基于LSTM与迁移学习的配电网负荷频率控制PI参数整定方法  

PI Parameter Setting Method for Distribution Network Load Frequency Control Based on LSTM and Transfer Learning

作　　者：郑舒 赵景涛 刘明祥 ZHENG Shu;ZHAO Jingtao;LIU Mingxiang(State Grid Electric Power Research Institute Co.,Ltd.,Nanjing 210000,China)

机构地区：[1]国网电力科学研究院有限公司,江苏南京210000

出　　处：《微型电脑应用》2024年第4期97-100,共4页Microcomputer Applications

基　　金：国网总部科技项目(5400-202219152A-1-1-ZN)。

摘　　要：在整定配电网络负荷频率过程中,控制器PI参数会受到局部控制迭代影响,在控制层与反馈层形成时间差量,进而造成频率里程偏差。对此,提出基于LSTM与迁移学习的配电网负荷频率控制PI参数整定方法。采用LSTM模型对调频(FM)里程参数进行模型量化,配合迁移学习对负荷频率控制相关参量进行差分更新,构建配电网负荷频率的LSTM模型,进行频率控制的迁移学习优化、PI参数控制约束及PI参数整定输出。实验结果表明,不论是在控制精度上还是在控制稳定性上,均具有良好的性能表现,整定结果满足预期。In the process of distribution network load frequency setting,the PI parameters of the controller are affected by local control iterations,forming a time difference between the control layer and the feedback layer,resulting in frequency mileage deviation.Therefore,a PI parameter setting method for distribution network load frequency control based on LSTM and transfer learning is proposed.The LSTM model is used to quantify the frequency modulation(FM)mileage parameters,and the related parameters of load frequency control are differentially updated with the transfer learning.The LSTM model of distribution network load frequency is constructed,and the transfer learning optimization of frequency control,PI parameter control constraints and PI parameter setting output are carried out.The experimental results show that it has good performance both in control accuracy and control stability,and the setting results meet the expectations.

关 键 词：LSTM 迁移学习 负荷频率控制 PI参数整定 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]