Model Predictive Direct Slope Control for Power Converters

Abstract

Model predictive control (MPC) schemes have become popular in the field of power electronics due to their intuitive formulation, flexibility, and ease of implementation. Typically, these schemes have been implemented with the prediction horizon limited to one time-step, and extension of the prediction horizon over multiple time-steps remains an ongoing area of research. In this paper, a variant of the MPC strategy is proposed wherein the slope of the output trajectories is used to emulate long prediction horizons. Each of the outputs, e.g., current, voltage, torque, or flux, is regulated within a set of symmetrical bounds. When switching is necessitated due to collision with a bound, the switching state that yields the set of output trajectories with the minimum slope, relative to the reference trajectory, is applied to the converter. The key benefit of this approach is its ability to achieve low switching frequencies with a minimal level of computational burden. The feasibility of the scheme, which can be adapted easily to different case studies, is demonstrated through simulations of both a medium-voltage induction machine drive and a grid-connected converter. Experimental results, which are presented for a 1.68 kVA prototype grid-connected neutral-point-clamped converter, further demonstrate the practical viability of the proposed strategy.