Temporal hierarchy forecasting improves day ahead electricity market profits

Stealing profits: Spread-based temporal hierarchy forecasting for day-ahead electricity markets

Machine Learning

Summary

Predicting the price of electricity for each hour helps people decide when to sell or store energy, especially for batteries. The authors show that predicting the differences (or spreads) between prices at different times, and making these predictions work together consistently, leads to better accuracy and more profit. They tested their method on data from Germany and Spain over five years and found improvements in both prediction and financial outcomes. This approach also worked well even with advanced pre-trained models, showing that better stats don’t always mean better money decisions.

What this means in practice

  • For energy traders: Make more profitable intraday trading decisions by forecasting coherent price spreads instead of individual hourly prices alone.
  • For grid storage operators: Optimize charging and discharging schedules for batteries using forecasts that better represent price differences throughout the day.

Authors

Arkadiusz Lipiecki, Nikolaos Kourentzes, Rafal Weron

Abstract

Day-ahead electricity price forecasts support trading and storage decisions, but for battery arbitrage predicting intraday price spreads is more relevant than predicting individual hourly prices. Here we show that a temporal hierarchy forecasting (THieF) framework that jointly reconciles forecasts of hourly electricity prices and all intraday price spreads consistently improves performance across two major European electricity markets and three different forecasting architectures. Using five years of out-of-sample data from Germany and Spain, we obtain accuracy improvements of up to 19.7% and profit gains of up to 10.4% relative to unreconciled hourly price forecasts. The gains persist even for a highly accurate pretrained TabPFN foundation model. Our results demonstrate that exploiting coherent relationships between economically relevant forecasting targets can improve both predictive accuracy and decision value, and that better statistical forecasts do not necessarily imply better economic decisions.