Robust Model-Predictive Control for High-Performance Electric Drives


Felix Plum

Wednesday, May 22, 2019, 13:00

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

"Though often invisible, induction machines permeate our everyday-lives in an increasing manner. With the advent of electrical cars, the requirements for motor control have further increased. Techniques from model predictive control have been shown to be suitable for e.g. achieving reduced energy consumption, noise-levels or increased driving performance. In this thesis, a strategy to robustly and optimally control a dual three-phase asynchronous machine in the sub-millisecond range was devised and implemented. The contribution of this thesis is threefold: Firstly, a starkly simplified robust optimal control problem is proposed and framed in the adjoint SQP framework to obtain an improved solution at a marginal increase of computational complexity. The main idea lies in propagating uncertainty along the previous state trajectory and to add a correction term to the objective, which is cheaply evaluated. Secondly, the effect of constraint robustification on the control of an asyn- chronous machine is investigated in simulation. Thirdly, the newly developed acados framework for rapid optimal control prototyping is cross-compiled, deployed on a dSpace platform and successfully used to robustly control a dual three-phase asynchronous machine with average MPC-turnaround times of 90µs. Contrary to the nominal controller, no violations of the current constraints are observed."