Wind power forecasting model fusion evaluation based on comprehensive weights

Studies show that fusion modeling can improve the forecasting accuracy of wind power. Fusion modeling is the process of selective use of information from individual forecasting models. The reasonable evaluation of the individual models is the premise and basis of model optimization so that the individual models with high forecasting accuracy can be selected to establish the fusion model. Because the results of a single index model evaluation may not be comprehensive, the multi-index fusion evaluation method based on maximizing deviations and subjective correction is proposed. The method is applied to the selection of short-term wind power forecasting models. Firstly, this method establishes the individual model base of wind power forecasting model. Secondly, it establishes the more comprehensive evaluation index system. Thirdly, it combines maximizing deviations with the subjective correction coefficient to determine the comprehensive weight of each model, which is used to calculate the fusion evaluation value and get the evaluation order to achieve the model optimization. Finally, based on five years of data from a wind power plant in Shanxi Province, the validated experiments by multiple sets of forecasting data have been done using MATLAB in this paper. The simulation results demonstrate that the evaluation based on the proposed fusion evaluation method is more comprehensive and stable compared to evaluation using a single index. More importantly, it can effectively guide the model optimization with simple operating steps.