Minimal radial basis function neural network based differential protection of power transformers

This paper presents a new approach for the protection of power transformers using a Minimal Radial Basis Function Neural Network (MRBFNN). 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 better or the same approximation/classification accuracy; without resorting to trial and error. The performance of this algorithm is compared with the Multilayer Feed-forward Networks (MFNs). The number of training patterns and the training time are drastically reduced and significant accuracy is achieved in the classification of different events and detection of faults in power transformers using computer simulated tests. Copyright (C) 2004 John Wiley Sons, Ltd.