基于改进LSTM的区域综合能源系统多元负荷短期预测研究  

Research on Multi-load Short-term Forecasting of Regional Integrated Energy System Based on Improved LSTM

作　　者：田浩含 张智晟 于道林 TIAN Haohan;ZHANG Zhisheng;YU Daolin(College of Electrical Engineering,Qingdao University,Qingdao 266071,China;State Grid Laiyang Power Supply Company,Laiyang 265200,China)

机构地区：[1]青岛大学电气工程学院,青岛266071 [2]国网莱阳市供电公司,莱阳265200

出　　处：《电力系统及其自动化学报》2021年第9期130-137,共8页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(52077108)。

摘　　要：冷热电负荷短期预测是区域综合能源系统优化调度的基础。针对区域综合能源系统多元负荷关联性和非线性的特点,本文构建了基于改进的长短期记忆神经网络的区域综合能源系统多元负荷短期预测模型,该模型采用灰色关联度法分析多元负荷之间和气象因素之间的耦合性,以此为依据,在改进长短期记忆神经网络预测模型中加入注意力层和dropout层,注意力机制可赋予模型隐含层不同的权重,dropout层可对模型正则化,并采用粒子群优化算法对预测模型参数进行优化。算例仿真结果表明,本文提出的预测模型具有较好的预测精度。The forecasting of cold,heat and electric loads is the basis for the optimization and regulation of a regional integrated energy system.According to the characteristics of correlation and nonlinearity of multi-load in this system,a multi-load short-term forecasting model of the regional integrated energy system based on improved long short term memory(LSTM)is constructed,which uses the grey correlation method to analyze the coupling between multiple loads and that between meteorological factors.On this basis,the attention and dropout layers are added to the improved LSTM prediction model,in which the attention mechanism can give different weights to the hidden layer and the dropout layer can regularize the model.In addition,the particle swarm optimization algorithm is used to optimize the parameters of the prediction model.The simulation results of a numerical example show that the proposed prediction model has a higher prediction accuracy.

关 键 词：区域综合能源系统 多元负荷短期预测 粒子群优化算法 注意力机制 长短期记忆神经网络 

分 类 号：TM715[电气工程—电力系统及自动化]