Research Of Short-term Wind Power Forecast Based On BP Neural Network Optimized By Improved Particle Swarm Optimization

As the installed capacity of wind power in China continues to increase,the proportion of wind power in the power grid is also rising.Effective wind power forecasting is an indispensable key measure for solving large-scale wind power grid integration,which can ensure the steady operation of the power system.The essence of wind power prediction is to establish a non-linear function mapping relationship between the influencing factors of the wind farm power and the output power of the actual wind farm.BP neural network has an strong learning generalization ability,and it has a great advantage in dealing with the extraction and approximation of nonlinear functions.This algorithm is suitable for solving wind power forecasting problems.However,there are some defect in BP neural network: it is easy to sink into local optimum,and the convergence speed is relatively slow.In this thesis,an improved particle swarm optimization algorithm is utilized to optimize the initial parameters of BP neural network,which can effectively compensate the defect of BP neural network.Then the wind power forecasting model is established by using the parameter optimized BP network.The main research contents of this thesis include:(1)Based on the dynamic development of wind power at home and abroad,the research background and significance of the topic of this thesis are discussed.The research status of the short-term power forecasting of wind farms at home and abroad is fully expounded,and the classification of short-term wind power forecasting technology is further discussed.(2)The main parameters affecting the output power of the wind farm are analyzed,and the pretreatment measures for the wind power data are introduced.(3)BP neural network algorithm has strong learning ability and can effectively deal with nonlinear problems.This algorithm is suitable for wind power prediction.The wind power forecasting model based on BP neural network algorithm is established.Then the parameter setting of the model is completed,and the short-term wind power forecast of the wind farm is implemented,and the prediction performance of the model is analyzed.(4)Due to the deficiencies of the BP neural network,the wind power forecasting effect is poor.This thesis uses the improved particle swarm optimization algorithm to optimize the initial parameters of the BP network and a combined prediction model based on the improved particle swarm-BP neural network algorithm is established.Firstly,based on the basic particle swarm algorithm,this thesis proposes an improved method based on the nonlinear decreasing of inertia weights.It can improve the shortcomings of the standard particle swarm algorithm by the nonlinear degressive weighting method and effectively improve the convergence speed and search performance of the particle swarm optimization algorithm.Then the improved particle swarm optimization algorithm is utilized to optimize the initial parameters of the BP neural network,which can effectively enhance the performance of the BP neural network,and then the wind power prediction model is established by using the parameter optimized BP network.The simulation results show that compared with the BP model and the BP model optimized by the basic particle swarm optimization algorithm,the BP model optimized by the improved particle swarm optimization algorithm has an better prediction accuracy,and the fitting degree between the predicted curve and the actual output power curve is more better.The example simulation results verify that the proposed method is feasible and effective for wind power prediction.