A data-driven merit order: Learning a fundamental electricity price model

TL;DR

The paper addresses the challenge of accurately forecasting electricity prices under increasing volatility and regime shifts by proposing a data-driven yet fundamentally interpretable merit-order framework. It embeds the classical merit order as a special case and learns plant-level parameters from data, while incorporating extensions for hydro, net imports, capacity corrections, gas stack splits, and must-run shares to better reflect market dynamics. Empirical results on the German day-ahead market show the approach outperforms traditional fundamental and data-driven baselines in MAE while maintaining causal interpretability, including insights into marginal technologies and fuel-switching. The framework is computationally efficient, enabling rapid forecasting and offering a flexible basis for policy analysis, cross-border flow modeling, and future enhancements such as probabilistic forecasting and storage integration.

Abstract

Electricity price forecasting approaches generally fall into two categories: data-driven models, which learn from historical patterns, or fundamental models, which simulate market mechanisms. We propose a novel and highly efficient data-driven merit order model that integrates both paradigms. The model embeds the classical expert-based merit order as a nested special case, allowing all key parameters, such as plant efficiencies, bidding behavior, and available capacities, to be estimated directly from historical data, rather than assumed. We further enhance the model with critical embedded extensions such as hydro power, cross-border flows and corrections for underreported capacities, which considerably improve forecasting accuracy. Applied to the German day-ahead market, our model outperforms both classic fundamental and state-of-the-art machine learning models. It retains the interpretability of fundamental models, offering insights into marginal technologies, fuel switches, and dispatch patterns, elements which are typically inaccessible to black-box machine learning approaches. This transparency and high computational efficiency make it a promising new direction for electricity price modeling.

Figures (13)

• Figure 1: Fuel prices and daily average day-ahead power prices in Germany.
• Figure 2: Average production given day-ahead power price intervals for 2023-2024 in Germany.
• Figure 3: Evolution of day-ahead electricity prices in Germany
• Figure 4: Comparison between existing model classes, and the newly introduced data-driven fundamental model type as a hybrid combination of the two.
• Figure 5: Fundamental vs data-driven models forecasts for Germany (2024).