Benchmark of Spatio-temporal Shortest-Term Wind Power Forecast Models

Many European energy supply systems are increasingly penetrated by wind energy. In order to be able to act optimally on the market or in the operation of electricity grids, it is necessary to have high-quality intraday forecasts of the expected wind power production. For this purpose, numerical weather forecasts and recent power measurements transmitted in real time are used. This provides a lot of information to the forecaster. It is, on the one hand, necessary to be able to decide which information are beneficial, and on the other hand, to be able to handle proper forecasting models. Suitable models to calculate wind power forecasts are power curve-based models and models from the field of statistics as well as machine learning models. In this work we benchmark (first) different models ranging from power curve based to machine learning like random forests, artificial neural networks and extreme learning machines, and (second) the value of spatio-temporal information from surrounding wind parks.