Permanent magnet synchronous motor (PMSM) is widely used in high-end equipment. Due to the influences caused by the time-varying physical parameters and the inherent values of these parameters in the a priori model of the plant, weak robustness is a serious problem in the model predictive control (MPC). To solve this problem, model-free predictive control is an improvement in realizing comprehensive performances as the third generation of advanced control technology. 

For the model-free predictive control, the plant and its current operating state are described as an online updating data-driven model which is built based on the sampled data only. Since the model accuracy directly affects the control performances, a suitable model structure and an estimation algorithm conforming to the motion characteristics of the motor are crucial to realizing good accuracy and meeting the requirements of the high-end equipment. 

In a study published in IEEE Journal of Emerging and Selected Topics in Power Electronics, Prof. WANG Fengxiang’s group from Fujian Institute of Research on the Structure of Matter of the Chinese Academy of Sciences designed an autoregressive moving average (ARMA) model-free predictive current control for PMSM driving system to realize enhanced robustness and comprehensive control performances, including good accuracy, well current quality and light calculation burden by the designed ARMA model and normalized least-mean-square (NLMS) estimation algorithm. 

The researchers introduced the implementation of the typical finite-control-set type MPC, which involves an a priori model and highlighted variables that impact parameter mismatches. Over-modulation and under-modulation were avoided in this type of strategy since all candidate vectors were selected by the possible switching states and accurately located on the hexagon center points or vertices of the vector diagram. 

They designed an ARMA model to reflect the motor driving system based on current errors with an expected value of zero. All undetermined coefficients in the model were estimated online by NLMS algorithm where the step length was adaptively normalized to achieve suitable static error and convergence speed according to the required distinct length of current operating states. Two special structures of ARMA including autoregressive and moving-average were analyzed to demonstrate their potential description for the system. Their stability was also discussed to enrich the effectiveness. 

An experimental platform of PMSM was used as an example to demonstrate the effectiveness of the presented control strategy. The simulation and experimental results showed that the presented control strategy obtains improved current quality and robustness compared with the conventional predictive controls under the same operating conditions, and achieves a low calculation burden by the NLMS estimation algorithm in the updating process. 

The presented method is highly compatible and can be applied to other motor driving systems to achieve better robustness and control performance. With the development of processors, there is a potential to further upgrade the control performance by using a more precise model. 

This study provides essential guidance for the improvements of the data-driven model and its accuracy for the model-free predictive control of the motor driving system.