Topological Attributes of Cascading Failures in Power Grids

Analysing cascading failure events and effectively predicting their impact on blackout-size distribution is critical in designing resilient future power grids. To do so in an efficient manner, cascade-aware predictive models are needed that can capture the complex dynamics of grid topology as failures progress. However, including topological attributes of the failed transmission lines in a direct manner would lead to an exponential growth in the model's complexity. In this paper, a previously reported Markov chain model for cascading failures is extended to incorporate the dynamical topological attributes of the failed transmission lines in a computationally efficient manner; the enlargement of the state space is linear in the number of transmission lines. The method relies on first partitioning the transmission lines using an existing power-flow based hierarchical clustering algorithm to substantially reduce the state-space of the topological attributes, and then extracting two statistical features from the reduced state-space that are used as topological state variables. The analysis shows a strong correlation between the statistical features of the topological attributes and the severity of cascading failures, as measured by the probability of the cascade stopping and the average number of failed lines in the steady state.