Leveraging data mining for critical branch identification through simultaneity and causality correlation analysis under cascading failures in power systems

Abstract

Identifying critical branches or propagation paths from cascading failure data can be an effective way to mitigate and even prevent cascading blackouts in power systems. Hence, this paper proposes a data mining-based identification framework to find critical correlations among propagating pathways. We define simultaneity correlation and causality correlation to comprehensively reveal the fault propagation features according to the temporal and synchronous dependence of critical branches during fault propagation. The itemset and sequence mining pattern is used to model the two types of correlations and then mine the critical correlations, respectively. To reduce the impacts of the incompleteness of initial conditions and improve the accuracy of mining, the Shannon diversity index is introduced to quantify the diversity of initial conditions. Moreover, we propose a recursion graph-based probability calculation model to fast predict the probability/risk of occurrence of the unknown correlations in cascading failures. Numerical simulation results based on the IEEE 118-bus system verify the effectiveness of the proposed method.