Data-driven Contingency Selection for Fast Security Constrained Optimal Power Flow

Security constrained optimal power flow aims to minimize power system operating cost while considering a defined set of outage scenarios. Contingency selection is the process of reducing the number of scenarios to be considered in the problem, thus reducing the computational complexity. In this paper we investigate whether two data-driven methods (artificial neural networks and k-nearest neighbors) can be used to perform extremely fast contingency selection. We demonstrate that both methods are orders of magnitude faster than standard methods. We show that k-nearest-neighbors works well when the algorithm is training on only data specific to the network, while neural networks are better when trained on multiple networks’ data.