Identification of Induction Machines Stator Currents with Generalized Neurons

A new approach to identify the nonlinear model of an induction machine using two generalized neurons (GNs) is presented in this paper. Compared to the multilayer perceptron feedforward neural network, a GN has simpler structure and lesser requirement in terms of memory storage which is makes it attractive for hardware implementation. This method shows that with less number of weights, GN is able to learn the dynamics of an induction machine. The proposed model is made by two coupled networks. A modified particle swarm optimization algorithm is designed to solve this distinctive GN training problem. A pseudo-random binary sequence signal injected to the induction machine operating at its rated value was chosen as the test input signal. For validation, the trained GN model is applied on the different operating conditions of the system.