Control of a Permanent Magnet Synchronous Motor (PMSM) with constraints
Abstract
The aim of the thesis is to design and implement a controller that controls the PMSMwithout violating the given constraints. This is especially with respect to the batterycurrent.
Two approaches has been implemented in this paper, where both of themhave shownto satisfy the given constraints. This was however on expense on the performanceof some of the responses, especially id (t ). As a Surface PMSM (SPMSM) has beenused, a nonzero value on id (t ) only contributes in heat increase in the motor. Theapproaches have shown different overall performance of these responses. The simulationsin this paper was carried for a given test case where an angular position referencewas followed. Both approaches have utilized a cascade of P- and PI-controllers,where one replaced the current controllers with a predictive controller, while theother utilized saturation blocks.
The predictive controller has been realized by using a Nonlinear Model PredictiveControl (NMPC) structure, as the constraint for the battery current was found to benonlinear. In this paper two versions of the NMPC has been looked at, where oneconsidered constant angular velocity in the prediction horizon and the other did not.Thiswas motivated by the assumption that a constant velocity in the controller mighthave significant reduction on the computational time. Based on the results however,this paper has shown that including angular velocity in the prediction horizon mightbe a better choice. This is in terms of the performance and computational time presentedin the results. However as a framework that does not support c-code generationhas been used, this paper cannot give any direct results on the computationaleffort the predictive controllers induce.
The work of this paper should therefore be regarded as an introductory on futurework. This is if it is of interest to use the NMPC presented here for controlling themotor. The results alone are not of very big significance as the control structure currently
is not deploy-able. However for further work, the results give insights in whatto expect when utilizing a framework that supports c-code generation.The other approach that have utilized saturation blocks can be used directly. Howeversome additional tuning might be required if the performance is not sufficient.