Design of New Transformer Protection Device Based on Wavelet Energy Entropy-Neural Network Theory and FPGA

Abstract

Transformer differential protection may be of malfunction at the emergence of inrush current, and it will affect the normal operation of the transformer. So the paper puts forward a new application of wavelet energy spectrum entropy-neural network theory in transformer microcomputer protection, in which the multi-resolution analysis of wavelet transform and information entropy technology are combined firstly and forms a new conception named wavelet energy spectrum entropy, and it will be put into neural network theory as the feature vector, forms the new algorithm in the end. This method decomposes signal through wavelet transform, and  extracts the high frequency part of energy in each scale of wavelet transform from inrush current signal and the short circuit current signal, and calculates the wavelet energy entropy value, which will be as the input feature vector of modified BP neural network. And this feature vector is used as training characteristic value for training in BP neural network. According to the measured data of the system, it has achieved good effect. At the same time, for the large amount of calculation and the high requirements of signal sampling rate in wavelet energy entropy - neural network algorithm, a new idea which uses the high-speed hardware platform of FPGA to realize the algorithm application is put forward, and it will break the bottleneck of traditional microcomputer protection that the MCU would not give consideration to both the speed and the accuracy of the protection at the same time.

 DOI: http://dx.doi.org/10.11591/telkomnika.v11i11.3524