Fault Detection and Classification in DC Microgrid Based on Convolutional Neural Network

A complex protection challenge has hampered the widespread acceptance of DC microgrids despite their numerous advantages, including increased efficiency and stability, as well as the ease with which they can be integrated with renewable energy sources (RES). With the lack of natural zero crossing, intermittent nature of RES, as well as the high likelihood of distributed energy resource (DER) outages, make the task of fault detection and classification even more difficult. A DC microgrid protection strategy is proposed in this paper that is more resilient to weather uncertainty and probable generation outages. In order detect and classify the fault, a convolution neural network (CNN) based technique is used in this paper. The proposed scheme minimizes the complexity and cost associated with the feature extraction process. The CNN classifier correctly differentiates between all events, and its performance is demonstrated by a variety of metrics. MATLAB/Simulink platform is used to model the DC microgrid system.