Modeling and parameter estimation of AC machines

A new approach to identify the linear and the nonlinear models of the induction machine using the time domain test data and the feedforward neural network model (FNN) is presented. The estimated machine model is suggested to be utilized for the power system transient stability study and the computer-controlled electric drive system.;For the linear machine modeling, a step voltage disturbance is applied across the stator terminals while the machine is in standstill and the transient responses of the stator voltage and current are recorded. The initial parameters are obtained from the time domain response model, and the maximum likelihood (ML) parameter estimation algorithm is used to estimate the transfer function and the equivalent circuit model parameters.;Free acceleration tests are performed for the nonlinear machine modeling and the resulting responses of the stator voltage, current and the rotor angular velocity are measured. A simplified nonlinear induction machine model in which all the parameters are considered to be nonlinear is proposed in the study, and the model parameter sets for various operating conditions are estimated using the ML algorithm. The FNN models are used to represent the nonlinear parameters as functions of the operating conditions.;For the validation, the performance of the estimated linear model is compared to the measured step voltage response, the standstill frequency response (SSFR) and the free acceleration test data. The estimated nonlinear model is also validated by comparing its performances to the measured free acceleration data using the measured and simulated operating condition data that are used as input patterns to the FNN nonlinear parameter model.;The study results show that the estimated linear model can be used for small disturbance studies, and the nonlinear model can be used for large disturbance transient response studies in which machine windings are to be deeply saturated.