [28] D. Henning, S. Amiri, K. Holmgren, Modeling and optimization of electricity, steam and district heating production for a local Swedish utility, European Journal of Operational Research 175 (2006) 1224– 1247. [29] N. J. Schenk, H. C. Moll, A. J. M. S. Uiterkamp, Meso-level analysis, the missing link in energy strategies, Energy Policy 35 (2006) 1505– 1516. [30] J.-M. Beaujean, J.-P. Charpentier, A review of energy models no. 4, Tech. rep., International Institute for Applied Systems Analysis (1978). 95 [31] S. Jebaraja, S. Iniyan, A review of energy models, Renewable and Sustainable Energy Reviews 10(4) (2006) 281–311. [32] D. Connolly, H. Lund, B. V. Mathiesen, M. Leahy, A review of computer tools for analysing the integration of renewable into various energy systems, Applied Energy 87 (2010) 1059–1082. [33] O. Amerighi, U. Ciorba, M. C. Tommasino, D2.1 models characterization report, Tech. rep., Analysing Transition Planning and Systemic Energy Planning Tools for the implementation of the Energy Technology Information System (2010). [34] J. Rosen, I. Tietze-Stockinger, O. Rentz, Model-based analysis of effects from large-scale wind power production, Energy 32 (2007) 575– 583. [35] Y. Matsuoka, M. Kainuma, T. Morita, Scenario analysis of global warming using the Asian Pacific Integrated Model (AIM), Energy Policy 23 (1995) 357–371. [36] K. Karlsson, P. Meibom, Optimal investment paths for future renewable based energy systems Using the optimisation model Balmorel, International Journal of Hydrogen Energy 33 (2008) 1777–1787. [37] Oak Ridge National Laboratory, Whole-Building and Community Integration, http://www.coolingheatingpower.org/about/bchpscreening-tool. [38] M. Chaudry, N. Jenkins, G. Strbac, Multi-time period combined gas and electricity network optimisation, Electric Power Systems Research 78 (2008) 1265–1279. [39] C. R. Hudson, ORNL CHP Capacity Optimizer User’s Manual, Tech. rep., Oak Ridge National Laboratory (2005). [40] A. Aldi, K. Anundson, A. Bigelow, A. Capulli, Decentralized Energy Master Planning, Tech. rep., Worcester Polytechnic Institute (2010). [41] Energy research Centre of the Netherlands, COMPETES, http://www.ecn.nl/units/ps/tools/modelling-systems/competes/. 96 [42] Aalborg University, EnergyInteractive.NET, http://energyinteractive.net. [43] I. Otero-Novas, C. Meseguer, C. Batlle, J. J. Alba, A Simulation Model for a Competitive Generation Market, IEEE TRANSACTLONS ON POWER SYSTEMS 15-1 (2000) 250–256. [44] T. Homma, S. Mori, K. Akimoto, H. Yamamoto, T. Kosugi, T. Tomoda, Development of multi-regional and multi-sectoral energyeconomic model and the analysis of CO2 emission reduction, in: 8th Annual Conference on Global Economic Analysis, Lbeck, Germany, 2005. [45] DER-CAM - Distributed Energy Resources Customer Adoption Model, http://der.lbl.gov/dercam.html (2009). [46] EWI, kets Dispatch in Europe and - Investment A brief Model overview, for Electricity Mar- http://www.ewi.uni- koeln.de/fileadmin/user/PDFs/DIME Model description .pdf. [47] K. Akimoto, T. Tomoda, Y. Fujii, K. Yamaki, Assessment of global warming mitigation options with integrated assessment model DNE21, Energy Economics 26-4 (2004) 635–653. [48] D. of Trade, Industry, Energy Paper 68, Energy Projections for the UK, Tech. rep. (2000). [49] D. J. Swider, C. Weber, R. Barth, The Value of Wind Energy in the European Electricity Market Application of a Stochastic Fundamental Model, in: 2004 European Wind Energy Conference & Exhibition, 2004. [50] H. Pollitt, T. Barker, et al., E3ME Manual, Tech. rep., Cambridge Econometrics. [51] H. M. Pant, GTEM - global trade and environment model, Tech. rep., Commonwealth of Australia (2007). [52] Y. Porat, R. Irith, R. Turvey, Long-run marginal electricity generation costs in Israel, Energy Policy 25-4 (1997) 401–411. 97 [53] M. R. Milligan, Modeling Utility-Scale Wind Power Plants - Part 1: Economics, Tech. rep., National Renewable Energy Laboratory (2000). [54] Decision and Information Sciences Division - Center for Energy, Environmental, and Economic Systems Analysis, Electricity Market Complex Adpative System (EMCAS) - Model Introduction, Tech. rep., Argonne National Laboratory (2008). [55] W. Lise, V. Linderhof, O. Kuik, C. Kemfert, R. stling, T. Heinzow, A game theoretic model of the Northwestern European electricity marketmarket power and the environment, Energy Policy 34(15) (2006) 2123–2136. [56] R. Segurado, S. Pereira, A. Pipio, L. Alves, Comparison between EMINENT and other Energy Technology Assessment Tools, in: EMINENT 2 Workshop Energy for Sustainable Future, 2008. [57] A. Helseth, Thermal B. Mo, Power Long-Term Systems Scheduling Using of Scenario HydroTrees, www.enercord.com/PSM2009/PSM 2009 Helseth.ppt. [58] California Environmental Protection Agency, Updated Economic Analysis of California’s Climate Change Scoping Plan, Tech. rep. (2010). [59] EMD International A/S, energyPRO - User’s Guide, Tech. rep. (2011). [60] ENPEP - ENergy and Power Evaluation Program, http://manhaz.cyf.gov.pl/manhaz/links/Environmental analysis tools /Environmental analysis tools.html (2009). [61] L. Barreto, Technological Learning in Energy Optimisation Models and Deployment of Emerging Technologies, Ph.D. thesis, Swiss Federal Institute of Technology Zurich (2001). [62] D. Belcredi, E. G. Bon, P. Bresesti, G. Bruno, R. Calisti, M. V. Cazzol, et al., Un ambiente integrato per la pianificazione strategica dei sistemi di trasmissione, Tech. rep., CESI RICERCA (2008). [63] R. F. Naill, A system dynamics model for national energy policy planning, System Dynamics Review 8(1) (1992) 1–19. 98 [64] M. Haller, Comparing CO2 Mitigation Options in the Electricity Sector: Nuclear Power, Renewable Energy and Carbon Sequestration, Ph.D. thesis, Technische Universitat Berlin (2006). [65] Argonne National transmission Laboratory, maximization Generation (GTMax) and model, http://www.dis.anl.gov/projects/Gtmax.html. [66] D. M. Steward, M. Penev, The Influence of Building Location on Combined Heat and Power/ Hydrogen (Tri-Generation) System Cost, Hydrogen Output and Efficiency (2009). [67] R. Martins, G. Krajacic, L. Alves, N. Duic, T. Azevedo, M. da Graça Carvalho, Energy storage in islands - modelling porto santo’s hydrogen system, Chemical Engineering Transactions 18 (2009) 367–372. [68] U.S. Department of Energy, Oak Ridge National Laboratory, HUD CHP Guide #2: Feasibility Screening for Combined Heat and Power in Multifamily Housing, Tech. rep. (2009). [69] N. Briguglio, M. Ferraro, L. Andaloro, V. Antonucci, New simulation tool helping a feasibility study for renewable hydrogen bus fleet in Messina, International Journal of Hydrogen Energy 33 (2008) 3077– 3084. [70] B. James, HyPro: Modeling the H2 Transition (2007). [71] D. Martinsen, J. Linssen, P. Markewitz, S. Vogele, CCS: A future CO2 mitigation option for Germany? - A bottom-up approach, Energy Policy 35 (2007) 2110–2120. [72] B. J. de Vries, D. P. van Vuuren, M. G. den Elzen, M. A. Janssen, The Targets IMage Energy Regional (TIMER) Model - Techincal Documentation, Tech. rep., The IMAGE Project (2001). [73] International sion 2050: Network visions for for Sustainable a renewable Energy, energy Viworld, http://www.inforse.dk/europe/Vision2050.htm. [74] Vienna University of Technology, INVERT, http://www.invert.at/. 99 [75] ICF International, Introduction to the Integrated Planning Model (IPM), Tech. rep. (2008). [76] R. Loulou, G. Goldstein, K. Noble, Documentation for the MARKAL Family of Models (2004). [77] M. Lisboa, L. Marzano, C. Saboia, M. Maceira, A. Melo, A MixedInteger Programming Model for Long Term Generation Expansion Planning of the Brazilian System, in: 16th PSCC, Glasgow, Scotland, July 14-18, 2008, 2008. [78] A. Susandi, Integrated Assessment Model for Indonesian Energy Forecast, in: Joint Convention Jakarta IAGI IAGI, 15 17 December 2003, Jakarta - Indonesia, 2003. [79] W. Krewitt, S. Simon, W. Graus, S. Teske, A. Zervos, O. Schfer, The 2C scenario A sustainable world energy perspective, Energy Policy 35(10) (2007) 4969–4980. [80] L. Schrattenholzer, The Energy Supply Model Message, Tech. rep., International Institute for Applied Systems Analysis (1981). [81] M. J. Scott, R. D. Sands, J. Edmonds, A. M. Liebetrau, D. W. Engel, Uncertainty in integrated assessment models: modeling with MiniCAM 1.0, Energy Policy 27(14) (1999) 855–879. [82] ENEA Ricerca sul Sistema Elettrico, SECURE - Assessment of the impact of gas shortages risks on the power sector, Tech. rep. (2009). [83] Deloitte, Deloitte Marketpoint - MarketBuilder Models and Data, Tech. rep. [84] NEMS - The National Energy Modeling System, http://www.eia.doe.gov/oiaf/aeo/overview/ (2009). [85] M. Maceira, L. Terry, F. Costa, J. Damázio, A. Melo, Chain of Optimization Models for Setting the Energy Dispatch and Spot Price in the Brazilian System, in: 14th PSCC, Sevilla, 24-28 June 2002, 2002. [86] S. Hadley, E. Hirst, ORCED: A Model to Simulate the Operations and Costs of Bulk-Power Markets, Tech. rep., Oak Ridge National Laboratory (1998). 100 [87] Nuclear and Electricity Analysis Branch, Energy Supply and Conversion Division, Office of Integrated Analysis and Forecasting, Energy Information Administration, Model Documentation - Electricity Capacity Planning Submodule of the Electricity Market Module, Tech. rep. (1994). [88] Iowa State University, PLEXOS Overview & Tutorial, Tech. rep. [89] P. Criqui, S. Mima, L. Viguier, Marginal abatement costs of CO2 emission reductions, geographical flexibility and concrete ceilings: an assessment using the POLES model, Energy Policy 27(10) (1999) 585– 601. [90] Energy research Centre of the Netherlands, POWERS, http://www.ecn.nl/units/ps/tools/modelling-systems/powers/. [91] National Technical University of Athens, Energyeconomicsen- vironment modelling laboratory research and policy analysis, http://www.e3mlab.ntua.gr/. [92] A. Gjelsvik, M. Belsnes, A. Haugstad, An algorithm for stochastic medium-term hydrothermal scheduling under spot price uncertainty, in: Proceedings of the 13th power systems computation conference. Trondheim, Norway, 28 June2 July; 1999, 1999. [93] Energy, Economics and Environment Modeling Laboratory, PROMETHEUS stochastic model, Tech. rep., National Technical University of Athens. [94] B. Hobbs, H. Rouse, D. Hoog, Measuring the economic value of demand-side and supply resources in integrated resource planning models, IEEE Transactions on Power Systems 8. [95] Danish Energy Agency, RAMSES, http://www.ens.dk/da-DK/Info/ TalOgKort/Fremskrivninger/modeller/ramses/Sider/Forside.aspx. [96] Australian Commonwealth Department of Industry, Tourism & Resources, Cogeneration Ready Reckoner - Software to assist with an initial evaluation of the viability of cogeneration, Tech. rep. (2002). 101 [97] CANMET Energy Technology Centre - Varennes, Solar Air Heating Project Analysis, in: CLEAN ENERGY PROJECT ANALYSIS: RETScreen Engineering & Cases Textbook. [98] Instituto de Investigación Tecnológica - Universidad Pontificia Comillas, ROM Model (Reliability and Operation Model for Renewable Energy Sources), http://www.iit.upcomillas.es/aramos/ROM.htm. [99] U.S. Energy Information Administration, Regional Short-Term Energy Model (RSTEM) Overview, Tech. rep. [100] R. Siemons, M. Vis, D. van den Berg, I. M. Chesney, M. Whiteley, N. Nikolaou, Bio-Energy’s Role in the EU Energy Market - A view of developments until 2020, Tech. rep., BTG biomass technology group BV (2004). [101] Energy Information Administration (US-DOE), System for the Analysis of Global Energy markets (SAGE), http://www.eia.doe.gov/. [102] K. S. Hornnes, A Model for Coordinated Utilization of Production and Transmission Facilities in a Power System Dominated by Hydropower, Ph.D. thesis, University of Trondheim (1995). [103] K. Illum, A SESAM Model of the Nordic Energy System - Methodology and the Modelling of the Nordic Energy System, Tech. rep., Greenpeace Nordic (2006). [104] Institute for Sustainable Solutions and Innovations, Energy rich Japan, http://www.energyrichjapan.info/. [105] Energinet.dk, Sivael, http://www.energinet.dk/en/. [106] W. B. Powell, A. George, H. Simao, W. Scott, A. Lamonty, J. Stewart, SMART: A Stochastic Multiscale Model for the Analysis of Energy Resources, Technology and Policy, Tech. rep. (2010). [107] B. Jankowski, W. Pellekaan, J. Winter, Volume III: A comparison of the energy models DORSEK, ENPEP and EFOM, in: The Netherlands - Poland Task Force on Integrated Energy and Environmental Planning, 1993. 102 [108] Ea Energy Analyses, The STREAM modelling tool, http://www.streammodel.org/. [109] M. G. Hoffman, A. Sadovsky, M. C. Kintner-Meyer, J. G. DeSteese, Analysis Tools for Sizing and Placement of Energy Storage in Grid Applications - A Literature Review, Tech. rep., Pacific Northwest National Laboratory (2010). [110] University of Wisconsin-Madison, A TRaNsient SYtems Simulation program, http://sel.me.wisc.edu/trnsys/. [111] J. D. Leaver, K. T. Gillingham, L. H. T. Leaver, Assessment of primary impacts of a hydrogen economy in New Zealand using UniSyD, International Journal of Hydrogen Energy 34(7) (2009) 2855–2865. [112] A. G. Kagiannasn, T. Didisz, D. T. Askounis, J. Psarras, Strategic appraisal of energy models for Mozambique, International Journal of Energy Research 27 (2003) 173–186. [113] Y. Cai, G. Huang, Q. Lin, X. Nie, Q. Tan, An optimization-modelbased interactive decision support system for regional energy management systems planning under uncertainty, Expert Systems with Applications 36 (2009) 3470–3482. [114] VTT Technical Research Centre of Finland, VTT-EMM, http://www.vtt.fi/?lang=en. [115] International Energy Agency, World Energy Model, http://www.worldenergyoutlook.org/model.asp. [116] Risoe National Laboratory, WILMAR: wind power integrated in liberalised electricity markets, http://www.wilmar.risoe.dk/. [117] Fondaziona Eni Enrico Mattei, WITCH, http://www.witchmodel.org/pag/model.html. [118] ETSAP ysis Program Energy Technology - Models Systems & Anal- Applications, http://www.etsap.org/Models&applicationsMainPage.asp (2009). 103 [119] R. Loulou, M. Labriet, ETSAP-TIAM: the TIMES integrated assessment model Part I: Model structure, Computational Management Science 5 (2008) 7–40. [120] M. Blesl, A. Das, U. Fahl, U. Remme, Role of energy efficiency standards in reducing CO 2 emissions in germany: An assessment with TIMES, Energy Policy 35 (2007) 772–785. [121] T. Alfstad, Development of a least cost energy supply model for the SADC region, Master’s thesis, University of Cape Town (2005). [122] Y. Lechon, H. Cabal, M. Varela, R. Saez, C. Eherer, M. Baumann, J. Dueweke, T. Hamacher, G. Toscano, A global energy model with fusion, Fusion Engineering and Design 75-79 (2005) 1141–1144. [123] M. I. Howells, T. Alfstad, D. G. Victor, G. Goldstein, U. Remme, A model of household energy services in a low-income rural african village, Energy Policy 33 (2005) 1833–1851. [124] K. Vaillancourt, M. Labriet, R. Loulou, J.-P. Waaub, The role of nuclear energy in long-term climate scenarios: an analysis with the World-TIMES model, Energy Policy 36 (2008) 2296–2307. [125] A. Kanudia, K. Vaillancourt, R. Loulou, G. Tosato, D. V. Regemorter, Long-term energy-emission scenarios with the World-TIMES, in: International Energy Workshop, 2004. [126] J. Cleto, S. S. oes, P. Fortes, J. Seixas, Renewable Energy Sources Availability under Climate Change Scenarios Impacts on the Portuguese Energy System, in: Proceedings of 5th International Conference on European Electricity Market, 2008. [127] J. Cleto, Climate Change Impacts on Portuguese Energy System in 2050, Master’s thesis, Universidade Nova de Lisboa (2008). [128] A. Fidje, E. Rosenberg, K. B. Lindberg, Regional TIMES model for Norway with high time resolution, in: International Energy Workshop, 2009. [129] L. Drouet, M. Labriet, R. Loulou, M. Vielle, A master program that will drive the coupling of GEMINI-E3 and MARKAL TIMES models (2008). 104 [130] National Renewable Energy Laboratory, Getting Started Guide for HOMER Version 2.1 (2005). [131] Applications of HOMER, http://nrelpubs.nrel.gov/Webtop/ws/nich/ www/public/ResultSet?upp=0&rpp=25&w=NATIVE(’KEYWORD2 +ph+words+”HOMER”’)&order=native(’pubyear/Descend’) (2009). [132] Applications of DER-CAM, http://der.lbl.gov/publications.html (2009). [133] H. Lund, Large-scale integration of optimal combinations of PV, wind and wave power into the electricity supply, Renewable Energy 31 (2006) 503–515. [134] H. Lund, E. Munster, Management of surplus electricity-production from a fluctuating renewable-energy source, Applied Energy 76 (2003) 65–74. [135] M. Blarke, H. Lund, The effectiveness of storage and relocation options in renewable energy systems, Renewable Energy 33 (2008) 1499–1507. [136] Encyclopedia dicators - of the Nations, Environment - World Energy Development production & Inuse, http://www.nationsencyclopedia.com (accessed: 2011). [137] E. A. Cherniavsky, Oil/gas supply modeling considerations in longrange forecasting. [Use of LEAP: Long-Range Energy Analysis Program], in: Symposium on oil and gas supply modeling, Washington, DC, USA, 18 Jun 1980, 1980. [138] K. Leimkhler, G. Egberts, OPTIMIZATION TECHNIQUES - Lecture Notes in Control and Information Sciences, Ch. The energy economics of the United Kingdom, the federal Republic of Germany, and Belgium. [139] O. B. Falls, A survey of the market for nuclear power in developing countries, Energy Policy 1(3) (1973) 225–242. [140] B. P. Hamilton, P. F. S. R. R. Cirillo, W. A. Buehring, Evaluating the environmental impacts of the energy system: The ENPEP (ENergy and Power Evaluation Program) approach, in: IAEA (International 105 Atomic Energy Agency) advisory group meeting on experience with using the agency’s model WASP for energy and nuclear power planning in developing countries, Vienna (Austria), 21-25 May 1990, 1990. [141] D. Berry, Least-cost planning: A ready tool for protecting air quality values, The Electricity Journal 3(4) (1990) 14–27. [142] A. Manne, R. Richels, On stabilizing CO2 concentrations cost- effective emission reduction strategies, Environmental Modeling & Assessment 2(4) (1997) 251–265. [143] S. Swaminathan, W. T. Flynn, R. K. Sen, Modeling of Battery Energy Storage in the National Energy Modeling System (1997). [144] C. Marnay, R. C. Richey, S. A. Mahler, S. E. Bretz, R. J. Markel, Estimating the Environmental and Economic Effects of Widespread Residential PV Adoption Using GIS and NEMS, Tech. rep., Lawrence Berkeley National Laboratory (1997). [145] L. Barreto, S. Kypreos, A post-Kyoto analysis with the ERIS model prototype, International Journal of Global Energy Issues 14 (2000) 262–280. [146] D. J. Lew, D. Corbus, R. Holz, L. T. Flowers, J. A. McAllister, Analysis of Village Hybrid Systems in Chile, Tech. rep., National Renewable Energy Laboratory (1996). [147] M. R. Milligan, A. H. Miller, The value of windpower: An investigation using a qualified production cost model, in: Windpower ‘93,San Francisco, CA (United States),12-16 Jul 1993, 1993. [148] C. Marnay, S. Kito, D. Kirshner, O. Sezgen, S. Pickle, K. Schumacher, et al., Restructuring and Renewable Energy Developments in California: Using Elfin to Simulate the Future California Power Market, Tech. rep., Lawrence Berkeley National Laboratory (1998). [149] S. W. Hadley, The impact of carbon taxes or allowances on the electric generation market in the Ohio and ECAR Region, Tech. rep., Oak Ridge National Laboratory (1998). 106 [150] M. B. Blarke, The missing link in sustainable energy: techno-economic consequences of large-scale heat pumps in distributed generation in favour of a domestic integration strategy for sustainable energy, Ph.D. thesis, Department of Development and Planning, Aalborg University (2008). [151] M. B. Blarke, H. Lund, The effectiveness of storage and relocation options in renewable energy systems, Renewable Energy 33(7) (2008) 1499–1507. [152] P. Meibom, C. Weber, R. Barth, H. Brand, Operational costs induced by fluctuating wind power production in Germany and Scandinavia, Renewable Power Generation, IET 3(1) (2009) 75–83. [153] J.-C. Altamirano, L. Drouet, A. Sceia, P. Thalmann, M. Vielle, Coupling GEMINI-E3 and MARKAL-CHRES to Simulate Swiss Climate Policies, Tech. rep., École Polytechnique Fédérale de Lausanne (2008). [154] M. Labriet, R. Loulou, M. Vielle, L. Drouet, Coupling TIAM and GEMINI-E3, in: ETSAP meeting, Sophia-Antipolis, December 17, 2008, 2008. [155] L. Qiang, J. Kejun, Low Carbon Scenario up to 2050 for China, in: 18th Asia-Pacific Seminar on Climate Change, 2-3 March 2009, Hanoi, Vietnman, 2009. [156] Altos Management Partners Inc., The Altos Integrated Market Model Suite, Tech. rep. (2009). [157] A. eling Garg, D. System: Ghosh, P. Energy R. and Shukla, Integrated Environment ModPolicies, http://www.pnl.gov/aisu/pubs/eemw/papers.htm (2001). [158] J. Rosen, The future role of renewable energy sources in European electricity supply - A model-based analysis for the EU-15, Ph.D. thesis, Universitatsverlag Karlsruhe (2007). [159] D. Most, W. Fichtner, Renewable energy sources in European energy supply and interactions with emission trading, Energy Policy 38 (2010) 2898–2910. 107 [160] F. Teng, A. Gu, M. Duan, Energy Models in China - A Literature Survey, Tech. rep., Global Climate Change Institute, Tsinghua University (2007). [161] J. Kejun, H. Xiulian, Energy Demand and Emissions in Building in China: Scenarios and Policy Options, in: ICEBO2006, Shenzhen, China, 2006. [162] J. Kejun, H. Xiulian, L. Qiang, Z. Songli, Z. Xing, Modeling Development and Emission Scenario Analysis in China, Tech. rep. [163] A. Fidje, Towards a Norwegian regional TIMES model, in: ETSAP Meeting, Nice, France, 15th December 2008. [164] DGGE, Consumos de Energia Eléctrica por Concelho e por Sector de Actividade, http://www.dgge.pt?cr=10171 (2009). [165] EDA, Caraterização das redes de transporte e distribuição de energia eléctrica da Região Autónoma dos Açores (2005, 2006, 2007, 2008). [166] Serviço Regional de Estatı́stica dos Açores, Statistical Yearbook of the Azores Region 2008, http://estatistica.azores.gov.pt/ (2009). [167] N. Hamsic, A. Schmelter, A. Mohd, E. Ortjohann, E. Schultze, A. Tuckey, et al., Increasing Renewable Energy Penetration in Isolated Grids Using a Flywheel Energy Storage System, pOWERENG, April 12-14, Setúbal, Portugal (2007). [168] Electricidade dos Açores, Caracterização das redes de transporte e distribuição de energia eléctrica na Região Autónoma dos Açores 2008, http://www.eda.pt/ (2009). [169] Rede Eléctrica Nacional, REN - Monthly Statistics, http://www.centrodeinformacao.ren.pt (2010). [170] República Portuguesa, Plano Nacional de Acção para as Energias Renováveis ao Abrigo da Directiva 2009/28/CE, Tech. rep. (2010). [171] EDA, Informação Estatı́stica (2005, 2006, 2007, 2008). [172] Quercus, ECOCASA, www.ecocasa.org (2009). 108 [173] Instituto Nacional de Estatı́stica, Inquérito á mobilidade da população residente (2000). 109