Energy-Economic. Alan H. Sanstad. University of Chicago Computation Institute. APS Short Course: Physics of Sustainable Energy June 17, 2016

Transcription

1 Energy-Economic Policy Modeling Alan H. Sanstad Lawrence Berkeley National Laboratory University of Chicago Computation Institute APS Short Course: Physics of Sustainable Energy June 17, 2016

2 Outline Terminology Economics in energy policy Theoretical foundations Examples of models and applications Uncertainty and other epistemological issues

3 Example of NEMS case (USEIA 2014b)

4 Terminology Energy policy model (energy model): A computational model based upon the economic principles i of optimization and equilibrium, representing: An energy system e.g., the electric power system in the Western U. S. An entire (regional, national, global) economy with emphasis i.e., greater detail - on the energy system Integrated assessment (IA) model: A coupled modeling system, linking energy-economic with (reduced-form) d climate (and in some cases other environmental) models

5 These models are dynamic, or intertemporal representing the system or economy over horizons ranging from hours (or less) up to decades (or more) This lecture will focus on those with multi-decadal horizons Why policy models? Because they are normative - Used to numerically study the hypothetical effects of potential energy and environmental policies, and to identify better and worse policy alternatives

6 Q: Why is economics important for energy policy? A: Because Behavior choices by consumers and firms Prices Market dynamics and interactions are first-order often if not always primary drivers of energy production and use, including energy technology innovation, development, and deployment

7 Why economics? Cont. Examples: Physical limits of fossil fuel resources These necessarily exist but the estimated magnitudes keep changing increasing - as a result of exploration and discovery efforts that are to a significant degree driven by economic factors E.g., the oil crisis of the 1970s, and its end in the 1980s The rebound effect in energy efficiency Indirect reductions in the costs of energy services resulting from efficiency-promoting policies and programs, which offset energy savings estimated t by purely physics/engineering i i methods (based on 1 st Law of Thermodynamics) (Gillingham et al. 2016)

8 Basic economic principles Rational behavior, represented mathematically as constrained optimization: Utility or profit maximization, cost minimization (including life-cycle cost minimization) These assumptions are best interpreted as justifiable and extremely useful approximations Market equilibrium: Supplies = demands In the present context, this is more an empirical fact In practice, these principles are very powerful and flexible tools for building theories and models and for understanding the potential and achieved effects of policies They do not, as such, imply that government policies or regulations are unjustifiable or unwarranted. On the contrary, standard economic neo-classical theory provides a rigorous way of understanding the circumstances in which policies can improve market outcomes (welfare economics)

9 Underlying theory Energy/IA models are, with some exceptions, very large, detailed, and complicated However, they are mostly based upon fairly straightforward mathematical frameworks - generally one of the following types (which incorporate the behavioral and equilibrium assumptions) Nonlinear systems of algebraic equations: F : R R R n m n Given (with differentiability, m convexity, etc. assumptions) and R, find such that F x, 0 x R n

10 A variation on the previous: Mixed, or Nonlinear, complementarity models Linear programs (LPs): n m m n c R, b R, A R, x R Given find to max cx t st.. Ax b n

11 Non-linear programs (NLPs): G R n f : R R : Given and, find n to x R max f s.. t G x x 0 n R m

12 Optimal control: x 0 Given, max u t st.. x f x, u t t t t 0 1 g x, u t 1 t t

13 Computational model types Relating these abstract schema to models in practice: Major economic types are Partial equilibrium: Represent systems of linked energy markets, but not the complete economy Non-linear systems of algebraic equations General equilibrium: Represent all energy and other markets in an economy as well as the optimizing agents, and the interactions in other words, the entire economy Computable General Equilibrium (CGE)

14 Types, cont. CGE: Non-linear systems, non-linear complementarity problems, non-linear programs Energy especially, electric power system - models: Linear and non-linear programs

15 Partial equilibrium Supply = Demand as a function of price:

16 General equilibrium schematic

17 Methodological aspects Most energy and IA models are deterministic and highdimensional Deterministic means neither uncertainty on the part of the modeler nor on the part of the agents in the model is explicitly represented With few exceptions, they are calibrated and parameterized by means other than statistical or econometric Their modal application is the projection of energy/ economic / environmental variables several decades or a century (or longer) into the future, in part as a function of policies, to analyze the potential consequences of the latter These deterministic projections typically take the form of scenarios: Sets of assumptions on key inputs

18 Current modeling landscape Energy and IA modeling has become the predominant analytical methodology in energy/ghg policy and regulation The models are now much more than tools: They serve to define the universe of discourse, and to determine what questions can be asked, what form answers can take, and what constitutes useful data to serve as model inputs Used by Federal agencies Dept. of Energy, Environmental Protection Agency e.g., for new GHG rules - and others State regulators energy commissions and others Operational authorities e.g., independent system operators in the electric power grid Academic and national laboratory researchers Advocacy groups

19 Example: NEMS The National Energy Modeling System (NEMS) Coupled partial equilibrium i and linear programming models, with exogenous macroeconomic inputs Created and maintained by Energy Information Administration i ti (EIA) Based in part on NRC recommendations (NRC 1992) NEMS is essentially the official national energy model of the U. S. federal government Projects U. S. energy supplies and demands (currently) to 2040, with time steps of 1 year Primary product is the Annual Energy Outlook report. Also used for specific analyses at request of Congress

20 Defining energy technology costs The basic behavioral assumption in NEMS is life-cycle cost minimization How much does it cost to purchase and operate some unit of energy-producing or energy-using equipment over a given period of time? Electric power plants End-use energy devices such as appliances The fundamental metric is life-cycle cost The basic elements let CC be how much the unit costs i.e., to build, purchase, install the capital cost OC t be the operating cost during time-period t typically, one-year This includes fuel costs e.g., of coal or natural gas for a power plant, or electricity for an appliance Also maintenance and possibly labor costs

21 Life-cycle cost Note that the operating cost is a function of hours of operation and/or annual electricity output T be the lifetime of the unit r be a discount rate consider this the rate paid on a loan of amount CC to pay for the unit Then LCC CC T t 1 1 OC t 1 r t

22 Structure of NEMS (USEIA 2014a)

23 The AEO is organized around a Reference Case As a matter of policy, EIA includes in the Reference Case only policies and regulations that are either already in place, or have been approved/ authorized e.g., by Congress and are certain to be implemented EIA also reports Low & High Economic Growth and EIA also reports Low & High Economic Growth, and Low & High Oil Price variations on the Reference Side Cases with different, e.g., technology assumptions

24 Examples of NEMS cases (USEIA 2014b)

25 NEMS cases, cont.

26 NEMS cases, cont.

27 NEMS cases, cont.

28 Integrated assessment (IA) model examples The Global Change Assessment Model (GCAM) Integrated assessment type, built around partial equilibrium global energy-economic model GCAM created, maintained, and run by Pacific Northwest National Laboratory, Joint Global Change Research Center (with U. of Maryland) Projects global energy, economic, climate, and ecosystems to 2100 (or beyond), in 5-year time steps

29 MiniCAM integrated assessment model, energyeconomic sub-model schematic (Kim et al. 2006) World Data Region supply supply get price Resource Supply Sector Demand Sector Sub-Resource Sub-Sector Sub-Sector input price input price input price Grades Technology Technology Market: Price Supply Demand Solution Algorithm ED = 0 GHG GHG demand set price demand

30 Examples, cont. The Integrated Global System Model (IGSM) Integrated assessment type, built around computable general equilibrium global economic model: Emissions Prediction and Policy Analysis (EPPA) Based on Organisation for Economic Co-operation and Development (OECD) General Equilibrium Environmental (GREEN) CGE model, circa early 1990s IGSM created, maintained and run by the MIT Joint Program on the Science and Policy of Global Change Projects global energy, economic, climate, and ecosystems to 2100, in 5-year time steps

31 EPPA schematic (Paltsev et al. 2005)

32 IGSM schematic (Paltsev et al. 2005)

33 DICE model Dynamic Integrated model of Climate & Economy Created by W. Nordhaus of Yale University; first version circa 1990 Underlying theory: Optimal control The single most influential, and used or adapted - as well as criticized - IA model Publicly available documentation and code; relatively simple structure esp., not too big (defined by ~20 equations) Straightforward to change assumptions/ parameters

34 Equations of DICE model (Nordhaus 2008)

35 DICE, cont.: Emissions

36 DICE, cont.: Climate

37 The Energy Modeling Forum (EMF) at Stanford University Created in the 1970s Pioneered multi-model inter-comparison and structured scenario analysis Very influential - In effect, the global intellectual center for energy and IA modeling Recently completed study #27, on technology and GHG emissions abatement 37

38 Energy Modeling Forum (EMF) 27: Global Technology and Climate Policy Strategies Overall aim: Understand how technology characteristics cost, availability, carbon output, etc. affect achievement of global CO2 reduction goals Subject of special issue of Climatic Change, April 2014 Both partial equilibrium (PE) and general equilibrium (GE) models (modelers) participated Policies analyzed included: 550 ppm CO2 concentration 450 ppm CO2 concentration 38

39 The models incorporate representations of a range of energy technologies Fossil fuel based, including carbon capture and sequestration (CCS) Wind and solar Nuclear Biomass Plus exogenous energy intensity-reducing technological change decline in ratio of economic output to energy input through unspecified mechanisms Basic design for scenarios: Examine the feasibility and cost Basic design for scenarios: Examine the feasibility and cost implications of limiting low-carbon technologies

40 Scenario feasibility under different assumptions (Krey et al. 2014) 40

41 EMF 27 cost estimates (Kriegler et al. 2014)

42 EMF 27: Global abatement costs 5% discount rate (Kriegler et al. 2014) 42

43 Interpreting and evaluating the models and the analyses Epistemology (from Oxford English Dictionary): The theory of knowledge and understanding, esp. with regard to its methods, validity, and scope, and the distinction between justified belief and opinion; [and/or] a particular theory of knowledge and understanding. 43

44 EIA view on NEMS epistemology Projections by EIA are not statements of what will happen but of what might happen, given the assumptions used for any particular scenario energy models are simplified representations of [the energy system]. Projections are highly dependent on the data, methodologies, model structures, and assumptions used in their development. Behavioral characteristics are indicative of real-world tendencies rather than representations of specific outcomes [These] projections are subject to much uncertainty [some of which is addressed via side cases] (USEIA 2014b).

45 The meaning of scenarios Scenarios are alternative images of how the future might unfold and are an appropriate tool with which to analyze how driving forces may influence future emissions outcomes and to assess the associated uncertainties The possibility that any single emissions path will occur as described in the scenarios is highly uncertain (IPCC 2000).

46 A synopsis In this modeling domain, there are no established or commonly applied theoretical, empirical, or computational methods or procedures for evaluating models and defining and determining model validity, verisimilitude, or quality The EIA, by law, conducts retrospective reviews of NEMS forecast accuracy, but this is an exception Model credibility is claimed on other grounds, including: Internal consistency, and plausibility of results Usefulness for generating g insights

47 The most dramatic contributions of models are found in the isolation of counterintuitive results, which may arise because of overlooked factors or complex system interactions that cannot be assimilated easily (Hogan 1979; see also Peace and Weyant 2008) Although not always stated explicitly, there is a revealed epistemology of increasing levels of model detail being identified with increasing validity, verisimilitude, and/or usefulness for policy applications

48 Calibrationist philosophy [Modelers view] the widespread use of a particular structure in the theoretical literature [as] an indication of its worth, so that they seek less to test or validate models and more to explore the numerical implications of a particular model, conditional on having chosen it [their] focus is to generate insights about the effects of policy or other changes conditional on a particular theoretical structure (Dawkins et al. 2001).

49 Intra-model uncertainty The deterministic energy and IA models embody a very high degree of Knightian uncertainty: Uncertainty that cannot be represented in terms of probability distributions This uncertainty is generally not formally recognized or addressed as such The scenario methodology can be interpreted as a strategy for in effect side-stepping this problem Only an extremely small subset of possible equally plausible sets of inputs and parameters are actually analyzed The models, and their outputs, can be seen abstractly as single points in a very high-dimensional space

50 Inter-model uncertainty Example: Modeling the potential impacts of the Kyoto Protocol on the U. S. economy Energy Modeling Forum conducted a structured multi-model scenario study of this topic (Weyant and Hill 1999)

51 EMF-16 Model Predictions of Marginal Abatement Costs for the United States Derived from the No Trade and Annex 1 Trading Scenarios for Implementing the Kyoto Protocol

52 Concluding remarks Energy policy modeling both reveals and reflects a rather profound dilemma: While models of this type (although not necessarily any particular model or type of model) are indispensable for developing, analyzing, and implementing policies, this modeling is unavoidably subject to fundamental, irreducible uncertainties. This is ultimately a problem for policy not for modeling: This is ultimately a problem for policy, not for modeling: Notwithstanding widespread belief to the contrary, our capacity to accurately predict the effects of energy policies including those directed at CO2 emissions is extremely limited.

53 Thank you.

54 References Dawkins, Christina, and T. N. Srinivasan, John Whalley Calibration. Chapter 58 in J. J. Heckman and E. Leamer, Eds., Handbook of Econometrics, Volume 5. Elsevier Science B. V. Hogan, William W Energy Modeling: Building Understanding for Better Use. In: Proceedings of the Second Lawrence Symposium on Systems and Decision Sciences, Berkeley, California, October. IPCC (Intergovernmental Panel on Climate Change) IPCC Special Report [on] Emissions Scenarios Summary for Policymakers. A Special Report of IPCC Working Group III. Gillingham, K., and D. S. Rapson, G. Wagner The Rebound Effect and Energy Efficiency Policy. Review of Environmental Economics and Policy 10 (1), Winter:

55 Kim, Son H., and Jae Edmonds, Josh Lurz, Steven J. Smith, Marshall Wise The ObjECTS Framework for Integrated Assessment: Hybrid Modeling of Transportation. In J-C Hourcade, M. Jaccard, C. Bataille, F. Ghersi, Eds., The Energy Journal, Special Issue Hybrid Modeling of Energy-Environmental Policies: Reconciling Bottom-Up and Top-Down, pp Krey, Volker, and Gunnar Luderer, Leon Clarke, Elmar Kriegler Getting from here to there energy technology transformation pathways in the EMF27 scenarios. Climatic Change 123: Kriegler, Elmar, et al The role of technology for achieving climate policy objectives: overview of the EMF 27 study on global technology and climate policy strategies. Climatic Change 123:

56 National Research Council (NRC) The National Energy Modeling System. Committee on the National Energy Modeling System, Energy Engineering Board, Commission on Engineering and Technical Systems, in cooperation with the Committee on National Statistics, Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Nordhaus, William A Question of Balance Weighing the Options on Global Warming Policies. Yale University Press. Paltsev, S. et al., The MIT Emissions Prediction and Policy Analysis (EPPA) Model: Version 4, Report No. 125, MIT Joint Program on the Science and Policy of Global Change, August USEIA (U. S. Energy Information Administration). USEIA (U. S. Energy Information Administration). 2014a. Integrating Module of the National Energy Modeling System: Model Documentation 2014, U. S. Energy Information Administration (EIA), July. USEIA (U. S. Energy Information Administration). 2014b. Annual Energy Outlook 2014 with projections to Publication DOE/EIA-0383 (2014), April Weyant, John P., and Jennifer N. Hill Introduction and Overview. The Energy Journal, Kyoto Special Issue: vii-xiiv.