In a typical overhead distribution system, a significant number of outages (specifically successful auto-reclosures with outage duration < 1 minute) will usually have its cause classified as ‘unknown’. With this in mind, Toronto Hydro has explored new technology advancements, such as a Fault Identification tool.
This tool leverages data already collected by the utility (e.g. power quality data) and applies machine learning algorithms that ‘learn’ from and make decisions on outage data with ‘known’ causes to determine the likely cause of ‘unknown’ outage events.
The tool’s accuracy is currently being assessed by withholding outage events with ‘known’ causes from training the machine learning model to see if it can correctly classify them. Based on 2018 data, it is showing promising results.
The Fault Identification tool is one of the first applications of Machine Learning in the Electric Utility space.
This tool will ultimately assist in investigating and diagnosing issues on the grid such as intermittent feeder outages.
