Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning-Enabled Large-Scale Capacity Expansion Planning under Uncertainty

Taehyeon Kwon, Anirudh Subramanyam

Abstract

Capacity expansion planning under uncertainty requires selecting a scenario count and representative operational horizon to estimate average production costs. Small choices risk unreliable plans, while large choices become intractable. We propose AutoSCEP, an automated, statistically grounded procedure that, for a fixed plan, selects the minimum sufficient scenario count and horizon length to estimate production costs to a given precision. Using these estimates, we train linear and neural surrogates to approximate expected production costs for arbitrary plans, and embed the surrogates within the planning model. On the continental-scale EMPIRE system, AutoSCEP attains 2% optimality gap on a reduced model and 8% gap on a large model, outperforming parallel progressive hedging under equal wall-clock budgets that include data generation, training, and solve times. Where the reduced model's optimum is available, investment patterns broadly align with the benchmark. Our approach enables high-resolution uncertainty modeling at realistic system scales.

Machine Learning-Enabled Large-Scale Capacity Expansion Planning under Uncertainty

Abstract

Capacity expansion planning under uncertainty requires selecting a scenario count and representative operational horizon to estimate average production costs. Small choices risk unreliable plans, while large choices become intractable. We propose AutoSCEP, an automated, statistically grounded procedure that, for a fixed plan, selects the minimum sufficient scenario count and horizon length to estimate production costs to a given precision. Using these estimates, we train linear and neural surrogates to approximate expected production costs for arbitrary plans, and embed the surrogates within the planning model. On the continental-scale EMPIRE system, AutoSCEP attains 2% optimality gap on a reduced model and 8% gap on a large model, outperforming parallel progressive hedging under equal wall-clock budgets that include data generation, training, and solve times. Where the reduced model's optimum is available, investment patterns broadly align with the benchmark. Our approach enables high-resolution uncertainty modeling at realistic system scales.
Paper Structure (15 sections, 5 equations, 2 figures, 3 tables, 2 algorithms)

This paper contains 15 sections, 5 equations, 2 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Stochastic Capacity Expansion Planning Problem Formulation
  3. ML Surrogate-Based Solution Methodology
  4. Generating Training Data
  5. Surrogate Model Architectures and Training
  6. Surrogate Model Embedding
  7. Computational Experiments
  8. Benchmark Model
  9. Comparative Baselines
  10. Implementation Details
  11. Comparative Results
  12. Adaptive Parameter Selection
  13. Capacity Expansion Solution Quality and Optimality
  14. Structural Reliability of Investment Plans
  15. Conclusion

Figures (2)

  • Figure 1: Comparison of fixed and adaptive parameter selection for generating a training label for a fixed $\boldsymbol{x}$. (Left) Heatmap of relative error in $\widehat{\mathcal{Q}}$ with respect to the largest configuration ($S=30, |\mathcal{H}|=48$) for varying scenario counts $S$ and operational horizon lengths $|\mathcal{H}|$. The star marks the parameters selected by Algorithm \ref{['alg:adaptive_labeling']} (2.2% relative error). (Right) Wall-clock time per label for representative fixed configurations and the adaptive method; bar color encodes relative error using the same colormap as the left panel.
  • Figure 2: Cumulative installed capacity by technology for the EMPIRE-sml case study; shaded regions represent 95% confidence intervals across ten solutions.