Research On High Performance Control Strategies For Pmsm Based On Parameter Identification Algorithm

In the industrial field, especially for the automatic production field such as robots and numerical control machines, the servo system with permanent magnet synchronous motor plays a pivotal role. Recently, along with the rapid development of automatic field, there is more and more demanding for high performance PMSM servo system. But among the high performance servo products, the servo products of Japan and Euro-America are occupying the most market share. The main gap of servo products between domestic and developed countries are: high performance current control function, and the parameter self-tuning advanced function, etc. For the current control strategies, most of the domestic servo products adopt the PID regulating method based on coordination transformation. But this method can not compensate the influence of the inherited cross coupling effect and the back electromotive force for the current responses.The goals of this paper are to improve the performance of servo system, and the main research topics are high performance current control strategies and the high performance algorithms in speed loop. The details are as follows,By comparing the currently mainstream current control strategies, it could be found that the predictive current control strategy based on dead-beat control can make the current loop bandwidth to be the optimal condition, i.e. one twelfth of the modulation frequency. Therefore the paper adopts the dead-beat predictive current control strategy to control the current loop of servo system. But it could be found by the theoretical analysis that, the accurate motor parameters are critical for the current control system, especially for the motor inductance parameters. If the inductance error exceeds two times, the current control system will be divergent. In order to improve the robustness of the predictive current control system, the robust predictive current control based on Luenberger observer is proposed. Then the quantitative relationship between system performance indexes and the observer coefficient are given. The robustness of the current control system will be improved obviously by adopting the current observer.Although the robust predictive current control decrease the sensitivity of the current control system for the model parameters, but it could be found from simulation and experiment that, in the speed dynamic process, the inductance error affects the direct-axis current response, the flux affects the quadrature-axis current response, and the effect of resistance error is small enough to be neglected. Otherwise the flux error will result in the difference between current references and feedbacks when the system working in steady state. In order to further improve the current loop performance, this paper proposes a predictive current control strategy based on recursive least square parameter identification method. Based on the characteristics of the motor current-voltage equations, the identification process is divided into two parts, firstly the inductance is identified, and then the identified inductance is treated as a known parameter to identify the motor flux and resistance. The simulation and experiment show that, on the basis of stable system, the proposed algorithm could further improve the current loop performance of the servo system.The regulator saturation is an inevitable phenomenon in the speed dynamic process, due to the limitation of system hardwares, which is also called Windup phenomenon. It will increase the overshoot and settling time of speed response. In order to compensate the speed regulator saturation, this paper proposes the segmented tracking back calculation Anti-Windup strategy based on the traditional tracking back calculation Anti-Windup strategy. It could restrict the saturation depth of the regulator, and makes the speed performance indexed independent on the system references. But the method is invalid when it is applied in a discrete controller. Furthermore, the paper proposes the integral state prediction AW strategy. The method calculates the steady-state value of integrator when the regulator is saturated, and treats the value as the initial value of the integrator when the regulator just enters into linear zone. The proposed method could decrease the overshot and settling time of the speed response, and makes the speed response to be the optimal. In order to realize the PI parameter self-tuning function in speed loop, the speed loop state function of the servo system is analyzed firstly, it can be seen that the load torque and load inertia are the most important two variables in the parameter self-tuning strategy. In order to get the load torque information in this paper, the Luenberger full order observer is adopted to realize the load torque observation function. Due to the algorithm is very complex and there are three adjustable observer coefficients, then the paper proposes another load torque observer based on Gopinath’s theory. Furthermore the observer performance is analyzed and the value range of the observer coefficient is also given. In the inertia identification algorithm based on recursive least square method, the identified result has a significant oscillation when the speed changed, and the algorithm is of poor immunity for the system noise. Hence the paper proposes the inertia identification strategy based on the FOREFOP algorithm, according to the simulation and experiment, the proposed algorithm could suppress the system noise effectively and improve the identification precision. At last, combine with the identified load torque and inertia information, the paper proposes an optimal parameter calculation method of the speed loop PI regulator in frequency domain, in order to realize the function of speed loop PI parameter self-tuning.