Design Of Wind Power Forecasting System Based On Improved Neural Network

At present,we are experiencing the transition period from the information technology era to the energy technology era.Wind energy is becoming the main new energy because of its abundant reserves and clean energy.However,the randomness and intermittence of wind will hinder the development of wind power.Therefore,the accurate prediction of wind power is obviously more and more important.In order to meet the higher requirements of the "two rules" issued by the National Energy Administration for wind power forecasting,and also provide technical support for the power trading in Northwest China.In this paper,the design of wind power prediction system is studied as follows:Firstly,the NWP data,the wind tower data and the actual power data in the wind power prediction are tested and processed by repeated rows,integrity and rationality respectively.Ensure to get high-quality data to provide good support for the prediction of wind power.Mean absolute error,root mean square error and error index correlation coefficient are selected to evaluate the error of wind power prediction.Then,based on the advantages of simple structure and clear learning steps of BP neural network algorithm,BP neural network is selected for short-term prediction of wind power.However,due to the shortcomings of BP neural network itself,such as easily falling into local minimum,immune genetic algorithm based on standard particle swarm optimization is selected to retain the characteristics of global optimization.Finally,in order to weaken the influence of the strong fluctuation of wind energy on the short-term power prediction,the "average value method" is used to optimize the input data in the PSO-IGA-BP prediction algorithm model.Through the comparison between the predicted power curve and the actual power curve simulated by the model,as well as the error evaluation and analysis of the prediction results,the simulation results are compared.It is proved that the PSO-IGA-BP prediction model optimized by "average value method" has the highest prediction accuracy and the best prediction effect for short-term wind power.Finally,based on the previous research and verification,the wind power prediction system is designed and implemented.Through the functional design of each part of the prediction system and the design of data interface and communication mode,the wind power prediction system can realize multiple functions stably and conveniently,and then provide technical support for dispatching end and power trading users.