Model Predictive Control for Electrical Drives

Abstract

Since more than 20 years the so-called "field oriented control" is a standard for controlled electrical drives. Today, strategies based on this principle fulfil nearly all demands of drive industry. However, the performance of field oriented controllers cannot be improved any more without limitations. Model predictive controllers (MPC) are based on a rather old approach whose first ideas have been published in the 1960s. Control strategies of this type are characterised by an explicitly and separately identifiable model of the controlled system. This model is used to precalculate the behaviour of the plant and therewith also to choose an optimal value of the control variable. In contrast to conventional predictive controllers used for drive control, which generally precalculate only for a single sample step in advance, model predictive controllers regard the control system behaviour over a long time into the future. Hence, these strategies are also known as "long-range predictive control" (LRPC) in literature. MPC methods have gained much importance in the field of process engineering. It is not known that they have been used for drive control so far, since the huge mathematical effort that is necessary to perform the calculation of the future system behaviour, collides with the demand for high sampling rates in drive control. The paper shows that it is possible to move a huge part of the calculational effort offline so that the remaining part can be easily calculated in a feasible time frame, even for drive control. The applicability of MPC for electrical drives is shown and proven by experimental results