基于DHNN指纹库的电网故障诊断方法  

Power Grid Fault Diagnosis Method Based on DHNN Fingerprint Database

作　　者：顾发英 罗旋 刘育 胡孔飞 王皓冉 GU Fa-ying;LUO Xuan;LIU Yu;HU Kong-fei;WANG Hao-ran(Dadu River Hydropower Development Co.Ltd.,Chengdu 610041.China;Dadu River Production Command Center.Chengdu 610041,China;Chengdu Dahui Zhilian T echnology Co.,Ltd.,Chengdu 610041,China;Tellhow Scitech Co.,Ltd.Nanchang 330029,China;Sichuan Energy Internet Research Institute,Tsinghua University,Chengdu 610000,China)

机构地区：[1]国电大渡河流域水电开发有限公司,四川成都610041 [2]国电大渡河流域生产指挥中心,四川成都610041 [3]成都大汇智联科技有限公司,四川成都610041 [4]泰豪科技股份有限公司,江西南昌330029 [5]清华四川能源互联网研究院,四川成都610000

出　　处：《水电能源科学》2021年第2期195-199,共5页Water Resources and Power

基　　金：国家自然科学基金项目(51709191);国家重点研发计划(2019YFB1310504);四川省重大科技专项(2018GZDX0043)。

摘　　要：针对高维度、非线性的电力系统存在难以实现快速、精准和有效的故障诊断问题,提出了基于DHNN指纹库的电网故障诊断方法。首先利用离散Hopfield神经网络(DHNN)算法对电网遥信数据分类和建立各种基准信号权值,构建了基准信号集合的信号模板;其次基于该模板建立各类故障指纹值编码及其编码组合为一体的指纹库;最后以瀑布沟水电站布坡三线“B相瞬时故障,重合成功”故障为例验证该诊断方法的正确性。结果表明,基于DHNN指纹库的电网故障诊断方法与实际电网故障描述完全一致,可正确有效地提升电网故障诊断水平,同时还可将该方法推广应用于国内大型流域梯级电站群组合成的大规模高压电网的故障诊断中。Aiming at the problem of high-dimensional,non-linear power system that is difficult to achieve fast,accurate and effective fault diagnosis,a power grid fault diagnosis method is proposed based on DHNN fingerprint database.First,the discrete Hopfield neural network(DHNN) algorithm was used to classify the power grid remote signal data and establish various reference signal weights,and a signal template for the reference signal set was constructed.Secondly,a fingerprint database of all kinds of fault fingerprint value coding and its coding combination included the signal template was established.Finally,taking the "B-phase instantaneous fault,reclosing success" fault of Bupo 3 rd line of Pubugou Hydropower Station as an example,the correctness of the diagnosis method was verified.The results show that the fault diagnosis method based on the DHNN fingerprint database is completely consistent with the actual description of the operation of the power grid,which concludes that the fault diagnosis method can accurately and effectively improve the fault diagnosis level of the power grid.Therefore,the proposed method can be extended to the cluster control center of large-scale cascade power stations in China.

关 键 词：电网 故障诊断 离散HOPFIELD神经网络 指纹库 指纹值编码 

分 类 号：TM764[电气工程—电力系统及自动化]