SVM based intelligent predictor for identifying critical lines with potential for cascading failures using pre-outage operating data

Abstract

The process of cascading failures is one of the most important threats to the security of the power system, which can lead to a number of blackouts. Therefore, predicting the potential of cascading failure occurrence (PCFO) using pre-outages information is very essential. In this study, PCFO, associated to each line outage which is an important factor for causing blackout, is predicted before line outage using pre-outage steady state operating data (PSSOD). In the proposed scheme, support vector machine (SVM) is used as an intelligent predicting tool with information of dominant operating variables (DOV) as its input vector. DOV is a subset of PSSOD which have rich information about line criticalness and associated PCFO of the line outage. DOV can be identified using the joint mutual information (JMI) mathematical method based on the theory of mutual information and entropy between operating variables. In a real-time environment, the DOV associated to each line outage are collected by wide area measurement system (WAMS) in the power system control center (PSCC) and given to the SVM as an input vector for PCFO prediction. Therefore, the proposed scheme can be used as an alarming system for evaluating the PCFO of lines. By predicting PCFO, there is a chance for changing operating point of the power system from insecure mode to secure one by preventive actions. The proposed scheme has been implemented on IEEE 39-bus test system and its performance is validated with the desired results.