Mobile Edge Computing Sensors and Cloud Machine Learning Advance Grid Predictive Maintenance

The resiliency of the nation's electric grid is under continuous stress. Impacts of weather, wear and tear, and the age of the grid all contribute to a decline in reliability and an increase in power outages. Aged and deteriorated overhead electrical equipment produce characteristic radio frequency (RF) emissions. Fixed sensors to monitor the condition of all grid equipment would be expensive and create an enormous data-handling problem. Mobile Edge Computing sensor (MECS) technology that can discriminate and locate the equipment that produce pre-failure RF emission failure signatures would enable equipment replacement before catastrophic failure causes an outage. Rebuilding or replacing the 5 million miles of the U.S. transmission and distribution grid is not an economical alternative. Maintaining the grid through a strategy of predictive, conditions-based maintenance results in improved grid performance, fewer outages, and addresses a worldwide problem. This paper and presentation detail a pilot project where MECS are installed on sanitation trucks. The vehicle mounted MECS evaluate changes in the state of electrical equipment on an entire city grid each week. Machine learning algorithms correlate a variety of grid impacts such as weather, load, season, and protective equipment operations to develop grid resilience forecast models. The forecasts allow financial impact evaluation of grid rebuilds, undergrounding programs, and maintenance strategies. The technology, data examples, and the results of the pilot project in a 65 square-mile Midwest city will be presented.