Wind speed forecasting at different time scales: a non parametric approach Guglielmo D’Amico Dipartimento di Farmacia, Universit`a ‘G. D’Annunzio’ di Chieti-Pescara, 66013 Chieti, Italy Filippo Petroni Dipartimento di Scienze Economiche ed Aziendali, Universit`a degli studi di Cagliari, 09123 Cagliari, Italy Flavio Prattico Dipartimento di Ingegneria Industriale e dell’Informazione e di Economia, Universit`a degli studi dell’Aquila, 67100 L’Aquila, Italy Abstract The prediction of wind speed is one of the most important aspects when dealing with renewable energy. In this paper we show a new nonparametric model, based on semi-Markov chains, to predict wind speed. Particularly we use an indexed semi-Markov model, that reproduces accurately the sta- tistical behavior of wind speed, to forecast wind speed one step ahead for different time scales and for very long time horizon maintaining the good- ness of prediction. In order to check the main features of the model we show, as indicator of goodness, the root mean square error between real data and predicted ones and we compare our forecasting results with those of a persistence model. Keywords: Wind speed; forecasting model; indexed semi-Markov chains; 1. Introduction The variations of wind speed, in a certain site, are strictly related to the economic aspects of a wind farm, such as maintenance operations, especially Preprint submitted to Renewable Energy May 17, 2013 arXiv:1305.3696v1 [physics.data-an] 16 May 2013