Mitigating Blackouts via Smart Relays: A Machine Learning Approach

In this paper, we investigate the protective relays used in electric power systems and their role in large-scale blackouts. After reviewing the state of the art, to mitigate future blackouts, we propose a new machine learning approach for protective relays based on binary hypothesis testing, support vector machines (SVMs), and communications between the protective relays and the supervisory control and data acquisition (SCADA), which we call smart protective relays. The goal of smart relays is to classify and discriminate the normal conditions from fault conditions via local measurements. It is shown that the proposed SVM-based smart relays can detect the location of an initial fault using local current, voltage, real power, and reactive power measurements, and by monitoring these metrics, they can make a correct decision even when the state of the system changes after some equipment failure. We show that by making an intelligent decision on whether and when to trip, and communicating the changes observed to SCADA for fast and intelligent decision making, SVM-based smart relays have the potential to mitigate large-scale blackouts and confine them to much smaller areas. By deploying SVM-based smart relays only at relatively few locations where they have the highest probability to be tripped incorrectly, the probability of cascade of failures and a blackout can be greatly reduced.