Predictive Analytics for electricity prices using Feed-INS from renewables

Since the liberalization of European electricity markets, stakeholders can actively participate in the trading of electricity. Successful participation in such markets requires an accurate forecast of future electricity prices. However, as large volumes of energy from renewable sources are fed into the system, electricity prices are highly volatile. While recent approaches put a strong focus on models from time series analysis using merely historic prices, only a few study the influence of exogenous predictors. This paper includes both expected solar power generation and expected wind power generation as exogenous predictors and improves state of the art by assessing their beneficial impact when evaluating our forecasting models with a two-pronged approach. First, we show that these externals decrease root mean squared errors by between 3.37 % and 9.86 %. Second, we apply a Diebold-Mariano test to prove statistically that the forecasting accuracy of the models including exogenous predictors is superior.