Research Of Wind Speed Prediction Method Based On Data Multi-stage Decomposition And Extreme Learning Machine

In the face of the rapid depletion of traditional fossil fuels and the deterioration of the environment,green and clean renewable energy has received widespread attention worldwide.As an important renewable energy resource,wind energy has the advantages of abundant reserves,wide distribution and no pollution.In wind energy development,due to the strong randomness,intermittentness and volatility of wind energy,the planning,scheduling,maintenance and control of wind energy systems depend on reliable wind speed prediction.Therefore,the accuracy of wind speed prediction is improved.The system plays a key role in adjusting the dispatch schedule,reducing spare capacity,reducing operating costs,and improving power quality.The wind has strong randomness,intermittentness and volatility.The nonlinear features contained in the original wind speed sequence increase the difficulty of model prediction.Therefore,in order to deal with the non-stationary nature of wind speed data,many studies use a variety of different signal decomposition techniques to pre-process the original data before prediction,which can improve the prediction accuracy to some extent.However,further research has found that data decomposition preprocessing can not fully exploit the characteristics of the original wind speed signal when decomposing the original data,which means that the single signal decomposition method has certain limitations in dealing with complex wind speed sequences.Therefore,it still exists.Large space for improving the accuracy of wind speed prediction.In response to the above problems,the research in this paper is mainly from the following aspects:1)The nonlinear variation of wind speed increases the difficulty of wind speed prediction,so it is very important to understand the change of wind speed.The levy is critical to the accurate prediction of wind speed.In this paper,the relationship between wind formation,wind speed characteristics,wind speed data acquisition and wind speed and wind power is analyzed,which lays a foundation for subsequent work.2)Different types of wind speed prediction models are studied,and the wind speed common prediction model is analyzed in detail and pointed out.The advantages and disadvantages of the prediction model are described in detail,and the theory and modeling steps of different wind speed prediction methods are described in detail.3)Aiming at the problems existing in the previous wind speed prediction model,a combined prediction model based on multi-level decomposition preprocessing and whale algorithm optimization extreme learning machine is proposed.In the first step,the original wind speed sequence is decomposed into a series of subsequences using the Empirical Wavelet Transform(EWT),and the Winner method is used to select the subsequence with the largest error.Further,the sub-sequence of the maximum error is decomposed by the variational mode decomposition to reduce the non-stationarity in the decomposition sequence,thereby improving the accuracy of the model prediction.Finally,all the decomposition modes(Mode)and variational modes(Vm)are predicted using an extreme learning machine optimized by the whale algorithm.In order to evaluate the predictive ability of the proposed model,this paper collected the wind speed data of three different wind farms in Ningxia.The selection of the baseline model,the existing single model and the traditional combination model are used for model prediction and comparison.The results of four error coefficients evaluation show that compared with the traditional model,the model established in this paper has higher prediction accuracy,indicating that the proposed model has Strong data mining capabilities and superior predictive performance.