Improvement in Transformer Differential Protection Using Singular Value Decomposition

This paper presents a new transformer differential protection technique based on singular value decomposition (SVD). Initially, transformer winding currents are converted into their respective real-time series hankel matrix, and the SVD of that matrix is calculated. The singular values of the transformer winding current on the primary and secondary sides are compared and used to detect internal faults in the transformer. The suggested SVD-based method is not only capable of detecting fault conditions faster than conventional differential protection scheme but also able to restrain its operation during magnetizing inrush, overexcitation, and CT saturation during a heavy through fault. The performance of the proposed scheme is evaluated using numerous simulation cases as well as practical field data. The simulation cases are generated by modelling the existing power system network in PSCAD/EMTDC software, and the protection scheme is developed in a MATLAB environment. The proposed scheme's average response time to detect internal fault is 5 ms, which is faster than the conventional differential protection scheme, indicating the superiority of the proposed method.