Fault Location Scheme for Cross-Country Faults in Dual-Circuit Line Using Optimized Regression Tree

Abstract

In the double-circuit transmission line (DCTL), location of cross-country faults (CCFs) is more wearisome due to its intricate nature. The CCFs can occur at miscellaneous locations on dissimilar phases at the same fault inception time. Furthermore, the CCFs encompasses different fault locations which can mislead the line patrolling team and not only takes long hours to attend the fault location. Therefore, this may also cause electrical stress on the various power system components owing to tripping of circuit breaker repeatedly because of inappropriate fault clearance. In this context, an ensemble of regression tree (ERT) model-based fault location scheme is proposed using different regression trees such as Bagged Regression Trees (BGRT) and Boosted Regression Trees (BSRT). These regression tree modules have been trained with optimized hyper-parameters such as minimum leaf size, leaning cycles, and learning rate by using Bayesian optimization. A 400 kV, 50 Hz Chhattisgarh state power system (CSPS) network has been designed and simulated in MATLAB/Simulink to implement the proposed fault location scheme. Exclusive datasets have been designed at atypical fault scenarios thereby applying an exploratory signal processing technique such as Discrete Wavelet Transform (DWT). Additionally, the performance assessment has been done by comparing different error metrics. The simulation results reveal the applicability of the proposed ensemble regression tree (ERT) model for location of CCFs and it gives a research insight to adapt the same in the real power system network.