Research On Short-term Combined Prediction Model Of Wind Speed Based On Improved Neural Network

As green renewable energy,wind energy has become an important part of energy development.Wind power is one of the main ways of wind energy utilization.However,due to the volatility of wind energy,wind power integration will affect the operation,scheduling,cost and other aspects of the power system.Therefore,accurate prediction is important to the development of wind energy,as well as the safe and stable operation of power system.Improving the accuracy of short-term wind energy forecasting is the key to wind energy development and utilization and grid connection.Therefore,in order to improve the accuracy,this paper studies the characteristics of wind speed,related factors affecting wind speed,data preprocessing methods,model optimization,etc.,and proposes a variety of models to improve the effect of wind speed forecasting.The specific research contents are as follows:(1)Establish the CEEMD-IGA-FNN-Markov prediction model.A method for selecting the number of CEEMD decomposition based on the principle of minimum reconstruction error is proposed to obtain sufficient decomposition components and solve the problem of the uncertainty of the number of CEEMD decompositions.At the same time,the improved CEEMD is used to decompose the original wind speed data to obtain different time The components of the frequency characteristics can better learn the different characteristics of wind speed;an improved genetic algorithm is proposed to optimize the fuzzy neural network,and the model is used to make a preliminary prediction of the wind speed;K-Means is used to divide the error sequence,and Markov is used to The prediction results are revised to further improve the prediction accuracy.(2)Establish the CEEMD-PSR-PCA-SVM prediction model.Taking into account the chaotic characteristics of wind speed,the CC method is used to calculate the phase space reconstruction parameters for each component of the wind speed,and then the phase space reconstruction of each component is carried out to further extract the effective information in the wind speed data;at the same time,the meteorological terrain that affects the wind speed volatility is introduced.Environmental factors,and use the principal component analysis method to determine the appropriate model input variables;construct a chaotic phase space prediction model,establish a multivariable support vector machine prediction model for prediction,and improve the accuracy of wind speed prediction.(3)Establish an optimal combination forecasting model based on SSA-CEEMD.In order to further decompose wind power data and obtain wind power characteristics,the SSA-CEEMD method is used to decompose wind speed;at the same time,because different prediction models learn and train wind speed from different aspects,it is impossible to predict wind speed well in all prediction environments.Therefore,in view of the difference in the prediction results of a single model,a comprehensive analysis of the prediction results of multiple methods is proposed;an optimal comprehensive prediction model is proposed,and the optimal comprehensive prediction model is obtained through BPNN training.The prediction results are weighted and combined to obtain the best prediction results.