The Hybrid Renewable Energy Forecasting and Trading Competition 2024

2025-07-02

AI Summary

How to translate meteorological and market-driven forecasting accuracy into tangible economic value for renewable energy participation in day-ahead electricity markets. Method: We develop an end-to-end forecasting–trading framework for a 3.6 GW UK wind–solar generation portfolio, applying gradient-boosting trees to jointly model numerical weather predictions and market clearing data; daily forecasts and bidding strategies are optimized over three months in 2024. Contribution/Results: This work presents the first real-world evaluation of both forecast accuracy and trading revenue simultaneously in an operational day-ahead market setting, revealing a nonlinear relationship between prediction error and economic value—thus shifting the paradigm from error minimization to value maximization in forecasting. Results show that while gradient-boosting trees achieve high predictive accuracy, revenue is critically contingent on bidding strategy design. All data and code are publicly released, establishing a new value-oriented benchmark for renewable energy forecasting research. (149 words)

Technology Category

• Reasoning under Uncertainty: Stochastic Optimization
• Search and Optimization: Mixed Discrete/Continuous Search
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

The Hybrid Energy Forecasting and Trading Competition challenged participants to forecast and trade the electricity generation from a 3.6GW portfolio of wind and solar farms in Great Britain for three months in 2024. The competition mimicked operational practice with participants required to submit genuine forecasts and market bids for the day-ahead on a daily basis. Prizes were awarded for forecasting performance measured by Pinball Score, trading performance measured by total revenue, and combined performance based on rank in the other two tracks. Here we present an analysis of the participants' performance and the learnings from the competition. The forecasting track reaffirms the competitiveness of popular gradient boosted tree algorithms for day-ahead wind and solar power forecasting, though other methods also yielded strong results, with performance in all cases highly dependent on implementation. The trading track offers insight into the relationship between forecast skill and value, with trading strategy and underlying forecasts influencing performance. All competition data, including power production, weather forecasts, electricity market data, and participants' submissions are shared for further analysis and benchmarking.

Problem

• Forecast electricity generation from wind and solar farms
• Optimize day-ahead energy trading strategies
• Evaluate performance using forecasting and revenue metrics

Innovation

• Hybrid wind and solar forecasting competition
• Gradient boosted tree algorithms for forecasting
• Forecast skill and trading strategy analysis