Abstract
An unexpected failure or outage of one or multiple system components can cause a new operational situation that requires remedial actions. An important remedial action to model correctly is islanding. Finding the transient stability of an island is computationally heavy, and it may be necessary with a trade-off between speed and accuracy in the classification of island stability. This is especially the case if one has to perform a large number of simulations.In this paper, a decision tree based ensemble method is used to predict the stability of islands in the power system during a contingency event. A comparison study shows that the trained model can contribute with a large reduction in time spent on the transient stability assessment, while being substantially more accurate than a static power flow simulation.