作者:陈富国1,2, 蔡杰1, 李中旗1
作者单位:1. 平高集团有限公司,河南 平顶山 467001;
2. 西安交通大学电气学院,陕西 西安 712000
关键词:高压隔离开关;故障诊断;经验模态分解;能量矩;长短时记忆网络;在线建模
摘要:
针对高压隔离开关故障诊断准确率低的问题,利用安装在252 kV高压隔离开关操动机构上的传感器采集不同状态下的机械振动信号,研究经验模态分解振动信号方法,计算得到高压隔离开关状态的特征量;并采用相关性及主成分分析相结合的特征量降维方法,提出一种基于长短时记忆网络的高压隔离开关故障在线建模与诊断方法。结果表明:采用相关性与主成分分析相结合的特征量降维方法分析得到的8维综合特征量可以代替25维特征量,实现特征量降维的目的;提出的在线故障诊断模型不仅离线状态实现正常和故障工况的准确分类,而且能够实时在线针对未知故障进行准确诊断,可为高压隔离开关实时在线故障诊断的实施提供技术支撑。
Study on fault diagnosis of high voltage disconnector based on long-short term memory network
CHEN Fuguo1,2, CAI Jie1, LI Zhongqi1
1. Pinggao Group Co., Ltd., Pingdingshan 467001, China;
2. School of Electrical Engineering, Xi’an Jiaotong University, Xi’an 712000, China
Abstract: In view of the low fault diagnosis accuracy of the high voltage disconnector, mechanical vibration signals in different states are collected by sensors mounted on the operating mechanism of a 252 kV high voltage disconnector. The signals decomposed method of empirical mode decomposition is studied, and the characteristic quantity of the high voltage disconnector is calculated. Then, a novel characteristic quantity dimension reduction method based on correlation and principal component analysis is adopted, and an on-line fault diagnosis modeling based on long short-term memory network is proposed. The results show that eight dimensional synthetic characteristic quantities obtained by the method of combining correlation analysis with principal component analysis can replace the twenty-five dimensional characteristic quantities, and finally achieving the purpose of reducing dimension of characteristic quantity. The proposed on-line fault diagnosis model not only realizes the accurate classification of normal and fault conditions off-line state, but also can accurately diagnose unknown faults on-line state. It provides technical support for the implementation of real-time on-line fault diagnosis on high voltage disconnector.
Keywords: high voltage disconnector;fault diagnosis;empirical mode decomposition;energy moment;long-short term memory network;on-line modeling
2022, 48(7):114-119 收稿日期: 2021-05-06;收到修改稿日期: 2021-06-29
基金项目: 国网平高集团科技项目(529110200000)
作者简介: 陈富国(1983-),男,河南周口市人,高级工程师,硕士,从事智能高压开关设备状态监测与评估诊断关键技术研究
参考文献
[1] 李少华, 张文涛, 宋亚凯, 等. 基于高压隔离开关振动信号的故障诊断方法分析[J]. 内蒙古电力技术, 2018, 36(1): 89-92
[2] 宋杲, 顾霓鸿. 高压开关设备运行及故障分析[J]. 供用电, 2007, 24(1): 6-9
[3] 肖荣, 徐澄. 220kV GW6型隔离开关导电回路过热故障分析及处理[J]. 高压电器, 2013, 49(1): 107-110
[4] 申超. 基于光纤Bragg光栅的高压隔离开关触指压力检测技术研究[D]. 昆明: 昆明理工大学, 2012.
[5] 黄聿琛, 阮江军, 邱志斌, 等. 隔离开关操动机构电机输出轴扭矩和电流关系的研究[J]. 高压电器, 2015, 51(6): 139-144
[6] 马宏明, 周涛涛, 彭诗怡, 等. GW4隔离开关机械故障仿真与诊断技术[J]. 电工技术学报, 2019, 34(S2): 501-508
[7] 邱志斌, 阮江军, 黄道春, 等. 基于电机电流检测的高压隔离开关机械故障诊断[J]. 中国电机工程学报, 2015, 35(13): 3459-3466
[8] 曹昆, 李国昌, 王艳松, 等. 基于大数据和随机矩阵理论的变电站状态评估[J]. 中国测试, 2020, 46(8): 149-157
[9] 陈军, 刘鑫, 王利平, 等. Petri网智能变电站保护控制故障诊断与定位方法研究[J]. 中国测试, 2019, 45(10): 128-134
[10] 陈士刚, 关永刚, 张小青, 等. 不完备故障类别下基于Multi-SVDD的高压隔离开关故障诊断方法[J]. 电工技术学报, 2018, 33(11): 2439-2447
[11] 刘仕兵, 宋陵灿, 郭文璟, 等. 基于定子电流特征与SVM高压隔离开关机构故障诊断[J]. 高压电器, 2020, 56(6): 289-295
[12] 王婷. EMD算法研究及其在信号去噪中的应用[D]. 哈尔滨: 哈尔滨工程大学, 2010.
[13] 林圣, 何正友, 罗国敏. 基于小波能量矩的输电线路暂态信号分类识别方法[J]. 电网技术, 2008, 32(20): 30-34
[14] 王艳松, 高鑫, 胡彩娥, 等. 基于核主成分分析和AP聚类算法的电力系统态势感知技术[J/OL]. 电测与仪表: 1-10[2021-07-08]. http://kns.cnki.net/kcms/detail/23.1202.TH.20201125.0902.006.html.
[15] ZHANG S L, WANG Y X, LIU M Q, et al. Data-based line trip fault prediction in power systems using LSTM networks and SVM[J]. IEEE Access, 2018, 6(3): 7675-7686