Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Parametric and Generative Forecasts of Day-Ahead Market Curves for Storage Optimization

Julian Gutierrez, Redouane Silvente

TL;DR

This work tackles forecasting and optimization in the day-ahead electricity market by developing two complementary forecasting frameworks. The fast parametric model delivers rapid, interpretable eight-parameter curve forecasts per hour using a degree-3 Chebyshev elastic component and XGBoost, suitable for daily storage decisions with typical errors around the low single-digit to mid‑single-digit percentages. The generative approach uses DDPMs to model bid-level curves via marked Cox processes, enabling probabilistic scenario generation conditioned on weather and fuel prices for deeper storage analysis; it demonstrates how order-level heterogeneity can be captured and aggregated into hourly supply/demand curves. When combined with a price-maker storage optimization grounded in a Supply Function Equilibrium framework, the results show that storage can compress price spreads and improve profitability, with diminishing returns as capacity grows, while the generative model provides richer uncertainty quantification and robustness insights. Overall, the paper offers a tractable, dual‑model framework that supports both fast operational use and thorough scenario-based analysis for storage strategies in a growing, volatile day-ahead market, with practical implications for market stability and investment decisions.

Abstract

We present two machine learning frameworks for forecasting aggregated curves and optimizing storage in the EPEX SPOT day-ahead market. First, a fast parametric model forecasts hourly demand and supply curves in a low-dimensional and grid-robust representation, with minimum and maximum volumes combined with a Chebyshev polynomial for the elastic segment. The model enables daily use with low error and clear interpretability. Second, for a more comprehensive analysis, though less suited to daily operation, we employ generative models that learn the joint distribution of 24-hour order-level submissions given weather and fuel variables. These models generate synthetic daily scenarios of individual buy and sell orders, which, once aggregated, yield hourly supply and demand curves. Based on these forecasts, we optimize a price-making storage strategy, quantify revenue distributions, and highlight the price-compression effect with lower peaks, higher off-peak levels, and diminishing returns as capacity expands.

Parametric and Generative Forecasts of Day-Ahead Market Curves for Storage Optimization

TL;DR

This work tackles forecasting and optimization in the day-ahead electricity market by developing two complementary forecasting frameworks. The fast parametric model delivers rapid, interpretable eight-parameter curve forecasts per hour using a degree-3 Chebyshev elastic component and XGBoost, suitable for daily storage decisions with typical errors around the low single-digit to mid‑single-digit percentages. The generative approach uses DDPMs to model bid-level curves via marked Cox processes, enabling probabilistic scenario generation conditioned on weather and fuel prices for deeper storage analysis; it demonstrates how order-level heterogeneity can be captured and aggregated into hourly supply/demand curves. When combined with a price-maker storage optimization grounded in a Supply Function Equilibrium framework, the results show that storage can compress price spreads and improve profitability, with diminishing returns as capacity grows, while the generative model provides richer uncertainty quantification and robustness insights. Overall, the paper offers a tractable, dual‑model framework that supports both fast operational use and thorough scenario-based analysis for storage strategies in a growing, volatile day-ahead market, with practical implications for market stability and investment decisions.

Abstract

We present two machine learning frameworks for forecasting aggregated curves and optimizing storage in the EPEX SPOT day-ahead market. First, a fast parametric model forecasts hourly demand and supply curves in a low-dimensional and grid-robust representation, with minimum and maximum volumes combined with a Chebyshev polynomial for the elastic segment. The model enables daily use with low error and clear interpretability. Second, for a more comprehensive analysis, though less suited to daily operation, we employ generative models that learn the joint distribution of 24-hour order-level submissions given weather and fuel variables. These models generate synthetic daily scenarios of individual buy and sell orders, which, once aggregated, yield hourly supply and demand curves. Based on these forecasts, we optimize a price-making storage strategy, quantify revenue distributions, and highlight the price-compression effect with lower peaks, higher off-peak levels, and diminishing returns as capacity expands.
Paper Structure (28 sections, 3 theorems, 23 equations, 41 figures, 3 tables, 1 algorithm)

This paper contains 28 sections, 3 theorems, 23 equations, 41 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Literature review
  3. Fast Parametric Demand-Curve Forecasting
  4. Dataset of interest
  5. Problem reduction
  6. Choice of features
  7. Time series
  8. Features from the original points
  9. External features
  10. Model and results
  11. Generative Supply Curves from Marked Point Processes and DDPMs
  12. The supply curve modeling
  13. Arrival price process generation
  14. Marks volumes generation
  15. Results
  16. ...and 13 more sections

Key Result

Theorem 4.1

Suppose that Assumptions ass.gamma.diff and ass.X.dens hold. There exists an optimal function $q^*$ that maximises problem.

Figures (41)

  • Figure 1: Aggregated bid and offers for 2022-11-03 at 3pm. The Demand curve, including an elastic part, is in orange and the Supply curve is in blue. They cross at the day-ahead electricity price.
  • Figure 2: Evolution of the day-ahead electricity price over time between 2018 and 2024
  • Figure 3: Progressive approximation steps: from interpolation to the final reconstruction.
  • Figure 4: Median value for every hour of the normalized Mean Absolute Error between the original points and the polynomial approximation for the demand (left) and the supply (right) curves.
  • Figure 5: Spearman's correlation heatmap between the aimed coefficients and the external features for the demand forecast.
  • ...and 36 more figures

Theorems & Definitions (4)

  • Theorem 4.1: Existence
  • proof
  • Proposition 4.1: Characterization of the optimal strategy
  • Corollary 4.1: Price-form of the optimal strategy