双通道多因素短期电力负荷预测模型  

Double channel multi-factor short-term power load forecasting model

作　　者：孙胜博 吴彬彬 陈晔 吴迪[2] SUN Sheng-bo;WU Bin-bin;CHEN Ye;WU Di(Marketing Service Center,State Grid Hebei Electric Power Limited Company,Shijiazhuang 050000,China;School of Information and Electrical Engineering,Hebei University of Engineering,Handan 056038,China)

机构地区：[1]国网河北省电力有限公司营销服务中心,河北石家庄050000 [2]河北工程大学信息与电气工程学院,河北邯郸056038

出　　处：《计算机工程与设计》2023年第6期1875-1884,共10页Computer Engineering and Design

基　　金：国家电网有限公司科技基金项目(5600-202019167A-0-0-00);河北省自然科学基金项目(F2020402003)。

摘　　要：针对传统负荷预测模型气象因素影响考虑不充分,特征获取能力欠佳的问题,提出一种BiLSTM-Att+CNN-CBAM双通道多因素短期电力负荷预测模型。利用灰色关联分析获取气象因素;采用皮尔逊相关系数筛选特征属性;构建BiLSTM-Att与CNN-CBAM双通道,分别获取全局时序特征与局部特征;使用全连接层进行特征拼接,通过Dense层输出预测结果。实验结果表明,该模型具有更优的均方误差(MSE)、均方根误差(RMSE)和平均绝对误差(MAE)。Aiming at the problems of insufficient consideration of meteorological factors and poor feature acquisition ability of the traditional load forecasting model,a BiLSTM-Att+CNN-CBAM dual channel multi-factor short-term power load forecasting model was proposed.The meteorological factors were obtained by grey correlation analysis.Characteristic attributes were screened by Pearson correlation coefficient.Acquisition of global temporal features and local features were obtained by constructing BiLSTM-Att and CNN-CBAM dual channels,respectively.The fully connected layer was used for feature stitching and the predictions were outputted through the Dense layer.Experimental results show that the proposed model has better mean square error,root mean square error and mean absolute error.

关 键 词：电力负荷预测 双通道 多因素 灰色关联分析 皮尔逊相关系数 注意力机制 特征拼接 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]