A Novel Signal Processing and Deep Learning Approach for Fault Diagnosis in Grid Integrated AC Microgrid

A novel approach for microgrid (MG) protection, utilizing Empirical Mode Decomposition (EMD) and Deep Convolutional Neural Networks (DCNN), is proposed to enhance fault detection for improved reliability and safety. The method applies EMD to decompose the raw signals into distinct intrinsic mode functions (IMFs) across various frequency scales, followed by signal reconstruction based on key IMFs that capture transient characteristics. These reconstructed signals are transformed into 2D scalogram images, which serve as input to the DCNN, allowing the model to extract both temporal and frequency-based features for comprehensive fault analysis. By combining EMD with DCNN, the proposed EMD-DCNN method achieves improved accuracy and reliability in detecting shunt and series faults in grid-connected microgrids, with simulations carried out in the MATLAB/Simulink environment.