The Research On Parameter Identification Of Permanent Magnet Synchronous Machines

Permanent magnet synchronous machines (PMSM) are now widely employed in industrial servo drives, electrical/hybrid electric vehicles, and wind power generators etc, due to high power/torque density and control performance. However, power/torque density and control performance are affected by parameters variation which caused by the temperature rise and flux saturation. So, there are so many control strategies designed for PMSM parameters online identification. The PMSM parameters online identification strategies are divided into three majors species which are based on the extend Kalman filter (EKF), model reference adaptive system (MRAS) and Neural Networks. The major challenges for the online parameter identification are the no precision identification results in the practical applications, computational complex of the identification algorithm and decoupling for the multi parameters identification. In this thesis, lots of work such as theoretical analysis, simulation and experimental work has been done to overcome these challenges.Finite Element Analysis (FEA) based identification strategy which was implemented by FEA software was investigated in this thesis from the machine design aspect. As the compensation of the control theory based identification strategy, the back EMF, cogging torque, output torque etc can be estimated by this strategy.Because of the computational complex of the EKF based identification strategy. A novel improved EKF based strategy which can dramatically reduce the computational complex by dividing the four dimension matrix computation into two dimension computation was introduced in this thesis.Two MRAS estimators which are based on Lyapunov stability theorem and Popov stability criterion, respectively,are developed for identifying the PMSM parameters. Due to its more functional design process and better performance in experimental performance, the strategy based on Popov stability criterion is preferable to use in the MRAS based PMSM parameter identification algorithm design Multi parameter identification decoupling strategy is based on the short time injection of negative d axis reference current, an Adaline NN based multi parameter identification algorithm with d axis current injection decoupling strategy was developed to prove this proposed decoupling strategy. What's more, further investigation on d axis current injection strategy was given, and the relationship between the identification error and injection d axis current value obtained.An inverter nonlinearity compensation strategy was developed in this thesis. This compensation strategy can eliminate the inverter nonlinear affection which can lead to the output voltage distortion and the identification result error. The inverter nonlinearity includes the switch turn on/off delay, switch dead time and voltage drop of the switch device.