Multifunctional Algorithm for Fault Analysis in Radial Distribution Network with Distributed Generation

Abstract

Distribution networks along with penetration of distributed generations shows complex operating characteristics that poses challenges to conventional fault analysis methods. To address this problem, this study proposes a multifunctional fault analysis framework that uses advanced signal processing with data-driven learning techniques. The approach starts with the use of the Hilbert transform to voltage and current signals that helps to extract disturbance associated sensitive transient features which+ are used as inputs to a two-stage classification structure. In the first stage, a Random Forest classifier is used for fast and reliable islanding detection and fault detection. In the second stage, an AdaBoost classifier performs the function of classifying fault type and locating faulted bus along with distance from it. Standard IEEE 33-bus distribution system with integrated distributed generators, modeled in MATLAB/Simulink is used to validate the proposed model, while for training and testing the classifiers Python is used. For a range of fault and non-fault scenarios performance is evaluated under grid-connected radial operation. Results confirm that the combined Hilbert–Random Forest–AdaBoost framework achieves higher accuracy in terms of fault detection and classification than conventional techniques such as decision trees and support vector machines. Also, K-fold cross-validation yields a mean accuracy above 90%, and performance metrics such as accuracy, precision, recall, and F1-score all exceed 90%, that confirms the robustness and practicability of the proposed scheme for fault analysis of distribution networks with distributed generation.