基于PCA-DBILSTM的多因素短期负荷预测模型  

Multi-factor Short-term Load Prediction Model Based on PCA-DBILSTM

作　　者：李泽文[1] 胡让 刘湘 邓裕文 唐鹏 王杨帆 LI Zewen;HU Rang;LIU Xiang;DENG Yuwen;TANG Peng;WANG Yangfan(Engineering Center for Power System Security and Supervisory Control Technology(Ministry of Education),Changsha University of Science and Technology,Changsha 410114,China;Yueyang Power Supply Branch,State Grid Hunan Electric Power Company Limited,Yueyang 414000,China;Loudi Power Supply Branch,State Grid Hunan Electric Power Company Limited,Loudi 417000,China)

机构地区：[1]长沙理工大学电网安全监控技术教育部工程研究中心,长沙410114 [2]国网湖南省电力有限公司岳阳供电分公司,岳阳414000 [3]国网湖南省电力有限公司娄底供电分公司,娄底417000

出　　处：《电力系统及其自动化学报》2020年第12期32-39,共8页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(51877012);湖南省教育厅重点资助项目(18A121);湖南省研究生科研创新资助项目(CX2018B557)。

摘　　要：针对传统神经网络在短期负荷预测中预测精度不高、预测时间较长的问题,提出了一种基于主成分分析法和深度双向长短期记忆神经网络的短期负荷预测模型。该模型运用主成分分析法对原始多维输入变量组成的时间序列进行主成分提取,实现原始负荷的降维;然后通过深度双向长短期记忆网络结合Adamax优化算法,对提取的主成分序列和负荷实际输出序列之间的非线性关系建立网络模型。以中国某地区的负荷数据作为实际算例,验证该方法预测精度达到了99.44%,并与传统预测模型进行对比,在保证预测精度的同时,大幅降低了预测时间。Aimed at the problem that the traditional neural network has low prediction accuracy and long prediction time in short-term load prediction,a short-term load prediction model based on principal component analysis(PCA)and deep bi-directional long-short-term memory(DBILSTM)neural network is proposed.This model uses PCA to ex⁃tract the principal components of the time series that consists of the original multi-dimensional input variables,thus achieving the dimensionality reduction of the original load.Then,the DBILSTM network is combined with the Adamax optimization algorithm to extract the nonlinear relationship between the extracted principal component sequence and the actual output sequence of load.As a result,the network model is established.The load data of one area in China are tak⁃en as an example,and the prediction accuracy of this method reaches 99.44%.Compared with the traditional prediction model,the prediction time of the proposed method is greatly reduced while ensuring its prediction accuracy.

关 键 词：主成分分析 双向长短期记忆网络 时间序列 负荷预测 Adamax算法 

分 类 号：TM71[电气工程—电力系统及自动化]