Potential of Ensemble Copula Coupling for Wind Power Forecasting

With the share of renewable energy sources in the energy system increasing, accurate wind power forecasts are required to ensure a balanced supply and demand. Wind power is,  however, highly dependent on the chaotic weather system and other stochastic features. Therefore, probabilistic wind power forecasts are essential to capture uncertainty in the model  parameters and input features. The weather and wind power forecasts are generally post-processed to eliminate some of the systematic biases in the model and calibrate it to past  observations. While this is successfully done for wind power forecasts, the approaches used often ignore the inherent correlations among the weather variables. The present paper,  therefore, extends the previous post-processing strategies by including Ensemble Copula Coupling (ECC) to restore the dependency structures between variables and investigates,  whether including the dependency structures changes the optimal post-processing strategy. We find that the optimal post-processing strategy does not change when including ECC and  ECC does not improve the forecast accuracy when the dependency structures are weak. We, therefore, suggest investigating the dependency structures before choosing a post- processing strategy.