基于二次分解特征矩阵和PCNN-BiLSTM的短期电价预测

Short-term Electricity Price Forecasting Based on Secondary Decomposition Feature Matrix and PCNN-BiLSTM

为了提高短期电价预测的精度,提出了1种基于二次分解特征矩阵、并行卷积神经网络(Parallel convolutional neural network,PCNN)、双向长短期记忆神经网络(Bi-directional long short term memory,BiLSTM)的预测方法。采用完全集合经验模态分解将归一化后的原始电价/负荷数据分解为一系列分量,用变分模态分解将第1次分解产生的最高频分量进一步分解成一系列模态分量;用第1次和第2次分解产生的所有分量构造2通道输入特征矩阵;利用PCNN提取各种特征,再将特征融合后输入到BiLSTM预测结构中,最终得出翌日预测值。预测结果表明,所提出的预测方法有效提高了短期电价的预测精度。

To improve the accuracy of short-term electricity price forecasting,a forecasting method based on secondary decomposition feature matrix,parallel convolutional neural network(PCNN),and bi-directional long short-term memory neural network(BiLSTM)is proposed.The normalized raw electricity price/load data are decomposed into a series of components by using full ensemble empirical modal decomposition,and the highest frequency components from the first decomposition are further decomposed into a series of modal components by variational modal decomposition;the input feature matrix of 2-channel is constructed by using all components from the first and second decompositions;the features are extracted by PCNN,and input into the BiLSTM prediction structure after feature fusion,finally the forecast value of next day is obtained.The prediction results show that the proposed forecasting method effectively improves the prediction accuracy of short-term electricity prices.