Decision tree analysis to identify harmful contingencies and estimate blackout indices for predicting system vulnerability

Abstract Cascading failure is the main mechanism for progressing large blackouts in power systems. Following an initial event, it is challenging to predict whether there is a potential for starting cascading failure. In fact, the potential of an event for starting a cascading failure depends on many factors such as network structure, system operating point and nature of the event. In this paper, based on the application of decision tree, a new approach is proposed for identifying harmful line outages with the potential of starting and propagating cascading failures. For this purpose, associated with each trajectory of the cascading failure, a blackout index is defined that determines the potential of the initial event for triggering cascading failures towards power system blackout. In order to estimate the blackout indices associated with a line outage, a three stages harmful estimator decision tree (HEDT) is proposed. The proposed HEDT works based on the online operating data provided by a wide area monitoring system (WAMS). The New England 39-bus test system is utilized to show the worthiness of the proposed algorithm.