Oskar Sjökvist, MPSYS
FEA-based modelling of an interior permanent magnet synchronous machine
Flux-based model of an IPMSM covering magnetic non-linearities and low speed estimation using mutual differential inductance
conducted at Aros Electronics supervised by Torbjörn Thiringer and Daniel Chädström
Examiner: Torbjörn Thiringer, Dept of Electrical Engineering
Opponent: Henrik Helmius
Interior Permanent Magnet Synchronous Machines are durable, high power-density machines. The motor geometry due to magnet placement inside the rotors introduces salience, and high stator currents causes the magnetic flux within the motor to take non-conventional routes depending on saturation and rotor angle. This thesis investigates a recently introduced motor modelling technique which is able to model the highly non-linear motor behaviour of the stator currents by the use of Finite Element Analysis (FEA) and implements this model in a Simulink environment. The model is based on the flux flow in the motor, and FEA is used to create mappings of the flux flows in the rotor as functions of stator currents and rotor position. The flux model is evaluated against reality and FEA. It shares high likeliness to the FEA-generated results, with a back-EMF peak error within 1% and pulse response error maximum of 8%. The flux model implementation is shown to be around 2 orders of magnitude faster at simulating dynamic motor behaviour than FEA, suggesting that it is an efficient way for accurate Simulation-in-the-Loop (SIL) development.
Further, the thesis introduces a signal injection technique considering non-zero mutual differential inductance and derives the exact response function of the direct-axis input voltage to quadrature-axis current for a square wave. It is shown that the conventional method of signal injection which assumed no mutual differential inductance will be offset from the real angular error by a phase scaling linearly with the mutual inductance term. Simulations using lookup tables for the instantaneous differential inductance in the mutual inductance model based estimator compared to a conventional estimator with constant direct- and quadrature-axis inductance show that the estimator implementation is at best a better choice in 47\% of the cases. This could possibly be improved by basing the selection of inductance values on something other than the previous estimation. Otherwise, it might be beneficial to keep the low speed estimator simple.
Student project presentation
Fredrik Lamm, meeting room, Hörsalsvägen 11, EDIT trappa A och B
21 February, 2020, 15:15
21 February, 2020, 16:15