Model predictive current control of modular multilevel converters

Abstract

This paper proposes a model predictive controller for high-power modular multilevel converter operating at low switching frequencies. The controller regulates the load currents along their reference trajectories, controls the circulating currents and controls the sum of the capacitor voltages per branch. Upper limits on the branch currents and capacitor voltages are imposed in the controller formulation. The controller manipulates the voltage references of a carrier-based pulse width modulator. A subsequent balancing controller maintains the capacitor voltages within each branch around their average voltage. Unlike hierarchical schemes based on multiple PI control loops, the proposed controller achieves not only a very good performance at steady-state operation but also very fast current responses during load transients, while maintaining the branch currents and capacitor voltages within their safe operating limits.

Model Predictive Control of Power Electronic Systems: Methods, Results, and Challenges

“…Regarding indirect MPC algorithms with a subsequent modulation stage, penalization of the control effort allows for less aggressive control actions and gives rise to smoother control [90], [99]. Moreover, given that λ u > 0, where λ u is the weighting factor of the control effort, the resulting QP is a strictly convex optimization problem [136], thus it guarantees the uniqueness of the solution [2].…”

“…As for indirect MPC, long horizons improve the performance of such algorithms as well [99]. Similar to direct MPC, the optimization problem of explicit MPC is an IP-namely an (M)IQP-meaning that the same computational challenges exist.…”

“…Many of the derived MPC algorithms and/or solvers assume linear systems of the form (4) with integer or real-valued inputs, see, e.g., [26], [105]. Given that many power electronic systems do not meet such a prerequisite, this means that they either have to be linearized [99], [203], thus detracting from the accuracy of the model, or other methods need to be developed and/or solvers employed. To tackle these issues, MPC needs to be addressed as a nonlinear problem [204].…”

“…In doing so, the optimization problem can be cast as a constrained or unconstrained QP. The former is relatively easier-compared to the aforementioned methods-to solve in real time owing to the several existing off-the-self solvers, as discussed later in this survey [99]- [104]. The latter allows for an analytical solution of the MPC problem [105]- [108].…”

A Survey of Low Switching Frequency Modulation Techniques for Medium-Voltage Multilevel Converters

“…However, minimization of switching losses has not been accounted in the cost function. Later, PWM-based MPC technique has been reported for MMC to achieve constant switching frequency and low device switching frequency [146]. Also, MPDCC technique with long prediction horizon has been implemented for MMC to track ac-side currents within hysteresis bounds, while minimizing circulating currents, SM capacitor voltage fluctuations and device switching frequency [147].…”