A novel forecasting model for wind speed assessment using sentinel family satellites images and machine learning method

Wind energy produced in near and offshore farms is constantly increasing, mainly thanks to the use of new technologies and the economic costs decrease. Data coming from satellites recently launched can be effectively used in near and offshore areas for wind speed mapping, and long and short-term analyses of wind regimes. Furthermore, the use of artificial intelligence methods is constantly increasing for predicting wind energy production. In this framework, a new forecasting model for wind speed assessment is presented integrating Sentinel satellite imagery analysis, in two phases using multi-sensor satellites, and machine learning methods. In the first step, wind speed and bathymetry have been analysed by means of sentinel-1 (S-1) and sentinel-2 (S-2) satellites images, respectively. Furthermore, a hybrid forecasting model has been proposed to assess and predict wind speed. The machine learning model consists of an integrated model using generalized regression neural network (GRNN) and the whale optimization algorithm (WOA). The new developed method has been then applied to assess wind energy potential around the Favignana island in Sicily, Italy. The results show that all the important primary parameters for a wind farm installations potential analysis, such as wind speed, water depth, and distance to the shoreline can be successfully analysed in a possible short time and freely. The following results have been obtained testing the new method: i) S-1 and S-2 satellite images have good potential for near and offshore wind speed assessment and bathymetry detection around Favignana island, ii) the root mean square error (RMSE), mean square error (MAE), and mean absolute percentage error (MAPE) of the proposed forecasting model using the whole dataset are 0.0205, 0.0159, and 6.8385 respectively, iii) the proposed forecasting model for wind speed has higher accuracy than other valid models.