Empirical Mode Decomposition and Optimization Assisted ANN Based Fault Classification Schemes for Series Capacitor Compensated Transmission Line

This paper presents two intelligent classifier schemes for classifying the faults in a series capacitor compensated transmission line (SCCTL). The first proposed intelligent classifier scheme is a particle swarm optimization-assisted artificial neural network (PSO-ANN). The second, proposed one is a teaching-learning optimization-assisted artificial neural network (TLBO-ANN). For each type of fault, the 3-phase current signals are acquired at the sending end and processed through empirical mode decomposition (EMD), to decompose into six intrinsic mode functions. The neighborhood component analysis is used to extract the best feature intrinsic mode functions. From the identified best feature intrinsic mode functions, the energy of each phase of the line is computed. The energy of each phase is fed as inputs for both PSO-ANN and TLBO-ANN classifiers. The practicability of the proposed intelligent classifier schemes has been tested on a 500 kV , 50Hz, and 300 km long line with a midpoint series capacitor using MATLAB/Simulink Software. The results demonstrate that the classifier schemes are able to accurately classify faults in less than a half-cycle. Furthermore, the efficacy of the proposed intelligent classifier schemes has been evaluated using Performance Indices including Kappa Statistics, Mean Absolute Error, Root Mean Square Error, Precision, Recall, F-measure, and Receiver Operating Characteristics. From the results of Performance Indices, it is concluded that the proposed TLBO-based artificial neural network classifier outperforms the PSO-based artificial neural network classifier. Finally, the efficacies of proposed intelligent classifier schemes are compared to existing approaches.