Research On Ultra-short-term Wind Speed Prediction Based On Improved Wavelet Least Squares Support Vector Machine

In recent years,clean and pollution-free renewable energy has developed rapidly in the increasingly severe energy environment.China strongly supports renewable energy,which is of great significance to improve the energy structure,achieve sustainable development,and promote the construction of the ecological civilization of the country.However,being a new type of renewable energy with a wide range of applications,the randomness and complexity of wind energy have greatly restricted the further development of wind power generation.Accurate wind speed prediction can not only increase the stability of wind power grid integration but also improve and effectively manage renewable wind power generation,maximizing the use of wind energy.At the same time,it ensures the stable output of wind power,as well as realizes the reasonable dispatching and safe operation of the power system.Consideration is given to the high non-determinacy and non-linearity of the wind speed series,a single model cannot meet the prediction accuracy any longer.Since this is the case,a novel hybrid prediction model,called MRSVD-SFLA-w LSSVM,was proposed in this paper.Among them,the wavelet least squares support vector machine(w LSSVM)was obtained by changing the traditional LSSVM kernel function into the wavelet kernel function and was taken as the basic prediction model.The multi-resolution singular value decomposition(MRSVD)and shuffled frog leaping algorithm(SFLA)is utilized to improve the predicting performance of w LSSVM from the decomposition and optimization respectively.One of the innovations of this paper is to apply MRSVD and single kernel w LSSVM to wind speed prediction,which makes up for the lack of prediction accuracy of existing algorithms.Another innovation of this paper is that the combined optimization model of MRSVD-SFLA-w LSSVM is constructed for the ultra-short-term wind speed prediction research for the first time.In order to verify the feasibility and superiority of this model in the ultra-short-term wind speed prediction,another12 contrastive models were established in this paper,and the actual wind speed data of the Weichang wind farm in Chengde City,Hebei Province were taken use to conduct a four-season empirical analysis,comparing the forecasting performance of all the models from the mean absolute percentage error(MAPE),the root mean square error(RMSE),the coefficient of determination(R~2),and their percentage improvement indexes.All the results showed that the three error indexes of the model are the best among all the models,declaring that the forecasting method and model proposed in this paper are feasible in the ultra-short-term wind speed prediction,and the model owns a better prediction performance compared with other models.Besides,the successful prediction of different season cases verifies the generality of the model,so it can be widely utilized in the future research of the ultra-short-term wind speed prediction,which provides a certain reference for promoting the development of wind energy and reducing the impact of wind on the power grid.