A Probabilistic Cascading Failure Model for Dynamic Operating Conditions

Failure propagation in power systems, and the possibility of becoming a cascading event, depend significantly on power system operating conditions. To make informed operating decisions that aim at preventing cascading failures, it is crucial to know the most probable failures based on operating conditions that are close to real-time conditions. In this paper, this need is addressed by developing a cascading failure model that is adaptive to different operating conditions and can quantify the impact of failed grid components on other components. With a three-step approach, the developed model enables predicting potential sequence of failures in a cascading failure, given system operating conditions. First, the interactions between system components under various operating conditions are quantified using the data collected offline, from a simulation-based failure model. Next, given measured line power flows, the most probable interactions corresponding to the system operating conditions are identified. Finally, these interactions are used to predict potential sequence of failures with a propagation tree model. The performance of the developed model under a specific operating condition is evaluated on both IEEE 30-bus and Illinois 200-bus systems, using various evaluation metrics such as Jaccard coefficient, $F_{1}$ score, Precision@ $K$ , and Kendall’s tau.