基于RA-CNN和同步相量的风电场次/超同步振荡参数智能辨识方法  

Intelligent Identification Method of Wind Farm Sub-synchronous/Supersynchronous Oscillation Parameters Based on RA-CNN and Synchrophasor

作　　者：陆友文 崔昊 陈佳宁 彭祥佳 冯双[2] 刘栋[3] LU Youwen;CUI Hao;CHEN Jianing;PENG Xiangjia;FENG Shuang;LIU Dong(College of Software Engineering,Southeast University,Nanjing 210096,China;School of Electrical Engineering,Southeast University,Nanjing 210096,China;NARI Technology Co.,Ltd.,Nanjing 211106,China)

机构地区：[1]东南大学软件学院,江苏南京210096 [2]东南大学电气工程学院,江苏南京210096 [3]国电南瑞科技股份有限公司,江苏南京211106

出　　处：《中国电力》2023年第4期46-55,67,共11页Electric Power

基　　金：国家自然科学基金资助项目(含高渗透并网变流器电力系统宽频强迫振荡机理及监测方法研究,51807025)。

摘　　要：近年来风电并网比例大幅提高,由此引发的次/超同步振荡的发生概率也大大提高,严重威胁系统的安全稳定性。准确辨识次/超同步振荡参数是抑制振荡的基础,提出基于注意力机制的残差卷积神经网络的辨识方法。卷积神经网络的局部相关性和权值共享决定了其具有更强的特征学习和表达能力,通过结合注意力机制可以更准确地辨识振荡参数。同时,引入残差连接,用以解决深层卷积神经网络存在的梯度消失和网络退化问题。仿真结果表明:相较于传统方法,该方法不仅能在较短时间窗数据上完整地辨识次/超同步振荡的参数,且能规避传统方法因主观因素带来的辨识误差,降低振荡参数辨识的复杂度。In recent years,the proportion of wind power connected to the grid has increased significantly,and the probability of occurrence of sub-synchronous/super-synchronous oscillations caused by this has also been greatly raised,which seriously threatens the safety and steadiness of the system.Accurate identification of sub-synchronous/super-synchronous oscillation parameters is the basis for oscillation suppression.Therefore,this paper proposes an identification method based on the attention mechanism of the residual convolutional neural network(CNN).The local correlation and weight sharing of the convolutional neural network determine its stronger feature learning and expression ability,and thus,the oscillation parameters can be more accurately identified when it is combined with the attention mechanism.Meanwhile,this method introduces residual connections to solve the problems of gradient vanishing and network degradation in the deep convolutional neural network.Simulations indicate that compared with the traditional method,this method can completely identify the parameters of sub-synchronous/super-synchronous oscillations on the data of a short time window.In addition,it can avoid the identification error caused by the subjective factors of the traditional method and reduce the complexity of oscillation parameter identification.

关 键 词：次/超同步振荡 同步相量 参数辨识 卷积神经网络 注意力机制 残差网络 

分 类 号：TM712[电气工程—电力系统及自动化] TM614