A data mining transmission switching heuristic for post-contingency AC power flow violation reduction in real-world, large-scale systems

Abstract

Transmission switching has proven to be a highly useful post-contingency recovery technique by allowing power system operators increased levels of control through leveraging the topology of the power system. However, transmission switching remains only implemented in limited capacity because of concerns over computational complexity, uncertainty of performance in AC systems, and scalability to real-world, large-scale systems. We propose a heuristic which uses a sophisticated guided undersampling procedure combined with logistic regression to accurately identify transmission switching actions to reduce post-contingency AC power flow violations. The proposed heuristic was tested on real-world, large-scale AC power system data and consistently identified optimal or near optimal transmission switching actions. Because the proposed heuristic is computationally inexpensive, addresses an AC system, and is validated on real-world large-scale data, it directly addresses the aforementioned issues regarding transmission switching implementation.