★ 国网南通供电公司 曹艺洋

摘要： 当前配电网故障诊断体系存在诊断延时过长的问题，传统依赖GIS平台的方法在大规模复杂数据场景下难以兼顾实时性与准确性。针对这一问题，本文提出了一种基于数据挖掘的混合诊断方法。该方法利用mPMU数据构建故障样本集，通过KICIC聚类实现故障类型划分，并结合粗糙集属性约简与改进LM-BP神经网络进行快速识别。实验结果表明，与传统GIS方法相比，该方法在保证诊断准确率的同时，将平均响应时延从约160ms缩短至95ms，最大延时压缩至120ms以内，整体耗时降低约40%。该研究有效验证了数据挖掘技术在提升配电网故障诊断实时性方面的优势，为电网快速隔离故障与恢复供电提供了可靠支撑。

关键词：数据挖掘；配电网故障；故障诊断；诊断分析

Abstract: Current fault diagnosis systems for distribution networks suffer from excessive diagnostic latency. Traditional methods relying on GIS platforms struggle to balance real-time performance and accuracy when handling large-scale, complex data scenarios. To address this challenge, this paper proposes a hybrid fault diagnosis method based on data mining techniques. The method constructs fault sample sets using μPMU data, classifies fault types via KICIC clustering, and integrates rough set attribute reduction with an improved LM- BP neural network for rapid identification. Experimental results demonstrate that, compared to conventional GIS-based approaches, the proposed method maintains diagnostic accuracy while reducing the average response delay from approximately 160 ms to 95 ms, with maximum latency compressed to within 120 ms—an overall reduction of approximately 40%. This study effectively validates the advantages of data mining techniques in enhancing the real-time performance of distribution network fault diagnosis, providing reliable support for rapid fault isolation and service restoration in power grids.

Key words: Data mining; Distribution network fault; Fault diagnosis; Diagnosis analysis

在线预览：数据挖掘技术在配电网故障诊断中的应用.pdf

摘自《自动化博览》2026年7月刊