Deep Learning (DL)-based methods are becoming of interest in the power community thanks to their ability to extract information and patterns in large sets of input data. In this paper, we exploit Convolutional Neural Networks (CNNs) to process the information of wind speed forecasts at different altitudes to improve the accuracy of the prediction of wind power generation. Through extensive comparisons on actual hourly data from four wind farms in Italy, we show that the forecasts of the wind components at different altitudes do provide an advantage in obtaining accurate forecasts. In doing so, we also evaluate the importance of other design aspects, such as the choice of filters and activation functions, in the implementation of the methodology. Finally, results are compared with other more conventional benchmark prediction algorithms as well.
Wind Power Forecast Using Wind Forecasts at Different Altitudes in Convolutional Neural Networks
Li Bai
;Emanuele Crisostomi;Marco Raugi;Mauro Tucci
2019-01-01
Abstract
Deep Learning (DL)-based methods are becoming of interest in the power community thanks to their ability to extract information and patterns in large sets of input data. In this paper, we exploit Convolutional Neural Networks (CNNs) to process the information of wind speed forecasts at different altitudes to improve the accuracy of the prediction of wind power generation. Through extensive comparisons on actual hourly data from four wind farms in Italy, we show that the forecasts of the wind components at different altitudes do provide an advantage in obtaining accurate forecasts. In doing so, we also evaluate the importance of other design aspects, such as the choice of filters and activation functions, in the implementation of the methodology. Finally, results are compared with other more conventional benchmark prediction algorithms as well.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.