Thanks to the technological developments, renewable energies are becoming competitive against fossil sources and also wind farms are growing more and more integrated into intelligent power grids. For this reason, accurate power forecast is needed and often operators are charged with penalties in case of imbalance. Yet, wind is a stochastic and very local phenomenon. Time and space variability therefore conspire and wind power forecast is still challenging. Statistical (typically Artificial Neural Networks - ANN) methods are often employed for power forecast but they have some shortcomings: they require vast data sets and are not fit for capturing tails of distributions. In this work a pure ANN power forecast is compared against a hybrid method, based on the combination of ANN and a physical method as Computational Fluid Dynamics (CFD). The test case is a wind farm sited in southem Italy in a very complex terrain, and having a vast layout.
展开▼