Dead-Time Compensation for Model-free Predictive Control in Multilevel Inverters

Model predictive control (MPC) has been widely applied to various medium-voltage power converters due to its advantages in good dynamic performance and multi-objective control. However, the control algorithm requires the accurate model in the controlled system. As a result, the parameter mismatch between the actual model and control module leads to an undesired performance of MPC. Therefore, model-free predictive control (MFPC) has been developed in the literature to solve this issue. This approach uses a recursive least square (RLS) algorithm to estimate parameters in the model based on the measured values. In such systems, the dead-time makes the model nonlinear and leads to a poor prediction in the output currents and leads to the current distortion. In this paper, a dead-time compensation algorithm based on MFPC method is proposed to obtain a linear model and reduce the negative effects of dead-time. This can be done by compensating the phase voltage in the controller with the voltage second theorem. The dead-time compensation method is verified by simulation results.