A transmission line fault classification scheme is proposed in this article using Poincaré-based correlation analysis of three-phase fault currents. The method segments each fault signal into two equal time-split components and computes correlation coefficient between these two time-split signals. The fault current signals of the directly affected line(s) observe an abrupt monotonic rise, compared to the indirectly affected phases. This sudden rise in magnitude is expressed with correlation coefficients between the two almost consecutive time-split components of signal, time shifted by delay index. This method emphasizes this monotonic nature of increment of the fault current, enabling prompt fault detection. Further analysis of three phases of fault signals independently yields a set of correlation coefficients for ten different fault prototypes, which are used to develop fault classifier signatures for direct classification. The proposed method yields high classification accuracy of 99.76% using only (1/6)th of the post-fault noisy signal with fault resistance varying from 0.01 to 100Ω. Besides, analysis of only one end single discards the requirement of time synchronous signal acquisition from both ends. Finally, use of simple analysis reduces the computational burden compared to several contemporary methods.

本文使用基于庞加莱的三相故障电流相关分析提出了一种输电线路故障分类方案。该方法将每个故障信号分割成两个相等的时间分割分量，并计算这两个时间分割信号之间的相关系数。与间接受影响的相相比，直接受影响线路的故障电流信号观察到突然单调上升。这种幅度的突然上升用信号的两个几乎连续的时间分割分量之间的相关系数表示，时间偏移了延迟指数。这种方法强调了故障电流增量的这种单调性质，从而能够进行及时的故障检测。对三相故障信号的进一步分析独立产生了一组十种不同故障原型的相关系数，用于开发用于直接分类的故障分类器签名。所提出的方法仅使用故障电阻从 0.01 到 100Ω 变化的故障后噪声信号的 (1/6)th，就产生了 99.76% 的高分类精度。此外，仅分析一端单，摒弃了从两端获取时间同步信号的要求。最后，与几种当代方法相比，简单分析的使用减少了计算负担。