Koopman原理内嵌MLP神经网络模型驱动的电力系统非线性振荡特征分析方法  

Analysis method of nonlinear oscillation characteristics for power system driven by MLP neural network model embedded in Koopman principle

作　　者：周一辰[1] 李金泽 李永刚[1] 陈鹏伟 郭通 孙浩潮 ZHOU Yichen;LI Jinze;LI Yonggang;CHEN Pengwei;GUO Tong;SUN Haochao(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding 071003,China;Department of Electrical Engineering,The Hong Kong Polytechnic University,Hong Kong 999077,China;College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China;Xiong’an New Area Power Supply Company,State Grid Hebei Electric Power Co.,Ltd.,Xiong’an New Area 071700,China;Foshan Power Supply Bureau,China Southern Power Grid Guangdong Power Grid Co.,Ltd.,Foshan 528000,China)

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003 [2]香港理工大学电机工程学系,中国香港999077 [3]南京航空航天大学自动化学院,江苏南京211106 [4]国网河北省电力有限公司雄安新区供电公司,河北雄安新区071700 [5]中国南方电网广东电网有限责任公司佛山供电局,广东佛山528000

出　　处：《电力自动化设备》2024年第10期132-139,共8页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(62303183);香江学者计划项目(XJ2022030);台达电力电子科教发展计划项目(DREG2022002)。

摘　　要：针对电力系统非线性动态特性表征与物理机理融合不清晰、精度低的问题,提出了一种Koopman原理内嵌多层感知机(MLP)神经网络模型驱动的电力系统非线性特性表征与分析方法。阐明了Koopman算子的基本原理,分析了Koopman算子在非线性系统时序演化中的作用。采用MLP神经网络构建编码、解码映射,进而形成Koopman原理内嵌的神经网络深度学习模型,通过深度学习实现非线性系统“编码映射-线性演化-解码映射”3种结构的演化逼近。分析了将所提方法应用于电力系统动态特性分析的物理机理,建立了所提方法的求解与应用流程。通过单机与4机系统算例对所提方法进行对比验证,结果表明所提方法可以精确表征平衡点稳定域内的系统动态过程,可用于电力系统非线性振荡动态特性解析。Aiming at the problems of unclear and low precision in the integration of nonlinear dynamic characterization and physical mechanism for power system,a novel nonlinear characterization and analysis method of power system driven by multi-layer perceptron(MLP)neural network model embedded in Koop⁃man principle is proposed.The basic principle of Koopman operator is clarified and the role of Koopman operator in the time sequence evolution of nonlinear system is analyzed.MLP neural network is used to construct encoding and decoding mapping,and then a neural network deep learning model embedded in Koopman principle is formed.Through deep learning,the evolution approximation of three structures of“encoding mapping-linear evolution-decoding mapping”of nonlinear system is realized.Then,the physical mechanism of applying the proposed method to analyze the dynamic characteristics of power system is analyzed,and the solution and application process of the proposed method is established.The proposed method is compared and verified by single-machine and four-machine system examples.The results show that the proposed method can accurately characterize the dynamic process of the system in the stable region of the equilibrium point and can be used to analyze the dynamic characteristics of nonlinear oscillation in power system.

关 键 词：电力系统 非线性振荡 Koopman算子理论 多层感知机神经网络 科学人工智能 

分 类 号：TM712[电气工程—电力系统及自动化]