首页|期刊导航|电气技术|换流变压器接地电流监测及诊断装置的设计与研究

换流变压器接地电流监测及诊断装置的设计与研究OA

Monitoring and fault diagnosis of grounding current in converter transformers

中文摘要英文摘要

针对换流变压器接地电流谐波畸变率高、故障识别困难的问题,本文开展接地电流监测与故障诊断研究.建立正常工况和故障工况下铁心及夹件接地电流分析模型,设计接地电流在线监测装置的硬件与软件系统.为提高检测精度,采用"中值滤波+滑动平均滤波"复合滤波方法,并结合窗函数实现频域分析.针对故障特征冗余问题,提出基于粗糙集属性约简和 K 折交叉验证优化概率神经网络的故障诊断方法.结果表明,所设计装置测量误差小于 5%,诊断模型能够实现换流变压器接地故障的有效识别,现场应用验证了该方法的可行性.

To address the high harmonic distortion and the difficulty in fault identification of grounding current in converter transformers,analytical models of the core and clamp grounding currents under normal and fault conditions are established,and an online grounding current monitoring device is developed.A combined median-filtering and moving-average method,together with windowed processing,is adopted for frequency-domain analysis.In addition,a fault diagnosis method based on rough set attribute reduction and K-fold cross-validation optimized probabilistic neural network is proposed.Results show that the measurement error of the developed device is less than 5%,and the proposed model can effectively identify grounding faults in converter transformers.Field application results further verify the feasibility of the proposed method.

田天;周秀;马宇坤;牛勃;刘英男

国网宁夏电力有限公司电力科学研究院,银川 750011国网宁夏电力有限公司电力科学研究院,银川 750011国网宁夏电力有限公司电力科学研究院,银川 750011国网宁夏电力有限公司电力科学研究院,银川 750011山东诚智电力科技有限公司,济南 250000

换流变压器接地电流在线监测故障诊断概率神经网络

converter transformergrounding currentonline monitoringfault diagnosispro-babilistic neural network

《电气技术》 2026 (10)

55-63,9

国网宁夏电力有限公司科技项目(5229DK25000K)

评论