Using temporal causal models to isolate failures in power system protection devices

Abstract

Smart Electric Grids and their underlying generation, transmission and distribution systems are constantly exposed to dynamic environments resulting from varying power flows, both direction and magnitude, changing operational requirements and conditions, physical component degradation, and software failures. Maintaining reliability of the power grid even in the presence of faults is one of the top national priorities [1]. Recent blackouts and Hurricane Sandy in 2012 demonstrated the grid vulnerability and reasons to look at existing defense mechanisms more closely. State of the art relies on a network of protection devices that include relays to detect anomalies and circuit breakers to isolate parts of the system that include the faulty components. These local protection schemes operate in short timescales to arrest the fault propagation and protect the remaining system. While the protection devices can mask the fault effects locally, it is important to analyze the events in a global context to improve the decision making. Protection malfunction and its correlation with major blackouts require a careful rethinking of its system-wide effects [2], [3]. This problem is often compounded due to loss of information from relays or Remote Terminal Unit (RTU) failure in the field. Such hidden (unobservable) relay failures are hard to locate and may be responsible for cascades [3].