Model Predictive Direct Current Control of MMC

Modular multilevel converters (M2LCs) are typically controlled by a hierarchical control scheme, which essentially requires at least two control loops: one to control the load current and another to control circulating currents. This paper presents an M2LC with a single controller, which is based on model predictive direct current control (MPDCC) with long prediction horizons. The proposed MPDCC scheme maintains the load current within tight bounds around sinusoidal references and minimizes capacitor voltage variations and circulating currents.

An internal prediction model of the M2LC is used to minimize the number of switching transitions for a given current ripple at steady state while providing a fast current response during transient conditions. A state-space model, which is generalized for an N number of modules per each arm of the M2LC, is also presented to investigate the dynamic behavior of arm currents and capacitor voltages. Simulated performance of the converter, under various operating conditions, is presented in comparison to measured performance of a single-phase, three-level 860-VA M2LC prototype to demonstrate the proposed MPDCC philosophy.