基于EWT-CNN-BiGRU的多特征电力负荷预测模型

针对目前多特征预测模型在短期电力负荷数据上精度不足的问题,提出一种基于经验小波变换(EWT)的卷积神经网络(CNN)融合双向门控循环单元(Bi GRU)预测模型。首先,从多维时序数据中提取强关联性特征,其次,对选定特征进行经验小波变换,将时序数据映射至频域以获取子序列,最后,通过卷积神经网络和双向门控循环单元融合模型实现对电力负荷数据的预测。该预测模型使用德国某联合循环电厂的时序数据进行实验验证。结果表明,该预测模型获得99.463%的拟合优度,具有较好的预测效果。

In order to solve the problem of insufficient accuracy of multi-feature forecasting models in short-term power load data,a convolutional neural network(CNN) fusion bi-directional gated cycle unit(Bi GRU) forecasting model based on empirical wavelet transform(EWT) is proposed.Firstly,strong correlation features are extracted from multi-dimensional time series data,and then the selected features are transformed by empirical wavelet transform,and the time series data are mapped to the frequency domain to obtain sub-sequences.the prediction of power load data is realized by convolution neural network and bi-directional gated cycle unit fusion model.The prediction model is verified by experiments using the time series data of a combined cycle power plant in Germany.The results show that the prediction model has a goodness of fit of 99.463% and has a good prediction effect.