Robust Calculation and Parameter Estimation of the Hourly Price Forward Curve

Deregulated energy market participants mostly must use theHourly Price Forward Curves (HPFC) to evaluate their long term energy price contracts. We propose a new framework to estimate hourly, daily and yearly energy price profiles, based on the median estimation, instead of the widely used mean value. Hourly and daily data used for estimation are normalized in order to minimize the seasonality bias in our predictions. Given the high dimensionality of the problem, we use the LAD-Lasso model selection as a way to prevent data overfitting. To test the proposed framework we used the German electricity spot price time-series. We discuss the effects of large negative hourly prices, residuals of daily estimation and data outliers. We show that our framework provides significantly improved estimation of the HPFC compared to the simple mean estimators.