基于attention机制的LSTM神经网络超短期负荷预测方法  

A LSTM Neural Network Method Based on Attention Mechanism for Ultra Short-term Load Forecasting

作　　者：李昭昱 艾芊[1] 张宇帆 肖斐[1] LI Zhaoyu;AI Qian;ZHANG Yufan;XIAO Fei(School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《供用电》2019年第1期17-22,共6页Distribution & Utilization

基　　金：国家自然基金-国家电网联合基金(U1766207)~~

摘　　要：随着电力系统中分布式能源及可调控柔性负荷等的增多,电力系统负荷的随机性增强,且电力系统智能化发展,使得可获得数据急剧增长。为充分利用历史负荷数据,提高超短期负荷的预测精度,提出一种基于attention机制的LSTM超短期负荷预测方法。首先,通过分析负荷数据的自相关性,选取预测点前168 h的负荷数据作为网络输入,并针对坏数据进行辨识修正,随后通过Adam算法实现网络训练。最后,通过与标准BP神经网络的预测结果的MAPE指标进行对比,验证了所提方法的可行性及效果。With the increase of distributed energy and regulatable flexible load in power system,the randomness of power load is enhanced,and the intelligent development of power system makes the available data increase rapidly.In order to make full use of historical load data and improve the accuracy of ultra-short-term load forecasting,a LSTM method based on attention mechanism for ultra-short-term load forecasting is proposed.Firstly,by analyzing the autocorrelation of load data,the load data of168hours before forecasting point is selected as network input,and the bad data is identified and corrected.Then the network training is realized by Adam algorithm.Finally,the feasibility and effect of the proposed method are verified by comparing the MAPE indexes of the predicted results of the proposed method and the standard BP neural network.

关 键 词：超短期负荷预测 LSTM网络 attention机制 相关性 标准BP神经网络 

分 类 号：TM74[电气工程—电力系统及自动化]