TRACE: A Reproducible Benchmark for Electricity Price Forecasting with Official Operational Text

2026-09-30

AI Summary

This study addresses the limitations of existing electricity price forecasting benchmarks, which often overlook the value of textual context and are susceptible to information leakage. We construct a reproducible benchmark comprising 7,300 regional daily instances, rigorously reconstructing pre-cutoff official operational texts to eliminate leakage risks. By integrating time-series foundation models with natural language processing techniques, we systematically evaluate the semantic alignment and predictive utility of such textual information. Ablation studies demonstrate that incorporating operational texts reduces the median upper-tail Pinball loss by 7.4%, validating the effectiveness of textual signals in capturing extreme price risks. This work establishes a rigorous new paradigm for tail forecasting in electricity markets.

Abstract

Electricity price forecasting (EPF) supports scheduling, bidding, and risk management in electricity markets, yet existing benchmarks focus mainly on numerical inputs, leaving the forecasting value of forecast-time textual context insufficiently evaluated. We introduce TRACE, a reproducible benchmark of 7,300 zone--day instances pairing prices from five zones in a major U.S. market with official operational text available at the forecast cutoff. TRACE reconstructs official operational text at each cutoff, preventing post-cutoff information leakage. We evaluate TRACE for semantic alignment and forecasting value. Semantic assessments align with central movement and both tail risks in ground-truth prices, most consistently for upper-tail price risk. Forecasting value is reflected in a median 7.4\% reduction in upper-tail pinball loss across time-series foundation models. A controlled cross-day text-mismatch ablation reverses the gains, falling below the no-text baseline.

Problem

• Electricity price forecasting
• textual context
• benchmark
• operational text

Innovation

• Electricity Price Forecasting
• Reproducible Benchmark
• Operational Text
• Information Leakage Prevention
• Time-series Foundation Models