Identifying electric power system fault types with Deep Neural Network

Abstract

Reliable electric power distribution is crucial in modern society. Modern society is vulnerable to power outages because so many functionalities are using electric power. All the modern communication systems require electric power, also transportation is becoming electrified in near future. It is possible to achieve information from various devices and other sources related to electric power distribution network. One source of information are the protective relays that creates an event recording when the device identify deviation in the network. This event data is not always used effectively and mostly the analysis is done manually and not in real-time. In this study the event recordings are analyzed with deep neural network to find out if it is possible to identify the fault type from the recording automatically. Event recordings are only one source of information from the distribution network. There is a lot of different data available from the network and recently different Internet of Things devices have become available to be used monitoring the network. This other data could be used in combination with event recordings to give more accurate situation analysis of the network.