Intelligent Fault Classification and Localization in Islanded DC Microgrids: A Discrete Wavelet Transform and Neural Network Approach

In this paper, a robust protection scheme is presented for fault detection, classification, and location identification in an islanded DC microgrid comprising PV arrays, wind turbines, battery energy storage system, and loads. The currents of individual Distributed Energy Resources (DERs) are monitored, and Discrete Wavelet Transform is applied with a non-overlapping rolling time window for feature extraction, yielding approximate and detailed coefficients. Utilizing the relative wavelet energy of these coefficients and an Artificial Neural Network (ANN) pattern recognition tool, the system accurately classifies faults-pole to pole, pole to ground, or load switching. MATLAB Simulink is employed for simulations across diverse fault cases, validating the effectiveness of the proposed scheme. The ANN, trained for three fault types and thirteen fault locations, demonstrates high accuracy, through a comprehensive analysis, including confusion matrix-based metrics.