基于DFT-DTW-k-means++与CNN-BiGRU电力数据降噪与负荷预测OA

In response to the issue that traditional power data analysis is prone to interference from cross-domain information,resulting in low accuracy of cross-domain load forecasting,a data denoising and forecasting model based on discrete Fourier transform and dynamic time warping(DFT-DTW)-k-means++combined with convolutional neural network and bidirectional gated recurrent unit(CNN-BiGRU)is proposed.DFT is used to denoise the power load data.The k-means++algorithm is adopted to cluster and divide the power load data,and DTW is used to measure the cross-domain power load data.The load forecasting model of CNN-BiGRU is constructed to forecast power load.The results show that the time domain signal obtained by using the denoising method of DFT is obvious.Under different forecasting models such as random forest regression,the root mean square error value of the data partitioning method of DFT-DTW-k-means++is low.Compared with the bidirection-al long short-term memory(BiLSTM)forecasting model,support vector regression prediction model,etc.,the average mean square error value of the proposed forecasting model is 0.085,which is lower than that of other forecasting models.This indi-cates that the proposed model can achieve cross-domain load data denoising and improve the prediction accuracy,thereby enhan-cing the utilization effect of power data.