An AI-Based Technique for Fault Location in Inverter-Based Active Distribution Networks

Fault location is increasingly essential in inverter-based active distribution networks. This is due to the large number of branches and laterals in such networks, as well as the presence of inverter-based distributed generators (IBDGs). Several techniques are used for locating faults in distribution networks, including impedance-based approaches, traveling wave-based schemes, and artificial intelligence (AI)-based approaches. AI-based schemes are superior to others in terms of speed and accuracy, and they do not demand high-frequency devices. However, there is a lack of AI-based schemes that can effectively address scenarios involving a high number of branches, a limited number of measurement instruments, the presence of IBDGs, and high fault resistance. Accordingly, this paper introduces a modified one-dimensional convolutional neural network (1-D CNN) that combines residual connections with 1-D CNNs. The suggested approach includes two elements for fault location: (i) determining the fault distance and (ii) identifying the section of the network that is faulty. The results indicate that this approach effectively pinpoints faults with varying resistance levels at different locations, even in the presence of IBDGs. Ultimately, the proposed solution demonstrates enhanced accuracy in networks featuring multiple distributed generators, numerous sub-branches, unbalanced load conditions, and diverse fault scenarios.