A Comparative Study of Data-Driven Power Grid Cascading Failure Prediction Methods

Cascading failures in power grids, where failures propagate from one component to another, are a major cause of large-scale blackouts. With renewed interest in enhancing power grid resilience, it is even more critical to predict cascading failures so that effective mitigative actions can be identified. The existing cascading failure prediction methods lack high accuracy and fast computation time and often face challenges due to unbalanced or unrepresentative datasets. In this work, a comparative study of various data-driven methods for failure prediction with a shorter computation time is provided. The problem is formulated as a binary classification, where the input features, such as the loading levels of the transmission lines, are mapped to the output, which is the failure status of the transmission lines. To validate the effectiveness of the proposed methods, the IEEE 30- bus system is used as a test case, and the results confirm the viability of the compared methods for failure prediction. This study could guide future research in developing fast and accurate data-driven cascading failure models.