Wind Speed Interval Prediction Based On Deep Learning And Error Prediction

With the increasing depletion of fossil fuels and continuous air pollution,renewable energy has become an important development direction for now and a long time in the future.As a clean energy with advantages of small footprint,large storage capacity and low fuel consumption,wind energy has been given broad development prospects.To help power system optimization,backup system arrangements,and power grid scheduling decisions,it becomes particularly important to accurately predict the fluctuation range of wind speeds in the future.In view of this research direction,based on the traditional upper and lower bound evaluation methods,combined with the wind speed point prediction and error prediction,the wind speed interval prediction model based on the ensemble gated recurrent model and the interval construction model based on error prediction are established respectively.On the basis of these models,the structure is optimized and the training speed of the model is further improved while ensuring the performance of interval prediction.The main research work of this article includes:(1)A wind speed interval prediction model based on the ensemble gated unit model is proposed.The preliminary prediction interval is obtained through interval width presetting,interval decomposition and interval prediction.The error between the upper and lower bounds of the preliminary prediction interval and the preset interval is calculated.The error correction strategy is studied to further improve the prediction The quality of the interval.(2)An interval construction model based on error prediction is proposed.The original wind speed data is processed through the variational mode decomposition algorithm.The predicted values and error prediction values of each sub-mode are obtained through point prediction and error prediction.Each sub-mode prediction error is given a weight and the upper and lower bounds of the prediction interval is accumulated by the point prediction values and the error prediction values with the given weights.The weight of each sub-modal prediction error is calculated by solving the constrained single-objective optimization problem.(3)An improved interval prediction model based on error prediction is established,and the interval coefficient is constructed from the historical error data calculated from the wind speed point prediction results and the real wind speed data,and the wind speed interval is predicted by combining the wind speed point prediction value and the interval coefficient.In order to solve the problem that the prediction interval indices are not derivable,two objective functions are introduced in the interval prediction model to enable the model to be trained by the back propagation algorithm.