Renewable Variability & Electricity Price Risk Decomposition
A quantitative case study in CWE electricity markets
Day-ahead electricity prices in Central Western Europe are increasingly shaped by one variable above all others: how much wind is blowing and how much sun is shining. This project quantifies that relationship — and what it means for revenue risk — using five years of hourly market data from Germany, Denmark, France, and their neighbours.
The Analytical Question
How much of the variance in German day-ahead electricity prices can be attributed to renewable generation variability, and what does this imply for hedging demand across bidding zones?
The analysis is structured as a four-stage pipeline, each building on the last.
Stages
1. Data Sources & QA
Ingestion and quality audit of SMARD generation data, ENTSO-E cross-border flows and prices, Copernicus ERA5 reanalysis weather, and TTF gas futures. Covers 17 bidding zones, 52,000+ hourly observations, 2020–2025. Includes a downstream variable audit table documenting what feeds each analytical stage.
2. Weather → Generation
ERA5 wind speed and solar irradiance mapped to capacity factors for onshore wind, offshore wind, and solar PV in DE-LU. Tidymodels workflows with temporal cross-validation. Cross-zone weather profiles for CWE and Nordic markets. Centred temperature features derived for the price model. ACF/PACF residual diagnostics confirm model adequacy.
3. Price Decomposition
LASSO regression with hour × season dummies decomposes DE-LU day-ahead price variance into contributions from renewable share, residual load, thermal dispatch (gas, coal, lignite), TTF gas price (level + daily change), centred temperature, solar irradiance, and cross-border net exports. ADF test confirms TTF non-stationarity. Expanding-window cross-validation. Post-crisis elastic net (2023–2025) fitted alongside full-sample model. Partial R² and permutation importance by season.
4. Risk Quantification & Hedging Implications
Block-bootstrap Monte Carlo revenue simulation for a stylised 400 MW onshore wind portfolio. Two variants isolate weather-driven versus total price risk. VaR and CVaR computed on realized historical prices (not model predictions). Cross-zone correlation analysis with geographic mapping. Post-crisis DK_1 comparison. Seasonal hedging strategy recommendations across CWE bidding zones.
Data: SMARD (CC BY 4.0) · ENTSO-E Transparency Platform · Copernicus ERA5 · Yahoo Finance