Support Vector Regression Machine with Enhanced Feature Selection for Transient Stability Evaluation

This paper presents a support vector regression machine (SVRM) to predict the energy margin (EM) of power systems subjected to severe disturbances. The nonlinear relationship between the pre-fault, during-fault and post-fault power systems parameters and the degree of stability of the system under post-fault state is captured by the SVRM trained offline. Significant generators are selected by feature selection based on the sensitivity of stability margin and the features other than generators are selected based on a step wise feature selection by three fold cross validation. The performance of the proposed SVRM predictor is demonstrated through the simulations carried out on 17 generator reduced Iowa system.