Intelligent Fault Diagnosis Using Deep Learning for a Microgrid with High Penetration of Renewable Energy Sources

—Fault diagnosis is essential to attain the microgrid's stable, reliable, and economical operation. However, a microgrid's fault current varies drastically based on the connected operating mode and connected renewable energy sources. This paper develops an intelligent fault detection and identification scheme for microgrids with high renewable energy penetration by combining Fourier-based Continuous Wavelet Transform (FTCWT) and Deep Learning (DL) methods. Several time-domain case studies representing different microgrid operating conditions were obtained and pre-processed in MATLAB/SIMULINK. The time-series data at the relay location are processed using an FTCWT to get time-frequency representation, and the sequence components waveforms were computed. The time-frequency representation and sequence components waveform contain helpful information, including fault conditions and healthy operating conditions for grid-connected and islanded microgrids that act as input datasets to the DL models. The proposed DL models perform fault location identification, fault type identification, and fault phase-detection for different operating modes. The development of the DL model for the fault detection/classification unit and fault location identification unit is carried out in the PyTorch framework. The experimental results show that the proposed model is more accurate when compared to the state-of-the-art methods.