Forecasting renewable production is a key activity in power systems. With the growing penetration of renewable energy sources, there is a pressing need for best manage supply/demand balance, therefore a reliable forecasting method of intermittent energy resources is an important issue. In this field, among renewable sources, the wind power one is characterized by the higher criticalities, due to the inherent intermittency not correlated with the day-night cycle. Moreover, the intermittent nature of wind power produces a heavy effect on the power system, because very often the production takes place in low-load conditions on the network, a condition for which a prediction error causes higher problems. The purpose of this work is to improve the wind forecasting developing a feed-forward neural network approach for wind power generation forecasting. Results from real-world case study, based on hourly meteorological data in South of Italy, are presented in order to show the proficiency of our proposed method. The effectiveness of our proposed methodology is clearly show by the value of three figure of merit: absolute percentage error (APE), mean absolute percentage error (MAPE) and mean square error (MSE). Obtained results are compared with their corresponding values generated by using the persistence model.
Artificial neural network application in wind forecasting: An one-hour-ahead wind speed prediction
Finamore, A. R.;Galdi, V.;Calderaro, V.
;Piccolo, A.;
2016
Abstract
Forecasting renewable production is a key activity in power systems. With the growing penetration of renewable energy sources, there is a pressing need for best manage supply/demand balance, therefore a reliable forecasting method of intermittent energy resources is an important issue. In this field, among renewable sources, the wind power one is characterized by the higher criticalities, due to the inherent intermittency not correlated with the day-night cycle. Moreover, the intermittent nature of wind power produces a heavy effect on the power system, because very often the production takes place in low-load conditions on the network, a condition for which a prediction error causes higher problems. The purpose of this work is to improve the wind forecasting developing a feed-forward neural network approach for wind power generation forecasting. Results from real-world case study, based on hourly meteorological data in South of Italy, are presented in order to show the proficiency of our proposed method. The effectiveness of our proposed methodology is clearly show by the value of three figure of merit: absolute percentage error (APE), mean absolute percentage error (MAPE) and mean square error (MSE). Obtained results are compared with their corresponding values generated by using the persistence model.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.