Convolution Neural Network for Fault Detection and Classification in Hybrid Overhead-Underground Distribution System

In distribution systems, a trend to use overhead (OH) lines along with underground (UG) cable is becoming popular due to densely populated areas which give rise to right of way and public safety issues. In these complex systems, maintaining continuous power supply and early detection and clearance of the fault are challenging tasks. Feature extraction from fault signals is the main concern of all detection schemes during occurrence of a fault. This article presents a Convolution Neural Network (CNN) which is a single learning body to detect faults in hybrid OH-UG cable. CNN rules are superior to other pattern-recognition techniques since they extract features and simultaneously detect faults as a single entity. The hybrid distribution system is simulated using MATLAB and the proposed scheme is compared with different soft computing techniques.