Enhanced Power System Resilience via Machine Learning for Critical Line Identification and Cascading Failure Prediction

The increasing frequency of extreme events and the high penetration of renewable energy sources have made power system resilience a critical challenge. This paper presents a hybrid framework for identifying critical lines and predicting cascading failure risk in power systems. First, using N-۱ contingency analysis on the IEEE ۳۹-bus test system, five critical lines are identified. Then, ۳۰ fault scenarios, including five critical and five non-critical lines under three different load levels, are simulated. Results show that ۱۰۰% of critical lines lead to network collapse, while this number is ۵۳.۳% for non-critical lines. A novel composite resilience index is also proposed that quantifies the effect of load level on network resilience. Finally, a Random Forest model, with ۱۰% noise applied to the key feature, achieves ۹۰.۰۰% accuracy in predicting high-risk lines. The findings demonstrate that combining cascading failure analysis with machine learning provides an effective tool for enhancing power system resilience.

کلیدواژه ها:

Power system resilience ، Cascading failure ، N-۱ contingency analysis ، Machine learning ، Composite resilience index