Method selection in different regions for short-term wind speed prediction in Japan

Recently, many approaches have been proposed in wind speed prediction. However they are verified only in certain areas, and have not been a quantitative verification in many different locations. Therefore, this paper used the data of the various parts of Japan for the one-step-ahead prediction, and applied a number of different approaches to each point. And then it was evaluated such as MAE. These models are persistent model, ARMA-GARCH model, non-linear autoregressive network with an external input (NARX), recurrent neural network (RNN), and support vector regression (SVR). From the results of the numerical simulation at each point, this paper presents the results that it is difficult to create the same model which minimize the error in all areas, and there is a need to create a predictor for each region.