Time-delayed Interaction Model for Causal Inference between Cascading Failures in Power Grid

The interaction model is widely used to study the propagation patterns of cascading failures in power systems. However, most existing models rely on the Markovian assumption, which limits interactions to adjacent generations of outages, thereby ignoring the potential influence of earlier outages. To address this limitation, this paper proposes an improved interaction model that captures time-delayed interactions between cascading line outages. A time-delayed parameter is introduced to account for multi-timescale interactions, and the Expectation-Maximization (EM) algorithm is employed to estimate the interaction tensor. Case studies conducted on the IEEE-RTS96 bus system demonstrate the effectiveness of the proposed method in identifying both immediate and delayed causal relationships. The results reveal a significant number of previously unidentified interaction links, highlighting the importance of considering time-delayed effects in cascading failure analysis. This model provides a more comprehensive understanding of failure propagation and offers a valuable tool for enhancing power grid resilience.