基于自适应ANN的孤岛微网群电压频率分布式协同控制  

Distributed Voltage and Frequency Cooperative Control for Isolated Microgrid Cluster Based on Adaptive Artificial Neural Network

作　　者：王海云 唐涛南 王卫 张雨璇 汤云帅 孙方霞[4] WANG Haiyun;TANG Taonan;WANG Wei;ZHANG Yuxuan;TANG Yunshuai;SUN Fangxia(Electric Power Research Institute,State Grid Beijing Electric Power Company,Beijing 100075,China;State Grid Beijing Power Dispatch and Control Center,Beijing 100031,China;Beijing Qingda Hi-tech System ControlCo.,Ltd,Beijing 102208,China;College of Marine Engineering,Electrization and Intelligence,JiangsuMaritime Institute,Nanjing 211170,China)

机构地区：[1]国网北京市电力公司电力科学研究院,北京100075 [2]国网北京电力调度控制中心,北京100031 [3]北京清大高科系统控制有限公司,北京102208 [4]江苏海事职业技术学院轮机电气与智能工程学院,南京211170

出　　处：《电源学报》2023年第4期105-114,共10页Journal of Power Supply

基　　金：国网北京市电力公司科技资助项目(52022320006H)。

摘　　要：针对孤岛微网群稳定运行难度大,电压和频率控制极为复杂的问题,提出了一种适用于含多逆变器分布式电源DGs(distributed generations)微网群MGC(microgrid cluster)的智能电压、频率协同控制方法。首先,该方法利用李雅普诺夫理论和基于逆变器DG的动态特性设计了基于模型化的控制器;然后,利用人工神经网络ANN(artificial neural network)来近似上述动态特性,从而得到不需要DG参数先验信息的智能控制器,此外,所提控制器不需要使用电压和电流PI控制器;最后,通过不同场景下的仿真分析,验证了所提控制器的有效性,还利用李雅普诺夫分析,证明了跟踪误差和神经网络权值最终一致有界,从而可实现较好的电压和频率动态调节。To solve the problems of difficulty in the stable operation of isolated microgrid cluster(MGC) and complexity of voltage and frequency control,an intelligent voltage and frequency cooperative control method which is suitable for an MGC with multiple inverter distributed generations(DGs) is proposed.First,a model-based controller is designed using the Lyapunov theory and the dynamic characteristics of inverter DGs.Then,artificial neural network(ANN) is used to approximate the dynamic characteristics,so as to obtain an intelligent controller which does not require the prior information about DG parameters.In addition,the proposed controller does not need to use a voltage and current PI controller.Finally,the effectiveness of the proposed controller is verified through simulation analysis in different scenarios.Moreover,Lyapunov analysis is also used to prove that the tracking error and neural network weights are ultimately uniformly bounded,thereby achieving a better dynamic regulation of voltage and frequency.

关 键 词：人工神经网络 协同控制 分布式电源 李雅普诺夫理论 

分 类 号：TM28[一般工业技术—材料科学与工程]