tristanfaure.com /projects / tsa-energy

TSA — Energy Price Analysis

Can EU electricity prices still be forecast through wars and crises?

2026 · ACADEMIC

Stack

Python · SARIMA · STL · Isolation Forest · FPCA · ENTSO-E

Key figures

Scope
FR · DE · HU · BG, 2015–2024
Extremes
+2,988 to −500 €/MWh
Forecasting
SARIMA, rolling eval
Anomalies
3 methods, cross-validated

Hourly day-ahead electricity prices for 4 European countries across a decade that contains +2,988 €/MWh spikes and negative prices — asking whether forecasting still works when the past no longer resembles the future.

Full statistical study: nested seasonalities, stationarity (ADF/KPSS), SARIMA identification via ACF/PACF and AIC/BIC, multi-step and rolling forecasts against a baseline, horizon-degradation analysis.

When prediction fails, detection takes over: STL residual thresholding, Isolation Forest on hourly contextual extremes, and FPCA for abnormally-shaped days — evaluated against real documented crisis events, plus a national-vs-European attribution via the 4-country panel.

Links

All 33 projects · Open the interactive portfolio · About Tristan