Impact of the Prediction Error on the Performance of Model Predictive Controllers with Long Prediction Horizons for Modular Multilevel Converters - Linear vs. Nonlinear System Models

The closed loop control performance of MMCs can be significantly improved by using Model Predictive Control (MPC). This paper evaluates an MPC algorithm based on a linearised MMC model regarding the performance limits caused by the prediction error due to the linearisation. To decrease the prediction error to a minimum and to improve the performance it is proposed to use a nonlinear MMC model as a prediction model for the MPC. The steady state and transient performance of the MPC with both MMC models is compared in detail using simulations to analyse the effect of the prediction error on the control performance.