基于CNN-BiLSTM-Attention的超短期电力负荷预测  

Ultra short term load forecasting based on CNN-BiLSTM-Attention

作　　者：宋珊珊 潘文林 王嘉梅[1] 梁志茂[1] SONG Shan-shan;PAN Wen-lin;WANG Jia-mei;LIANG Zhi-mao(School of Electrical Information Engineering,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学电气信息工程学院,云南昆明650500

出　　处：《云南民族大学学报（自然科学版）》2022年第2期235-240,共6页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(41961053).

摘　　要：在长短期记忆神经网络(LSTM)的基础上,运用双向的长短期记忆神经网络(BiLSTM),结合卷积神经网络(CNN)提出了一个预测模型,对超短期电力负荷预测.运用合肥市2019年全年数据对该模型进行训练及预测,结果显示,CNN-BiLSTM预测精度高于CNN-LSTM预测模型,为进一步提升预测精确度,在BiLSTM神经网络后面连接了一个Attention在输出,发现其预测精度进一步提升了.Based on the requirements of feature extraction and timing dependence in the neural network prediction model,this study proposes to upgrade to a bidirectional long and short-term memory neural network on the basis of the long-and short-term memory neural network,and combine it with a convolutional neural network.Forecast model is created to achieve ultra-short-term power load forecasting.On the premise of scientifically selecting feature vectors,and using Hefei s 2019 annual data to train and predict the model,the results show that its prediction accuracy is higher than that of CNN and LSTM(Long Short Term Memory Neural Network)prediction models.For prediction accuracy,an Attention is connected to the output behind the BiLSTM neural network,and it is found that the prediction accuracy is further improved.

关 键 词：超短期电力负荷预测 卷积神经网络(CNN) 双向长短期记忆神经网络(BiLSTM) Attention机制 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]