Research On Wind Power Prediction Model Based On Deep Learning

Wind power prediction system is used to predict the power generation capacity of wind power plant in the future,which is convenient for the grid to make power generation plan.The system is necessary condition for wind power plant to generate electricity.At present,the prediction accuracy of most domestic power prediction systems cannot meet the requirements of grid dispatching center.In order to improve the accuracy of wind power prediction,the application of deep learning in wind power prediction model was studied,and the error of power prediction model was reduced from the perspectives of data quality control,numerical weather prediction matching,and power prediction model design.In addition,aiming at the problem that few wind power plants can't generate electricity caused by congelation in winter,the congelation prediction method based on LSTM is proposed.The research achievement of this study can improve the power generation capacity of the wind power plant,which have high practical application value.The main research contents are as follows:(1)Aiming at the problem of a large amount of abnormal data and missing data in the historical data of wind power plant,the data quality control process including data anomaly detection and data missing value filling is studied.The method of data anomaly detection based on LSTM neural network is studied and the result was compared with other methods based on machine learning algorithm.The missing data filling method based on multi-view learning is studied and the result was compared with the traditional interpolation method and the missing data filling method based on ARIMA model.(2)Aiming at the low accuracy of existing wind power prediction systems,the power prediction model based on numerical weather forecasting is studied.Due to the poor pre-universality of numerical meteorology,a numerical meteorological matching method based on similarity ranking is proposed.A power prediction model based on four-layer BP neural network is designed.Compared with other models,this model establishes a more accurate mapping relationship between wind and power,which greatly reduces the errors caused by the model.(3)Aiming at the problem that few wind power plants can't generate electricity caused by congelation in winter,the congelation prediction method based on LSTM neural network is studied.Due to the temperature,humidity and other factors that directly affect the probability of congelation in numerical weather prediction are inaccuracy,the real-time measurement data of wind tower of wind power plant are used in LSTM neural network model to predict temperature and humidity data in the future.Experimental results show that compared with the time series prediction method based on ARIMA,the prediction results of LSTM network have a higher degree of fitting with the actual data.The power prediction model studied in this paper has been tested in many wind power plants,with an average accuracy of over 83%,which is relatively high in China.The congelation prediction model in this paper can accurately predict the temperature and humidity in the future.According to the prediction results,the wind power plant can reasonably arrange the starting time of the wind generator's heating system to prevent the losses caused by the wind generator's congelation.The practical application research on this aspect is the first in China.