Assessment of voltage stability margin by comparing various support vector regression models

Voltage stability assessment and prediction of loadability margin are the major concerns in real-time operation of power systems. This paper proposes a support vector machine (SVM) regression network for the voltage stability assessment for normal condition as well as for contingency cases. The loadability margin of any given operating conditions is obtained for pre-contingency and post-contingency based on the computation of a stability index. SVM takes real and reactive power at all buses of the system and gives the loading margin. The validity of the proposed SVM-based index is tested on IEEE 30 and Indian 181 bus systems. The results of the proposed method are compared with neural network, extreme learning machine, online sequential extreme learning machine and extreme support vector machine regression methods. The feasibility of application of the proposed SVM regression network for real-time stability assessment is discussed. Also, FACTS devices are produced to improve the system loadability and their results are discussed.