A modified game theoretic self-organizing map for wind speed forecasting

This paper proposes a hybrid wind speed forecasting framework using a modified game theoretic self-organizing map (GTSOM) clustering method. Cluster selection using correlation analysis is used to determine the inputs to Bayesian neural networks (BNN), which are used to forecast wind speed. Iowa wind speed data are used to assess the performance and accuracy of the proposed forecasting method. Furthermore, a comparison of the forecast results using the proposed clustering method with those of the K-means and traditional self-organizing map clustering shows an improved accuracy of wind speed forecasts.