A Predictive Nearest Level Control of Modular Multilevel Converter

Model predictive control (MPC) is an emerging control technique which can include the multiple control objectives, nonlinearities and system constraints easily. The main drawback of MPC for modular multilevel converter (MMC) application is its large computation, since the high level MMC has huge combination of switching states. Choosing the weighting factors of the cost function for MMC is also very complex. In this paper, a new predictive nearest level control (PNLC) is proposed, which uses the predicted values of control objectives to directly calculate the optimal number of on-state sub-modules (SMs) in each arm, then determines the specific SMs to be switched on according to sorting strategy for capacitor voltage balancing and average switching frequency reduction. In this improved predictive control, no calculation of cost function under different combination of switching states is needed. It is suitable for low and high level MMC. The results are verified through simulation, which shows the proposed control technique has better performance than conventional MPC.