IMPROVED WEIGHTED GRU WIND POWER INTERVAL PREDICTION BASED ON QUANTILE REGRESSION

In order to improve the accuracy of short-term wind power prediction, an improved entropy weighted GRU wind power interval prediction model based on quantile regression(QR-EGRU)is proposed to solve the problems that point prediction is difficult to describe wind power uncertainty and the traditional GRU cannot accurately track the data mutation when data changes. Firstly, the improved adaptive wavelet threshold denoising method is used to reduce the noise of the original data. Then two update gate weight matrices are introduced to replace the traditional update gate weight matrix. The new weight matrices adopt information entropy to dynamically adjust the matrix changes, quantify the change degree of the weights, and construct the information entropy weighted GRU (EGRU) network. Finally, the probability interval under different quantiles of the point prediction is obtained based on quantile regression algorithm. Experimental results show that the proposed model can improve the prediction accuracy compared with other comparison methods under the same experimental conditions and has a better interval prediction performance. © 2024 Science Press. All rights reserved.