Short Term Wind Speed Prediction Model Based On Hybrid Algorithm

With the rapid growth of population and the rapid development of society,the demand for energy is increasing day by day,and the development of traditional non-renewable energy is facing enormous pressure.Improving the development and utilization of renewable energy has become a key way to alleviate the current energy crisis.As a kind of renewable energy with wide distribution,abundant resources,clean and pollution-free,wind energy has good social benefits,economic competitiveness and environmental friendliness,and has become one of the fastest-growing renewable energy sources in the world.However,the nonlinearity and instability of wind speed will lead to unstable power generation of wind turbines,affecting the quality of output wind power,bringing great challenges to the grid connection of wind power,and also affecting the operation safety of wind turbines.Therefore,it is of great practical significance to accurately predict the wind speed of wind farms to increase the power of wind power generation and reduce the risk and cost of wind power grid connection.At present,the application research of wind speed prediction models for wind farms at home and abroad is to continuously improve the accuracy of the prediction models for traditional prediction models,and to further improve the performance of prediction models by combining the advantages of different algorithms with combined prediction models.In this thesis,the wind speed data recorded every 1h from a wind farm in Iowa and a domestic meteorological monitoring station are taken as the research objects,and the predictions are made from data preprocessing,multiple single prediction models,single model and hybrid algorithm for parameter optimization.Four aspects of the model make detailed analysis and comparison of short-term wind speed forecasting models.In this thesis,the non-linearity and non-stationarity of wind speed time series are firstly analyzed,and the preprocessing method of wind speed data is deeply analyzed.The algorithm principles and steps of Empirical Mode Decomposition(EMD),Ensemble Empirical Mode Decomposition(EEMD)and Variational Mode Decomposition(VMD)are introduced respectively,and two wind speed data sets are used as experimental objects to compare the decomposition of different methods result.For the selection of basic forecasting algorithms for short-term wind speed forecasting models,this thesis studies the theoretical basis and algorithm steps of back-propagation neural network,deep belief network and longshort-term memory neural network respectively.Prediction performance comparison.At the same time,for the optimization of the input parameters of the back-propagation network model,the sparrow search algorithm(SSA)is introduced to improve the parameter input of a single prediction model,and an improved sparrow search algorithm(ISSA)is proposed to improve the initialization population strategy with chaotic mapping and nonlinear weight update strategy,which improves the global search ability of optimal parameters and the late convergence speed.In this thesis,a hybrid model for short-term wind speed prediction based on VMD-ISSA-BP is proposed,and VMD-SA-DBN,VMD-GWO-BP,VMD-ISSA-LSTM and other single models and hybrid models are constructed for simulation experiments and comparisons.The VMD-ISSA-BP models have better prediction performance and algorithm stability.In this thesis,two sets of historical wind speed data sets are used to conduct experiments.The proposed comprehensive short-term wind speed prediction model based on VMD-ISSABP can predict the wind speed data after 1 hour with 24 wind speed information at intervals of 1 hour.The prediction accuracy indicators RMSE,MAR and MAPE are respectively0.205m/s,0.165m/s,0.0258.Compared with a single prediction model,the prediction performance is improved by 70%-80%,the prediction performance of other comprehensive models is improved by about 15%,and the model has better prediction performance and algorithm stability.