Harmonic restraint differential protection of power transformer based MRBFN

This paper presents a minimal radial basis function neural network (MRBFNN) scheme for harmonic restraint differential protection of power transformers. The minimal resource allocation network ( M-RAN) learning algorithm which is a sequential learning radial basis function neural network is shown to realize networks with far fewer hidden neurons with the better or same approximation/classification accuracy without resorting to trial and error. Performance of this model is compared with the usual one, i.e., the feedforward backpropagation (FFBP) model. The results show that this new algorithm is better in terms of accuracy and speed with respect to detection of faults and requires less training time. The proposed protection scheme has been evaluated using simulated data obtained through the EMTP/ATP package.