Design of power distribution network fault data collector for fault detection, location and classification using machine learning

Abstract

The protection and maintenance of a power transmission system during fault condition is indispensable to ensure efficient and reliable power supply to consumers. Most methods of fault detection and location rely on measurements of electrical quantities provided by current and voltage transformers. In this paper, a prototype data collecting device was built for collecting data during different faulted conditions in a single-phase distribution network. Machine learning algorithms were developed for fault detection, location and classification on single-phase distribution lines. The transmission line was modelled using resistor network in the device; the current and voltage sensors were used in the prototype model with the data collection device for current and voltage readings under open-circuit and short-circuit faulted conditions. Training data was collected by varying the load on the line during the simulation of the fault type, sensor location on the node and analyzed. The test data was assessed using three (3) machine learning algorithms namely: K-Nearest Neighbor (KNN), Decision Trees and Support Vector Machines (SVM) for prediction of fault, location and classification within the single-phase distribution network. Test results showed that a higher accuracy rate of 99.42 % was obtained by using the Decision Trees algorithm compared to the others investigated.