Advancing Fault Detection in Distribution Networks with a Real-Time Approach Using Robust RVFLN

English (OpenAlex)

In this paper, the fault type and location of high-impedance short-circuit faults, which are difficult to detect in distribution networks, are determined in real time using the Real-Time Digital Simulator (RTDS). In this study, an IEEE 39-bar system model is created using the Real-Time Simulation Software Package (RSCAD). In this model, a short-circuit fault is generated at different fault impedance values. For high-impedance short-circuit fault detection, 14 feature vectors are created. Six of these feature vectors are newly developed, and it is found that these six new feature vectors contribute 10% to the detection of hard-to-detect high-impedance short-circuit faults. We propose a data-driven online algorithm for fault type and location detection based on robust regularized random vector function networks (ORR-RVFLNs). Moreover, the robustness of the model is improved by adding a certain amount of noise to the detected short-circuit fault data. In this study, the method ORR-RVFLN for the 39-bus system IEEE detects the average error type for all error impedances, with 92.2% success for the data with noise added. In this study, the fault location is shown to be more than 90% accurate for distances greater than 400 m.