Leveraging Streaming Analytics for Anomaly Detection in the Power Distribution Grid
The electric power distribution grid is a complex collection of devices and systems. The Utilities industry recognizes the value of Monitoring and Advanced Analytics to preserve the reliability of power distribution grid. But extracting historical data from different systems and normalizing the data so that one can then also train AI and ML models is often an expensive and a difficult process. In this session we will discuss a patent-pending technique that uses real-time data streams to learn and identify anomalies in the power distribution grid. It is designed on a learning system that ingests streaming data to detect anomalous behavior in real time. The AI and ML models are easy to deploy and designed to scale across a large fleet of like assets. This new technique will alloy utility companies to mitigate the effects of power outages caused by transformer failures and save billions of dollars (to quantify the magnitude of this industry problem, annual cost of short power interruptions (i.e. five minutes or less) is US$60 billion per year.) The standard smart meter measures how much electric power and electric energy a home or facility uses, but it also measures other electric parameters which characterize the quality of your power (such as voltage). Anomalies in voltage and other electric parameters are leading indicators of equipment failures. Early anomaly detection by our ML-enhanced Subspace Tracking (SST) technique can help alert on a problem before a failure occurs. This can help avoid unplanned interruptions, employee exposure to safety hazards, and high restoration costs, among others. SAS technique does not require installation of new sensors and meters, we deploy our ML-enhanced SST method to profile and monitor already-installed smart meters. Join us at the session to learn more about the technique to see how one can deploy advanced anomaly detection on power distribution grid in a much faster, simpler, and less historical data intrusive method.
Priyadarshini (Priya) Sharma is a Senior Solutions Advisor at SAS Institute’s Internet of Things Division with specialization in machine learning, artificial intelligence, model management and operationalization of analytical lifecycle. Over the years she has worked on financial analysis, financial crimes and risk management, in-memory analytics, and in-database technologies.
James Caton is Principal BDE at SAS Institute’s Internet of Things Division. He is responsible for AI and IOT Partnerships in Smart Infrastructure, Cities for SAS. He has led Analytics efforts for IBM globally (Industrial IOT and Smart Cities) and Larsen Toubro, a $16B Indian conglomerate (India Smart Cities).