DEVELOPMENT OF A NEURAL-NETWORK MODEL FOR ROTOR ANGLE ESTIMATION

This paper presents a neural network model to measure the rotor angle of the synchronous machine. The network model is established based on the mapping capabilities of multi-layer feedforward neural networks an d weight estimation of the back-propagation learning rule. A neural ne twork is then developed using simulation data, obtained by applying Pa rk's equations for the dynamics of the synchronous machine in the d-q axis reference frame. Following successful application of the network model based on simulation data, data from laboratory measurements of t erminal variables of a 5 kVA test machine is then applied to develop t he neural network model for rotor angle measurement.