Power Transformer Condition Monitoring and Fault

Abstract

Differential protection is the best protection technique for power transformers. However, Differential relays cannot provide discrimination between internal fault and inrush currents; they always detect the inrush current as an internal fault state. This paper presents a proposed algorithm based on Support Vector Machines (SVM) for three phase two windings transformer. The algorithm diagnosis different transformer conditions: normal, over excitation, inrush current, internal and external faults. It uses the primary and secondary currents and second order harmonics of these currents to distinguish between these conditions. The transformer conditions are simulated using PSCAD/EMTDC in order to obtain the primary and secondary current signals. These signals are then used to train and test the SVM using MATLAB/Simulink software. Results obtained using the proposed SVM algorithm are compared with corresponding results obtained using Artificial Neural Network (ANN)-based method. The results show that the proposed algorithm is accurate and very fast in detecting and classifying different transformer conditions.