Research On Combination Model Based On EWT Decomposition And DE-GWO Hyperparameter Optimization For Short-term Wind Speed Forecasting

With the large-scale development and utilization of fossil fuel and serious pollution it caused,wind energy,as a clean renewable energy,had exhibited great potential to develop and utilize.However,the nonstationary and intermittent properties of wind speed cause the engineering problems of the integration of wind power into a multisource energy network.In order to solve this problem,researchers have proposed many solutions,among which short-term wind speed forecasting is an important solution.Referring to the research status locally and abroad,this paper proposes a combination model based on Empirical Wavelet Transform(EWT)decomposition and Differential Evolution-Grey Wolf Optimizer(DE-GWO)hyper-parameter optimization for forecasting short-term wind speed.Firstly,the decomposition-based hybrid model is optimized by using EWT which can adaptively represent the processed signal and has complete mathematical theory as decomposition method and by replacing the single forecasting method with combination forecasting method.So,the combination forecasting model based on EWT decomposition which can matches the characteristics of signals in different frequency bands with the application range of prediction method is established.Then through three groups of comparative experiments,it is demonstrated that EWT decomposition and combination prediction method are better than traditional decomposition method and single prediction method for short-term wind speed forecasting.the prediction effect of the model is verified.Next,the hyper-parameters in the prediction methods Least squares support vector machine(LSSVM)and Gaussian Process Regression(GPR)need to be set manually,and improper setting of these parameters will affect the prediction results.Then Differential Evolution(DE)algorithm has advantages of keeping of multiplicity for production and enhancing the capability for local search.Grey Wolf Optimizer(GWO)has good convergence and avoidance of local optima.And DE-GWO algorithm has the advantages of DE and GWO at same time.So this paper selected DE-GWO hyper-parameter optimization method to optimize the hyper-parameter of LSSVM and GPR,and established the forecasting model based on DE-GWO hyper-parameter optimization.By comparing with the control model,the effect of DE-GWO hyper-parameter optimization method on LSSVM and GPR is proved.Finally,this paper combines the combination forecasting model based on EWT decomposition with the forecasting model based on DE-GWO hyper-parameter optimization,and obtains the combination forecasting model based on EWT decomposition and DE-GWO hyper-parameter optimization of short-term wind speed.The model is compared to 13 control models to systematically verify the predication effect of the model.