We're testing two forecasting methods for volatile electricity prices on the day ahead electricity markets. Gaussian Processes work well for normal conditions but struggle with sudden price spikes. Student-T Processes handle extreme price movements better because they expect more volatility. Both models track seasonal patterns and market trends. We're evaluating them on European electricity markets, especially Slovenia, focusing on prediction accuracy, performance during price shocks, reliability of uncertainty bands, and ability to capture market volatility patterns.
|
