Prediction Interval For Short-term Wind Speed In Wind Farm Based On GWO

With the increasingly prominent problems of energy security,environmental issues and climate warming,wind power as an important clean energy,with its renewable,environmental protection and other advantages is in the process of vigorous development.However,the wind speed has the characteristics of strong fluctuation,which has great influence on the safe and stable operation of the electrified wire netting,so it is important to predict the wind speed accurately or give the possible range of wind speed.Most of the existing wind speed prediction methods are point prediction,while the research on wind speed interval is not thorough.Considering the inevitable error of the current wind speed prediction methods,if wind speed interval prediction based on the point prediction can be accurately carried out,we can provide more information about the uncertainty of wind power for the wind power department and offer more reference information for the decision makers.Under this background,this paper adopts the comprehensive optimization objective function PPA as optimization criterion,and put forward the hybrid prediction model CEEMD-GWO-PPA-BP which based on BP neural network and GWO algorithm.The proposed model involves the following three steps:(1)decomposing the original short-term wind speed time series into several intrinsic mode functions(IMFs)and one residual with complementary ensemble empirical mode decomposition(CEEMD);(2)individually predicting each IMF and the residual with back propagation neural network optimized by GWO;(3)integrating all predicted IMFs and residual for the ensemble result as the final prediction by another BP neural network with multiple inputs and two outputs optimized by GWO which can obtain the upper and lower bounds of the prediction interval.In the third steps,I adopt the comprehensive optimization objective function which include prediction interval coverage probability,prediction interval normalized average width and accumulated width deviation as the optimization criterion.Besides,this paper also analyzes the model from three aspects: Data decomposition methods,optimization algorithms and optimization criterions.The empirical study indicates that the proposed hybrid model CEEMD-GWO-PPA-BP has a better performance than the contrast models.In the case of ensuring the prediction interval coverage probability and prediction interval normalized average width,the proposed hybrid model CEEMD-GWO-PPA-BP can obviously reduce the accumulated width deviation and obtain a more accurate clearer prediction interval.