Performance analysis of machine learning algorithms for power system dynamic security assessment

Abstract

This paper presents performance analysis of decision tree (DT) algorithms for real-time dynamic security assessment (DSA) application for power system. The scope of this study is limited to three widely implemented DT algorithms in practice namely: random forest, reduced error pruning (REP), and J48. The performance of three decision tree (DT) algorithms are evaluated in assessing the dynamic security assessment (DSA) of the power system network. The algorithms are evaluated in terms of its accuracy, error, and computational time. The performance of these algorithms are demonstrated on a modified IEEE 30-bus test system model using PowerWorld software. From this study, the results show that random forest algorithm outperform other min method in terms of accuracy and error. On the other hand, REP outpace other algorithms in terms of computational speed in assessing the dynamic security of the system.