中文摘要：

针对线路单端电流行波故障测距因波头标定困难所导致的测距可靠性不高、难自动执行等不足，提出基于历史数据进行案例推理的经验学习式改进思路。首先分析线路故障实录行波的形态特征和影响故障距离折算的因素。针对故障的复发性和多厂家、跨平台、不同电压等级线路故障波形的相似性，对测距过程中的波头标定和距离折算2个关键阶段进行参数化、案例化描述，根据相应特征属性于历史案例库搜索最相近样本进行算法参数复用从而实现智能测距。该方法能够对复发性故障实现测距结论复用，能够对高相似度故障样本实现测距方法复用，测距效果随历史故障样本积累而提升。不同厂家行波装置获得的大量故障实录数据表明该方法可行、有效。

英文摘要：

In order to solve the problem that single-ended traveling wave fault location is hard to implement automatically and reliably, because of the difficulties in wavefront detection, based on historical data, an experiential learning improvement method via multi-stage case-based reasoning was proposed in this paper. The characteristics of actual fault induced traveling wave currents and the factors affecting the converted fault distance were analyzed at first. According to the feature of similarity and repetition of the faults induced traveling waveshapes, the parameterized cases of wavefront detection and converted coefficient were described separately, and then intelligent fault location was performed reusing the most similar known cases key parameters by searching the nearest neighborhood historical cases in each feature space. The proposed method is capable of reusing the final conclusions for repetitive faults, and reusing the solving methods for the cases which are similar with historical cases. The performance of the fault location increases gradually with the accumulation of actual fault samples. The proposed method is proved feasible and effective by actual datasets captured by travelling wave fault location devices of different manufacturers.