A supervised principal component analysis-based approach of fault localization in transmission lines for single line to ground faults

Abstract

Fault localization in power transmission line has been a key area of research, especially with the advent of modern soft computation-based digital relaying schemes. In this article, a principal component analysis (PCA)-based method of fault detection and localization in an overhead transmission line is proposed. The analysis is carried out for single line to ground faults only as these faults encompass about 70–90% among all the faults in transmission lines; although this generalized method could also be extended to cater other fault prototypes too. A fault in a line causes disturbances in the form of high-frequency transient oscillations which are developed immediately after the fault. The peaks and crests of these fault transients are investigated as key features from the receiving end three-phase voltage signals. These feature points are fed to a PCA-based fault localizer algorithm as supervised inputs. The proposed model is tested with a large set of fault signals, varying all the three major influencing parameters like fault locations, fault resistances as well as fault inception angle. Power system noise is also incorporated additionally in the fault signals order to incorporate more robustness to the localizer model. An average localization percentage error of 0.871% ranging from 0.011 to 2.413% is obtained from this work. The proposed work requires only a maximum of nearly 10.1% of post-cycle fault signal for analysis, which is remarkably less compared to several other schemes.