Model predictive control of power converters: Control algorithms and numerical optimization methods

Tobias Geyer

ABB Corporate Research

Tuesday, February 20, 2018, 13:30

"Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany"

Building on the introductory lecture, the second lecture provides a more in-depth view of MPC algorithms for power converters by focusing on direct and indirect MPC. Direct MPC combines the inner current control loop with the modulation stage, thus generating the converter gating signals without the use of a separate modulator. To simultaneously reduce the inverter’s switching losses and the current distortions, long prediction horizons are required. The underlying integer optimization problem can be solved relatively quickly by adopting sphere decoding, which is a branch and bound method.

Alternatively, indirect MPC schemes require a modulator. The current control and (some of) the outer control loops are addressed in one multiple-input multiple-output controller that ensures operation within state and input constraints. This MPC formulation is particularly promising for systems that are intrinsically difficult to control, such as modular multilevel converters. Both MPC concepts will be reviewed and results for inverters with LC filters and modulator multilevel converters will be discussed.