Convolutional Neural Network Based Fault Location Detector for Power Grids

Monitoring of power distribution networks for the purpose of identifying system faults has become more important due to the continuous demands for reliable and sustainable power supply. Several works are found in literature which focus on optimal placement of sensor units to achieve full monitoring and observability of power grids. In our previous work [1], a data driven approach has been introduced to determine the location of measurement units to achieve a desired fault location detection accuracy. In [1], the location of phasor measurement units (PMUs) has been identified via several feature selection based measures, and the performance of several machine learning based detectors has been evaluated. Simulation results have shown that the mutual information (MI) measure is capable of placing minimum number of PMUs for sufficient grid observability to achieve a desired detection accuracy using a decision tree (DT) based detector. In this work however, a new approach in the form of a convolutional neural network (CNN) based detector is introduced, and its performance is compared to the detectors in [1]. Several benchmark distribution systems are used in this work in order to evaluate the performance of the fault location detectors.