Machine learning models for online dynamic security assessment of electric power systems

In this paper we compare two machine learning algorithms (Support Vector Machine and Multi Layer Perceptrons) to perform on-line dynamic security assessment of an electric power system. Dynamic simulation is properly emulated by training SVM and MLP models, with a small amount of information. The experiments show that although both models produce reasonable predictions, the performance indexes of the SVM models are better than those of the MLP models. However the MLP models are of considerably reduced complexity.