| Abstract |
Considering the sensorless control of PMSMdrives, a novel observer for improving the estimationperfomance is proposed in this paper. An Adaptive linearneuron (ADALINE)-based feed-forward harmonic compensationblock is integrated into the classic PLL, whichtogether form an ADALINE-PLL. Firstly, given a generalinput, the estimation error of a classic PLL is analyticallyinvestigated. It is shown that the estimation error containsseveral distinctive harmonics, which are related to thePLL non-linear structure and input harmonics. Theseharmonics are able to pass through the loop-filter andappear in the estimated phase and frequency as extraripple. In light of the estimation error harmonics, theADALINE-PLL is used for suppression of undesired harmonics.Secondly, the superior perfomance of ADALINEPLL,compared to PLL, is verified through simulationsin several different conditions. In case of acoustic noisebasedor EMF-based PMSM sensorless control, the observerinput signals contain several harmonics, which caneasily distort the estimation performance if not properlyrepressed. As the result, the positive impact of ADALINEPLLon the estimation performance is studied in details. |
