Novel classifier design for optimising the accuracy for identification of disturbance in power system

Abstract

Origination of the disturbances in the power system is not a new problem and various technologies have evolved while various research-based contributions have addressed this problem to a large extent. Reviewing existing techniques shows much usage of iterative optimisation while designing the classifier. Hence, higher accuracy arrives at the cost of computational complexity in the existing system. Therefore, the proposed system introduces a novel framework where short-time Fourier transform is applied on the numerical-modelling on power signal generation corresponding to practical stages of various disturbances. A simple and novel feature extraction process is applied, which causes the system to enhance its potential for the classification process. Adopting IEEE 1159 standard, the disturbances of power system has been modelled, which is trained using dual-layer feedforward network. Comparative analysis shows proposed system outperforms conventional techniques by offering higher accuracy with lower computational complexity. From the analysis, it is observed that the proposed concept of detection offers higher accuracy of the classification within extremely very less iteration with less execution time of 2.6551 sec.