基于Bi-LSTM网络的时变综合负荷模型参数辨识  

Bi-LSTM-based Time Varying Parameter Identification forComposite Load Modeling

作　　者：陈谦[1] 冯源[1] 陈嘉雯 徐旸 CHEN Qian;FENG Yuan;CHEN Jia-wen;XU Yang(Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学,电气与动力工程学院,江苏南京211100 [2]国网江苏省电力有限公司盐城供电分公司,江苏盐城224000

出　　处：《电力电子技术》2024年第11期67-71,共5页Power Electronics

基　　金：国家自然科学基金(51837004)。

摘　　要：考虑到实际电网负荷的组成会随着系统运行方式、环境状况等因素发生变化,以及各类分布式电源的接入,负荷模型中增加了具有各种时变特性的负荷分量,对其进行参数辨识的难度日益加大。这里提出了一种基于深度学习的时变参数辨识模型,用于综合负荷模型时变参数的辨识。采用两个并行的双向长短期记忆(Bi-LSTM)网络,利用时变参数以及有功、无功功率和正序电压的时序特性,综合考虑它们对时变参数的影响,并在系统测量范围的情况下,辨识综合负荷模型的所有时变参数。Considering that the composition of the actual power grid load will change with factors such as system operation mode and environmental conditions,as well as the connection of various distributed power sources,load components with various time-varying characteristics have been added to the load model,making it increasingly difficult to identify their parameters.A time-varying parameter identification model based on deep learning is proposed for the identification of time-varying parameters of system-wide load model.Two parallel bidirectional long-term and shortterm memory(Bi-LSTM)networks are used to comprehensively consider the temporal characteristics of time-varying parameters,active power,reactive power and positive sequence voltage,then identify all time-varying parameters of the composite load model in the case of system measurement range.

关 键 词：负荷模型 时变参数 双向长短期记忆网络 

分 类 号：TM74[电气工程—电力系统及自动化]