永磁同步馬達速度控制之參數估測與擾動補償

Abstract

現今交流伺服馬達驅動器大多是針對固定負載慣量或是無載所調整的最佳控制參數，所能承受的負載並不足以應付工廠自動化設備的平穩、定速等精度要求。實際上，系統慣量變動(System Inertia Variation)與外界擾動扭矩(External Disturbance Torque)的不確定性嚴重影響伺服系統的動態響應。然而，永磁同步馬達(Permanent Magnet Synchronous Motor, PMSM)接上負載機構時，系統慣量的改變與運轉產生的衝撞、壓力等擾動因素通常是同時存在。本論文探討在系統慣量變化與外部擾動同時存在的情況下，如何維持伺服系統類似無載時的良好性能。研究方法利用擴展型卡門濾波器(Extend Kalman Filter, EKF)線上估測系統慣量，有別以往採用離線式的系統鑑別(例如：最小平方法須收集系統大量數據)，同時藉由估測出的系統慣量值適應性地修正全階擾動觀察器(Full Order Disturbance Observer)內的參考模型參數與速度PI控制器增益，並將估測的擾動扭矩值作前饋式補償電流命令，以彌補伺服系統的動態響應。 在本研究論文中由Matlab/Simulink模擬永磁同步馬達伺服系統針對負載慣量變動與外部擾動介入的情形，詳細探討與分析各個估測參數響應。再由實驗測試觀察速度與電流迴路的重要參數響應與模擬圖形比較。藉由數位訊號處理器(Digital Signal Processors, DSPs)的快速運算，實現永磁同步馬達的速度控制。期望能對工業界與伺服馬達控制領域上有所助益。

In practice, the dynamic performance of PMSM (Permanent Magnet Synchronous Motor) servo driver is influenced by uncertainties of inertia variation and external load disturbance, both would degrade performance seriously. For several industry applications, there are inevitably existed some kind of backlash and stress which would deteriorate original machines. When PMSM join to load of mechanisms, the system of inertia would be changed. In order to satisfy the high performance for speed tracking and load disturbance compensation, accurate values of mechanical parameters are necessary. Even if it was seen on datasheets, moment of inertia should be identified again in the control scheme. As the system of inertia is identified accurately, we use the information of the estimated parameter to adjust the disturbance observer reasonably. This paper proposed an integrated control algorithm which a full order observer to estimate the instant disturbance and online system inertia identification by EKF (Extended Kalman Filter) to auto-tune the disturbance observer and speed PI controller. This control system was tried to ensure system stable in different actual situations. Due to the development of DSPs (Digital signal processors), that can execute accurate computations in short sampling time. The experiment was implemented by DSP platform, and the computer simulation of this paper was performed by the software Matlab/Simulink.