Predictive control has been recognized as a highly adaptable and accurate methodology within various power electronics applications, particularly in motor drives. Nevertheless, obstacles persist in its practical application specific to motor drives. A critical concern associated with model predictive control is its dependence on the load model, which, in motor drive scenarios, often displays heightened uncertainties and fluctuations. The uncertainties present in motor drives can be classified into two distinct categories: 1) Parameter mismatch, which pertains to alterations in the electrical or magnetic characteristics of the motor or inverter, such as the escalation of stator resistance in response to temperature increases or core saturation effects. 2) Operating point, which includes deviations in torque or speed, with load disturbances acting as a representative illustration.

Recently, a new methodology has been developed to reduce the method's dependence on the physical model, which is called Model-Free Predictive Control. The model-free predictive method can be categorized into three main groups: 1) data-driven model-based predictive methods, 2) ultra-local model-based predictive control, and 3) combined model-based predictive control. These methods will be introduced in this tutorial, and a comprehensive comparison will be performed.