Bachelor's thesis · Aarhus University · 2026 · Grade 10
Electricity Price Forecasting in the Era of Foundation Models
A head-to-head benchmark of classical econometrics, machine learning and zero-shot foundation models for forecasting day-ahead electricity prices, on hourly data from 2020–2024 across six European bidding zones.
- 12 models benchmarked, from seasonal naïve and SARIMAX to CatBoost, NHITS, TimeGPT, Chronos, TimesFM and Lag-Llama
- Full dataset built from APIs (ENTSO-E, yfinance) with a Python/pandas pipeline that respects the 12:00 gate closure
- CatBoost cut forecast error (RMSE) by about 43% compared with the best classical model in Germany
