Research On Short-term Wind Power Prediction Based On Combination Analysis Method

With the rapid development of the national economy,the continuous aggravation of pollution and the proposal and implementation of China's energy-saving and emission reduction economic development strategy,wind power has become an important part of renewable energy.Wind power is the power of wind turbine.Because of the characteristics of wind and the principle of power generation,the output of wind power is volatile and intermittent,so the grid-connected wind power will have a great impact on the power system.Wind power forecasting can effectively solve the problem of power grid dispatching,and plays an important role in wind power grid integration and safe operation of power grid.Aiming at the problem that the prediction accuracy of the existing prediction methods is not ideal,this paper adopts a short-term wind power prediction method based on combination analysis.This method uses the generative adversarial network as the system framework.Compared with the traditional neural network,this method is trained repeatedly by the interaction of the generative model and the discriminant model.Therefore,compared with the single prediction method,this method is more effective.The prediction accuracy of this method is high.In view of the chaotic characteristics of wind energy,such as disorder,randomness,instability and uncontrollability,chaotic networks can better represent the internal relationship between wind energy parameters and wind power than other types of neural networks.Therefore,this paper chooses chaotic neural network as the generation model of generative adversarial neural network algorithm.In order to ensure the prediction accuracy of the combination method,BP neural network with good fitting performance of the non-linear function is selected as the discriminant model of the generative adversarial network.This paper chooses the measured data of a wind farm in Ningxia,China,for training and forecasting experiments.The time series of wind power output and local wind energy data are selected as training samples to predict the short-term wind power of the wind farm for the next 72 hours.The experimental results are compared with the results of a single chaotic neural network algorithm.The combined forecasting method based on the framework of generative countermeasures network is compared.Single prediction method can effectively improve the prediction accuracy.There are many input data variables in wind power forecasting,which will increase the computational complexity of forecasting algorithm and reduce the forecasting efficiency.In this paper,the input variables of wind power forecasting are analyzed,and it is found that there are large correlations and redundancies among many variables.In order to reduce the number of input variables of wind power prediction on the basis of guaranteeing the correctness of prediction,this paper uses mutual information theory to do correlation analysis and redundancy elimination of input variables of wind power prediction,thus reducing the number of input variables of the algorithm,thereby reducing the complexity of the algorithm and improving the efficiency of the algorithm.In this paper,a comparative experiment is carried out before and after the input variables are de-redundant.The experimental results show that the computational efficiency is improved while the input variables are reduced,and the prediction accuracy of the original data is guaranteed.