Neuro-fuzzy Technique for Power Transformer Protection

Abstract

In the quest of more reliable power transformer protection, differential protection is considered the best scheme. In the proposed scheme, various operating conditions of transformers are distinguished by virtue of the signatures of differential current. As the conditions of internal fault and magnetizing inrush do have some of the signatures common among them, it is becoming increasingly important and difficult to distinguish between magnetizing inrush and fault conditions for differential relaying. In this direction, both feature-based, as well as pattern-based, approaches are used. In this article, a new approach, based on neuro-fuzzy technique, is presented for power transformer protection that ensures relay stability against external fault, magnetizing inrush, sympathetic inrush, and over-excitation conditions and its operation on internal faults. This approach is able to handle the “vague” information rather than only the “crisp” information. In the proposed method, fuzzy back-propagation neural network (FBPNN) is used as a core classifier to discriminate between magnetizing inrush and internal fault of a power transformer. An algorithm has been developed using an optimal number of neurons in the hidden layer as well as in the output layer. The effect of hidden layer neurons on the classification accuracy is analyzed. The algorithm makes use of voltage-to-frequency ratio and amplitude of differential current for detection of transformer operating conditions. The performance of BPNN, radial basis function neural network (RBFNN), and probabilistic neural network (PNN) are compared with the proposed fuzzy BPNN. Extensive simulation studies have been performed to demonstrate the efficiency of the proposed scheme using PSCAD/EMTDC and MATLAB.