PaperlayerView exact revision on arXiv

Source-linked AI summary

Energy Scenario Exploration with Modeling to Generate Alternatives (MGA)

Joseph F. DeCarolis, Samaneh Babaee, Binghui Li, Suyash Kanungo

arXiv:1912.03788v1physics.soc-phecon.EM

TL;DR

Deep uncertainty limits what singular ESOM projections and conventional scenarios can establish for energy planning. The paper incorporates MGA into Temoa to explore near-optimal alternatives in a simplified U.S. electricity and light-duty transport system. The results show alternative technology configurations, while the authors note that MGA does not provide probability-weighted outcomes or value of imperfect information.

  • Problem

    Deep uncertainty about model structure, future parameters, and outcome values, alongside limitations of conventional scenario analysis, complicates the use of ESOMs for energy planning.

    Source: §1 · pp. 2–3

  • Method

    The paper incorporates MGA into Temoa to generate near-optimal alternatives and customizes its objective using technology activity, including normalized sector weights.

    Source: §1 · p. 3, §3.2 · pp. 8–9

  • Results

    MGA solutions include significant IGCC, biomass, and wind deployment in electricity and BEVs in light-duty transport, with the model changing electricity generation more readily than the vehicle mix.

    Source: §5 · p. 17

  • Takeaways & Limitations

    The authors conclude that interactive exploration of model decision space can challenge mental models and yield new insight beyond conventional scenario results.

    Source: §6 · p. 19

    MGA results do not include probability-weighted outcomes or the value of imperfect information, so sensitivity analysis and stochastic optimization should also be used.

    Source: §5 · p. 18

Abstract

from arXiv · show

Energy system optimization models (ESOMs) should be used in an interactive way to uncover knife-edge solutions, explore alternative system configurations, and suggest different ways to achieve policy objectives under conditions of deep uncertainty. In this paper, we do so by employing an existing optimization technique called modeling to generate alternatives (MGA), which involves a change in the model structure in order to systematically explore the near-optimal decision space. The MGA capability is incorporated into Tools for Energy Model Optimization and Analysis (Temoa), an open source framework that also includes a technology rich, bottom up ESOM. In this analysis, Temoa is used to explore alternative energy futures in a simplified single region energy system that represents the U.S. electric sector and a portion of the light duty transport sector. Given the dataset limitations, we place greater emphasis on the methodological approach rather than specific results.

1. Introduction

Deep uncertainty about models, future parameters, and how to value outcomes makes singular projections and conventional scenario analysis inadequate guides for energy planning. The paper presents MGA as an interactive way to explore near-optimal energy-system alternatives and inform decisions.

1. Introduction

  • ESOMs face deep uncertainty about model structure, future parameters, and the value of outcomes, so no single model structure fully captures the real world.The paper characterizes these ill-defined public planning problems as “wicked.”Source: §1 · p. 2
  • Scenario storylines can appear more plausible than warranted, and traditional scenarios may leave decision makers with disparate futures or fail to build consensus in broad public debates.The cited limitations are especially relevant where stakeholders have divergent interests and values.Source: §1 · pp. 2–3
  • The paper identifies a need for a framework that lets users iteratively run models and formulate insights for decision making, including through visualization and search.The cited ideal uncertainty-analysis outputs include probability-weighted results, decisions accounting for imperfect information, outcome-risk measures, and value of information.Source: §1 · p. 3
  • MGA changes model structure to generate sequences of near-optimal solutions that differ in decision space, offering alternatives for analyst evaluation.The paper presents this as the first published application of MGA to an ESOM.Source: §1 · p. 3
  • The demonstration uses Temoa with a simplified dataset for U.S. electricity and light-duty transportation, emphasizing method over specific results because of dataset limitations.Plug-in electric vehicles connect the sectors through charging demand.Source: §1 · p. 3

2. An Electric and Transportation Sector Case Study

The case study applies Temoa to a single-region U.S. energy system spanning electricity and light-duty transport, using a simplified dataset and scenarios that benchmark alternative low-carbon pathways. Its model and data assumptions constrain the scope and temporal detail of the exploration.

2. An Electric and Transportation Sector Case Study

  • The study compares a base case with moderate and aggressive CO2 caps, then applies MGA to the moderate cap to explore alternative cost- and emissions-constrained solutions.The caps target 40% and 80% reductions below 2015 emissions by 2050, respectively.Source: §2 · p. 4

2.1 Model Description

  • Temoa minimizes system-wide energy-supply cost while meeting end-use demands, with technology, energy-conservation, and optional policy constraints.It assumes rational decisions, perfect information, and perfect foresight; fixed demands make it a simplified partial-equilibrium model.Source: §2.1 · p. 4

2.2 Data Description

  • The model horizon runs from 2015 to 2050 in five-year periods, represents daily renewable variation with four time segments, and omits seasonal variability.Each year within a five-year period is assumed to have identical characteristics.Source: §2.2 · p. 4
  • The single-region database draws on NUSTD and AEO updates, representing fuel supply, electricity generation, and light-duty vehicles with endogenous electricity prices and separate electricity and travel demands.The electricity sector includes 32 generation technologies and two CCS retrofit technologies; the LDV sector includes 48 technologies.Source: §2.2 · pp. 5–6, Figure 1 · p. 5
  • Alternative vehicle technologies use a 10% technology-specific hurdle rate, while the MGA runs retain 2015 technology activity from the base case.The hurdle rate is intended to partially capture non-market factors affecting alternative-fuel vehicle deployment; the dataset also assumes EIA reference fuel prices and a battery cost of 135 $/kWh in 2035.Source: §2.2 · p. 6

3. Methods

The methods use Hop-Skip-Jump MGA to search for distinct near-optimal configurations under a cost constraint, then customize its objective for energy-system technologies. Two weighting schemes and varying slack values support exploration of alternative configurations and uncertainty.

3. Methods

  • MGA changes model structure to search the feasible near-optimal region for solutions that differ in decision space, complementing scenario analysis and addressing structural uncertainty.The paper also interprets MGA results as objective-coefficient perturbations relevant to parametric uncertainty.Source: §3 · p. 7

3.1 Hop-Skip-Jump MGA

  • Hop-Skip-Jump MGA obtains an optimum, adds objective slack as a constraint, then iteratively minimizes activity from prior solutions until changes are no longer significant.The revised objective is constrained to remain within the prescribed slack of the original optimum.Source: §3.1 · pp. 7–8

3.2 MGA Customization

  • The ESOM customization uses total technology activity over the model horizon, rather than installed capacity, because capacity may not be used to meet demand.The derived variable V_ActivityByTech sums each technology’s total output across the horizon.Source: §3.2 · p. 8
  • The MGA objective includes electric-generation and vehicle technologies but excludes upstream fuel-supply, blending, and emissions-accounting processes to avoid penalizing technologies linked to more such processes.Source: §3.2 · p. 8
  • The integer weighting scheme increments a technology’s objective weight by +1 after each iteration with positive activity, while normalized-sector weighting accumulates its fractional activity within its sector.The latter accounts for technologies’ relative roles when activity units differ between sectors.Source: §3.2 · pp. 8–9
  • Varying slack allows costlier configurations, which the authors treat as potentially reflecting either energy-technology cost uncertainty or complex unmodeled issues.Source: §3.2 · p. 9

4. Results

The base and CO2-cap scenarios show how the modeled sectors respond to emissions constraints, while MGA reveals alternative technology configurations under modest cost increases. Further tests show how weighting, slack, and discount-rate choices shape those alternatives.

4.1 Results from the base and CO2 cap scenarios

  • In the base case, coal generation stays relatively constant as nuclear and wind gain share; gasoline vehicles remain dominant, and no plug-in electric vehicles are deployed.Increased natural-gas prices toward mid-century accompany declining gas-turbine activity; E85 gains share by 2050.Source: §4.1 · p. 9, Figure 2 · p. 9
  • Under the 40% cap, wind and IGCC-CCS offset declining coal, while the 80% cap displaces coal with gas turbines and IGCC-CCS; BEV160 reaches 21% of 2050 vehicle share.Changes in the light-duty vehicle sector are minimal in the 40% cap scenario; gasoline vehicles remain dominant under the 80% cap.Source: §4.1 · p. 10, Figure 3 · p. 10

4.2 MGA weighting scheme test

  • The integer weighting method deploys more technologies, but the normalized sector method yields 2050 emissions from 16% to 60% of 2015 levels; the authors examine it further.With integer weighting, the CO2 cap always binds at 60% of 2015 emissions.Source: §4.2 · p. 11, Figure 4 · p. 12

4.3 In-depth MGA results

  • With 1–2% cost slack, MGA produces 40–75% emissions reductions in 2050; the 1% slack fourth iteration reaches 63% at 0.6% more cost than the moderate CO2-cap scenario.The reductions rely largely on wind, IGCC-CCS, and nuclear; gasoline remains dominant in light-duty transport, with modest BEV contributions.Source: §4.3 · p. 13, Figure 5 · p. 13
  • MGA-3-5% achieves an 84% emissions reduction using natural gas and renewables, while MGA-4-10% uses nearly 60% natural-gas market share and alternative vehicles to meet the 40% target.These examples illustrate how system designs vary across slack values; Fig. 6 shows MGA-3-2% and MGA-4-10%.Source: §4.3 · p. 14, Figure 6 · p. 14
  • Higher slack increases variation in technology deployment; solar PV remains insignificant and alternative vehicles other than BEVs see little deployment until slack reaches 10%.Higher slack gives the model more room under the cost constraint to deploy more expensive technologies.Source: §4.3 · pp. 14–15
  • At 2% slack, a 10% discount rate enables a wider technology suite than 5%, with deployment patterns qualitatively similar to 5% slack at the 5% rate.The 10% rate permits greater 2050 deployment of capital-intensive coal- and biomass-based IGCC and BEVs than the 5% rate at 2% slack.Source: §4.3 · p. 15, Figure 7 · p. 16
  • With 2% slack, maximizing solar PV and plug-in vehicle deployment yields 18% solar market share, but only modest PHEV20 deployment as solar uses much of the cost slack.The annual 10% solar growth constraint binds, allowing about 500 GW of installed solar capacity by 2050.Source: §4.3 · p. 17, Figure 8 · p. 17

5. Discussion

MGA generates diverse, cost-constrained energy futures, while its results depend on slack and weighting choices and do not provide a complete uncertainty analysis.

5. Discussion

  • Varying MGA slack and discount rate can produce diverse futures, but higher slack often extends existing technology patterns rather than introducing new technologies.A targeted run found an 18% solar PV market share possible at 2% slack, at a cost similar to the 80% cap scenario.Source: §5 · pp. 17–18
  • MGA results include significant deployment of IGCC, biomass, and wind in electricity and BEVs in light-duty vehicles; alternative vehicle technologies appear only with 10% slack.The model varies electricity generation more readily than vehicle technologies because it can make larger activity changes at lower cost penalty.Source: §5 · p. 17
  • The MGA objective weighting scheme should fit the model context; normalized-sector weights favor technologies with smaller roles in prior solutions.The authors suggest testing technology-group and sector-specific weights when an analysis aims to focus on particular sectors or technology sets.Source: §5 · p. 18
  • MGA observations suggest renewables incentives could be explored as an alternative to a politically untenable CO2 cap, while lower alternative-vehicle hurdle rates could be tested.The authors propose running additional scenarios with reduced renewable investment costs and lower vehicle hurdle rates.Source: §5 · p. 18
  • MGA does not provide probability-weighted outcomes or the value of imperfect information, so the authors recommend also using sensitivity analysis and stochastic optimization.Source: §5 · p. 18

6. Conclusions

The conclusions argue that deep uncertainty makes precise-looking energy projections problematic and call for interactive exploration of model decision spaces. The authors present MGA as one useful approach and envision a Temoa interface for applying multiple uncertainty methods.

6. Conclusions

  • Under deep uncertainty, energy models should be used iteratively to search decision spaces rather than produce precise-looking projections.Source: §6 · p. 19
  • The authors envision an interface for Temoa that would let users apply MGA, sensitivity analysis, and stochastic optimization with visualization to explore energy futures.They argue that model-based analysis may be most useful as a tool for user interrogation rather than as a set of projections.Source: §6 · p. 19

Appendix A: Case Study Input Data

The appendix materials identify the modeled energy-system network and list input-database technologies and case-study data tables.

Appendix A: Case Study Input Data

  • The appendix includes a network flow diagram for the modeled energy system.Source: Figure A.1 · p. 24
  • The input-database materials include technology names and descriptions, including light-duty vehicle technologies and electricity-generation technologies.Source: Table A.1 · p. 25, Appendix A · p. 25
  • The case-study data tables cover capacity and growth limits, electricity and light-duty transportation demand, vehicle size-class shares, and selected generation and fuel-import bounds.Source: Table A.3 · p. 26, Table A.4 · p. 27, Table A.5 · p. 27

Appendix B: Complete Results from MGA Integer and Normalized Sector Weighting Methods

The supplied passage provides publication and licensing information for this appendix.

  • The manuscript is identified by DOI 10.1016/j.envsoft.2015.11.019 and made available under the CC-BY-NC-ND 4.0 license.Source: Appendix B · p. 29

MGA: Integer Weighting with 1% Slack

The supplied passages identify electricity generation (TWh / yr), the value 5000, and several technology labels, but provide no readable comparison of results.

MGA: Integer Weighting with 2% Slack

The supplied passages contain electricity-generation labels and technology names across the listed MGA settings, but no readable results comparing those settings.

MGA: Integer Weighting with 2% Slack

MGA: Integer Weighting with 5% Slack

MGA: Integer Weighting with 10% Slack

MGA: Normalized Weighting by Sector with 1% Slack

MGA: Normalized Weighting by Sector with 2% Slack

MGA: Normalized Weighting by Sector with 5% Slack

MGA: Normalized Weighting by Sector with 10% Slack

The supplied passages label electricity generation and list several technologies, but provide no readable comparison of results.

Appendix C: Results From Discount Rate Tests

The supplied passages identify discount-rate settings and electricity-generation labels, but do not provide readable comparisons across the tests.

Appendix C: Results From Discount Rate Tests

  • The appendix labels tests at Moderate CO2 Cap with 0.1% and 10% discount rates and lists IGCC-CCS, Nuclear, Natural Gas, Hydro, and Wind.Source: Appendix C · p. 41

MGA: Normalized Weighting by Sector – 2% Slack – 10% Discount Rate

Loading 1912.03788v1…