Weather-based machine learning technique for Day-Ahead wind power forecasting

This paper presents the development of forecast models for a wind farm producibility with a 24 hours horizon. The aim is to obtain accurate wind power predictions by using feedforward artificial neural networks. In particular, different forecasting models are developed and for each of them the best architecture is researched by means of sensitivity analysis, modifying the main parameters of the artificial neural network. The results obtained are compared with the forecasts provided by numerical weather prediction models (NWP).