Modelling to generate alternatives with an energy system optimization model

Joseph F. DeCarolis, Samaneh Babaee, Binghui Li, S. KanungoEnviron. Model. Softw 2016 · DOI

TL;DR

This paper introduces a novel application of the "Modeling to Generate Alternatives" (MGA) technique to energy system optimization models (ESOMs). By systematically exploring the near-optimal decision space, MGA helps uncover a diverse set of energy futures that account for deep uncertainty, offering more robust insights than traditional scenario analysis.

What Problem Does This Paper Address?

Energy system planning faces "deep uncertainty," meaning we don't know the exact models to use, the probability distributions of key parameters, or how to value future outcomes. Traditional energy system optimization models (ESOMs) often produce single, precise projections that can be misleading. Scenario analysis, while useful, can be limited by cognitive biases and may not explore the full range of possibilities. This paper addresses the challenge of generating a diverse set of plausible energy futures that acknowledge and explore this deep uncertainty, moving beyond single-point projections.

Methods

The core methodological contribution is the application of the "Modeling to Generate Alternatives" (MGA) technique to an energy system optimization model (ESOM). Specifically, they use the "Hop-Skip-Jump" (HSJ) MGA algorithm, which iteratively modifies the optimization problem to find solutions that are distinct from previous optimal solutions while staying within a defined cost tolerance (slack). They also explore two MGA objective function weighting schemes: an "integer" method that increments weights for technologies appearing in previous solutions, and a "normalized sector" method that weights technologies based on their contribution within each sector. This approach allows for a systematic exploration of the decision space, uncovering alternative configurations that might be missed by conventional methods.

Datasets

The study utilizes a single-region U.S. database compatible with the Temoa model. This database includes projected fuel prices, technology cost and performance estimates, and end-use demands. Specific data sources mentioned include:

  • U.S. EIA (2012) for fuel price projections.
  • NUSTD (National TIMES dataset) as the primary source for technology cost and performance data.
  • U.S. EIA (2015) for updates to existing capacities and lifetimes.
  • AEO (2015) for technology updates and upper bounds on renewable capacity.
  • EPA (2013a, 2012, 2014, 2013b) for air quality regulations and CO2 rules.
  • U.S. EIA (2015) for electricity and light-duty vehicle sector data.

Key Insights and Ideas

  • MGA reveals a richer set of alternatives: By allowing for a modest increase in cost (slack), MGA can uncover a wide array of low-carbon technology configurations in both the electric and transportation sectors that are not apparent in standard scenario analyses.
  • Cost tolerance unlocks diverse futures: Even small increases in the allowable cost (e.g., 1-2% slack) can lead to significant emissions reductions (40-75%) and reveal different technology pathways compared to base or capped scenarios.
  • Discount rate and slack have similar effects: Both increasing the MGA slack value and increasing the discount rate can lead to the deployment of more capital-intensive technologies, particularly in later time periods. This suggests these parameters can be used to explore different risk appetites or societal preferences.
  • Sector-specific weighting matters: The "normalized sector" weighting scheme, which considers a technology's contribution within its sector, provided more nuanced results regarding emissions reductions compared to the "integer" method.
  • Policy implications are clearer: MGA can help identify policy levers, such as the potential for renewables to achieve emissions reductions without strict caps, or the impact of hurdle rates on alternative vehicle adoption.

Key Results

  • With a 1-2% cost slack, MGA can achieve 40-75% CO2 reductions by 2050 relative to 2015 levels, often utilizing renewables, IGCC-CCS, and nuclear in the electric sector.
  • The "normalized sector" weighting method, at 2% slack, produced emissions reductions ranging from 16% to 60% of 2015 levels, demonstrating a broader range of outcomes than the "integer" method which consistently met the 40% reduction target.
  • A specific MGA run (MGA-3-2%) showed a balanced approach using IGCC-CCS, nuclear, natural gas, renewables, and BEVs for a 41% reduction.
  • Another run (MGA-3-5%) relied heavily on natural gas and renewables for an 84% reduction.
  • Solar PV could reach an 18% market share with 2% slack and a 5% discount rate, constrained by annual growth rate limits.

Limitations and Future Work

  • Model Simplification: The study acknowledges the use of a simplified model, which may limit the generalizability of specific results.
  • Focus on Method: The paper emphasizes the methodological approach rather than deep dives into specific results due to dataset limitations.
  • No Probability Weighting: MGA results do not include probability-weighted outcomes or the value of imperfect information, suggesting that techniques like sensitivity analysis and stochastic optimization should also be used.
  • Future Work: The authors plan to design an interface that allows users to apply various uncertainty analysis techniques (sensitivity analysis, MGA, stochastic optimization) interactively within Temoa, enabling more robust exploration of energy futures. They also suggest further tests to explore the balance between solar PV and PEV deployment and to revisit hurdle rates for alternative vehicles.

Summary updated

BibTeX
@article{decarolis2016modelling,
  author = {Joseph F. DeCarolis and Samaneh Babaee and Binghui Li and S. Kanungo},
  title = {{Modelling to generate alternatives with an energy system optimization model}},
  journal = {Environ. Model. Softw.},
  year = {2016},
  doi = {10.1016/j.envsoft.2015.11.019},
}

Loading related content...

Figure