A Comparative Evaluation of Data Mining Techniques For Fault Classification in TCSC Lines Using Multiclass Classification Approach

This paper presents a comparative evaluation of data mining techniques such as Random Forest (RF), Support Vector Machine (SVM), and Naïve Bayes Tree (NBTree) classifiers for classifying faults in Thyristor Controlled Series Compensated (TCSC) transmission lines. The performance of these classifiers has been evaluated as a multi-class classifier for classifying shunts faults in TCSC lines. The inputs given to the classifiers have been derived from instantaneous currents. The validity of the suggested technique has been tested by modeling a power system with 300 km long, 400 kV TCSC line in PSCAD/EMTDC. Wide variation in fault parameters and system conditions have been considered to generate faults cases for evaluating performance of the classifiers. The training of the classifiers has been carried out with 3,840 fault cases. The performance evaluation with 14,400 test fault cases indicates that the performance of RF classifiers has been consistent in comparison to SVM and NBTree classifiers for classifying faults in TCSC lines.