Neural network synchronous machine modeling

Abstract
A set of training data is generated from a simulation of the dynamics of the synchronous machine in order to train the multilayered feedforward neural network. Two different structures are compared. The first structure is a three-layer feedforward network, the hidden layer containing ten nodes. The second structure is also a three-layer feedforward network, the hidden layer containing 20 nodes. A step response to a step input on the field voltage is simulated. The simulation results show that neural networks can basically capture the synchronous machine dynamics, and an increased number of hidden nodes per layer can increase the accuracy of the model.

This publication has 6 references indexed in Scilit: