Power system probabilistic security assessment using Bayes classifier

Abstract

This paper presents a method for power system security assessment based on the Bayes classifier. This method can be applied to calculate probabilistic security indices as well as on-line security assessment. In general, the determination of security breach is a cumbersome and time-consuming process due to dynamic and steady-state effects. These effects can be incorporated by considering transient stability, satisfaction of system load without violation of constraints, and voltage stability studies. The variation of the system load, as well as contingencies, may cause system transition to a different operating state. It is impractical if not impossible to evaluate all these situations, such as contingencies resulting from load variation. The straightforward Monte Carlo simulation, one of the possible methods in power system reliability analysis, requires the evaluation of a system operating state for each sampled state and is computation-intensive. In the proposed method, first the joint probability density of feature vectors has been obtained by using some training data. Once this joint distribution is obtained, the Bayes classifier provides the assessment of system security without complicated contingency analyses and can reduce the computational burden. Security status of a given feature vector can be determined by a posteriori probability rule called Bayes rule, which can be implemented in on-line security assessment or power system reliability studies.