Wind Speed Prediction Based On Neural Network And Time Series

The increasing shortage of fossil energy and environmental pollution has become one of the main problems facing the world,which limits the social and economic development.The problems of energy shortage and environmental pollution can be alleviated effectively by developing the wind energy.The wind energy is developing rapidly because of its clean,efficient,low-cost and renewable.The fluctuation,intermittence and uncontrollability of wind energy have a serious impact on the safe and stable operation of power grid,so the wind speed prediction of wind farm can effectively reduce the cost of wind power grid connection and improve the security and stability of the grid.In order to improve the accuracy of wind speed prediction,the actual wind speed of wind farm is taken as the research object in this paper,the chaotic characteristics of the time series is analyzed,the wind speed time series model is reconstructed by the phase space method,the parameters of the chaotic time series model is optimized by the long-term and short-term memory neural network(LSMT),and the wind speed prediction proposed algorithm is verified in the actual wind farm.Firstly,the general model of wind speed is established,and the time series fluctuation characteristics of wind speed and wind power are analyzed.The wind speed data are pretreated by the natural spline interpolation method and normalization method to solve the the problem of missing and abnormal time series data,which provides the basis for the subsequent neural network and chaotic time series wind speed prediction model.Secondly,the chaos attribute of original data is identified according to the fluctuation of wind speed time series data,and the delay time and embedding dimension are determined by phase space reconstruction technology,which can reconstruct the wind speed prediction model of first-order weighted chaotic time series.Taking the wind speed time series data of a wind farm as an example,it shows that the chaotic time series model can better describe the wind speed fluctuation.Finally,aiming at the problem that the parameters of the first-order weighted chaotic time series model are adjusted difficultly,the LSTM neural network is used to.optimize the model parameters for wind speed prediction of wind farm.The wind speed prediction method based on LSTM neural network and chaotic time series is verified,it shows that the proposed method in this paper has a higher tracking accuracy and faster tracking speed compared with BP neural network,ARIMA model and first-order chaotic mathematical model.