Model predictive control (MPC) is an optimal control technique that is capable of taking nonlinear systems and system constraints into account. It can be designed to use a modulation scheme (PWM/SVM), which is called the convex control set (CCS) approach, or actuating directly a switch state, which is called the finite control set (FCS) approach. MPC is defined by a constrained optimization problem that uses a cost function to define the control goals and takes N future sampling periods into account, which is called the prediction horizon.
MPC can be used to achieve a high control performance in dynamic (fast transient response) and steady-state (low switching and conduction losses) operation of power electronic and drive systems. This is achieved solving the constrained optimization problem at each sampling instant. Since power electronic sampling periods are typically small, the online execution time is critical. Thus, the problem statement needs to be formulated in a sufficiently simple manner and computationally efficient solvers are required. Also, MPC relies explicitly on the dynamic system model (other controllers usually rely implicitly on a model), which requires the generation of acceptable models. Last, the “cardinal” stability theorem of MPC can be tricky to apply for systems with short prediction horizon and can require the use of alternative approaches.