Sensorless Control of AC Machines

1. Sensorless ControlofAC Machines Marie Curie ECON2 Nottingham Summer School 08 Cedric Caruana

2. Objectives • To review the sensorless control of ac machines at low and zero speed • To present two techniques: • Zero Vector Current Derivative Technique for PMSM • Use of PWM Harmonics for IM Rotor Position Detection Slide: 2

3. 1.Sensorless Control of AC Machines at Low and Zero Speed Slide: 3

4. Topics • Background • Injection and Demodulation Techniques • Multiple Saliencies • Saturation Saliency Shift with Load Slide: 4

5. Why go Sensorless • Objective is to enable vector control without the need of the encoder on the machine shaft • Gains • remove drive dependency on sensors that are external to itself • cost, robustness, reliability • Which market • precision drives, possibly integrated solutions • lower cost, general purpose drives Slide: 5

6. Methods for Sensorless Control • Fundamental model based Methods • simple realization, however: • parameter dependent • generally fail at low and zero frequency • Signal Injection Methods • exploit saliencies that are not seen by fundamental signals • excite machine at much higher frequency than fundamental • injection setting can ensure that high frequency effects are superimposed to fundamental machine operation • rotor- or flux- position obtained indirectly through response of machine (reflects hf impedance) • parameter independent Slide: 6

7. L qr(elec) AC Machine Saliency • Salient Pole machine: geometric saliency • Symmetric machines • geometric saliencygenerally negligible • saturation saliency (main fieldand local saturation) • rotor slotting saliency (IMs) ls Slide: 7

8. Signal Injection Methods • High Frequency Carrier Injection • Rotating Carrier Injection • Pulsating Carrier Injection • Transient Injection • Test Voltage Vector injection superimposed on fundamental PWM • Standard PWM Switching • exploit the switching of the fundamental PWM waveforms Slide: 8

9. Signal Injection Methods:High Frequency Rotating Carrier Injection • derives two orthogonal position signals • can detect instantaneous rotor / flux position • different demodulation schemes: heterodyning, synchronous filters • easy to implement • requires no additional sensors Slide: 9

10. derives frame position error signal • tracking observer to obtain • easy to implement • requires no additional sensors Signal Injection Methods:High Frequency Pulsating Carrier Injection • Injection in estimated dqe frame Slide: 10

11. Signal Injection Methods:Transient Injection • test voltage vector superimposed on fundamental PWM • can detect instantaneous rotor / flux position • needs to measure current derivative • requires synchronous sampling of current derivative • simple combination of readings to obtain position • can be realized WITHOUT test vector injection, using fundamental PWM switching Slide: 11

12. Comparison of Methods • Different levels of complexity in setting up the injection • hf carrier injection obtained easily through same hardware but observing demodulation scheme complex. Tracking scheme is easy. • transient and PWM switching schemes require synchronization of sampling but algorithm is very easy • Latter techniques require extra sensor. However these can be integrated in the drive. • industrial current transducers have a current derivative signal available internally (Kennel) Slide: 12

13. Corrupting Harmonics • Ideally machine will only exhibit one saliency • Practical machines will exhibit multiple saliencies; the saliency that is not tracked acts as a disturbance corrupting the position signal/s • Similar effect if the saliency distribution is not sinusoidal • Need to couple the unwanted saliencies to improve the estimation accuracy Slide: 13

14. Corrupting Harmonics:Saturation and Rotor Slotting Effects (IM) • both rotor geometry and saturation saliency present • saturation saliency acts as a disturbance Slide: 14

15. Corrupting Harmonics:Higher Order Saturation Harmonics • Ideal and experimental position loci over the stator current angle • Harmonic spectrum of position signal pa for closed slot IM under motoring conditions: (a) 20% and (b) 100% rated torque at 60r/min Slide: 15

16. Corrupting Harmonics:Nonlinearity of the PWM Inverter • Effect caused by the inverter, not the machine • Generates additional harmonics that coincide with the harmonics of the saturation saliency in high frequency signal injection drives • standard, simple dead-time compensation strategies not effective • complex compensation schemes published in the literature • Less effect on transient excitation schemes Slide: 16

17. Decoupling of Corrupting Harmonics • Various engineering solutions proposed • Harmonic compensation table (frequency approach) • table stores frequency, amplitude and phase of different harmonics • not effective against inverter nonlinearity effects • Space Modulation Profiling (SMP) (time approach) • table stores corrupting magnetic signature of the machine (obtained during commissioning stage) • effective against inverter nonlinearity effects • tedious to commission • Neural Networks • ease the commissioning of the table • Synchronous filters with memory Slide: 17

18. Decoupling of Corrupting Harmonics:Space Modulation Profiling (SMP) • Open slot IM under rotating hf carrier injection • shows clear signs of inverter nonlinearity effects • Closed slot IM under transient excitation • not influenced by inverter nonlinearity effects • Both profiles quite complex Slide: 18

19. Decoupling of Corrupting Harmonics:Real Time Implementation using SMP Table • SMP table referenced through measurable variables like stator current angle • Can compensate saturation saliency, higher order saturation harmonics and inverter nonlinearity effects Slide: 19

20. Decoupling of Corrupting Harmonics:Considerations • How complex is this compensation • more drive memory required; not considered a problem • Commissioning required. What tests can be done? • How often do we need to commission • might not be portable to different machines • Do we go for a high quality inverter with less ‘nonlinearity’ effects ? • will be more costly • Do we go for off-the-shelf machine or custom machine? • will depend on the application • Can we define criteria for ‘sensorless friendly’ machines? • lcd≠ lcq not sufficient Slide: 20

21. Load Dependent Saturation Saliency Effects:Phase Displacement • Load dependent phase displacement between the identified and real flux position • Displacement will depend on: • machine: type, construction • injection scheme used • Closed slot IM • relationship is nonlinear • Upper: hf injection (max of around 30°) • Lower: transient injection (max of around 18°) Slide: 21

22. Load Dependent Saturation Saliency Effects:Compensation of Phase Displacement • Linear approximation possible based on armature reaction effect • parameter dependent • not applicable to closed slot IMs • Lookup table referenced through isq* • Might be less on PMSM due to design (Kennel) Slide: 22

23. Load Dependent Saturation Saliency Effects:Considerations • More memory required • Not portable to different machines, hence need of commissioning • Do we go for purposely designed machine where this phenomenon is insignificant or predictable Slide: 23

24. 2.Zero Vector Current Derivative Technique for PMSMs Slide: 24

25. Objectives • Sensorless Operation of a PMSM Drive without Additional Test Signal Injection • To define a position error signal utilizing both back emf and PMSM magnetic saliency • Analysis and Decoupling of Inverter Nonlinearity Effects • Sensorless operation Slide: 25

26. Mathematical Model • Consider the voltage equations for a PMSM, in a dq frame orientated to the rotor flux: • On the application of a zero voltage vector: Slide: 26

27. Saliency Back EMF Definition of Position Error Signal • Under sensorless operation, drive operates in estimated dqe frame • current control forces current in estimated de axis to zero (i.e. ide =0) • In the dqe frame, assuming ide=0 Slide: 27

28. Examining the Position Error Signal Slide: 28

29. Proposed Tracking Controller Slide: 29

30. Test Rig Slide: 30

31. Acquisition of Current Derivative • TPWM = 266.7s (fPWM = 3.75kHz) • TADC = 66.7s Slide: 31

32. perr with iqe = 1 A at 1 rpm • perr with iqe = 0 A at 6 rpm • perr with iqe = 1 A at 6 rpm Position Error Signal Components:Plotting perr against dqe frame position error Slide: 32

33. in real dq frame versus stator current angle Inverter Nonlinearity Effects Slide: 33

34. Inverter Non-linearity Effects Slide: 34

35. Inverter Non-linearity Effects • Theoretical • Experimental Slide: 35

36. Sensorless Torque Control at higher speeds • Torque control in quadrants II and III, -12rpm  55% rated • Torque control in quadrants I and IV, 12rpm  55% rated Slide: 36

37. Speed change from 0 to 6 (0.4 Hz ele) to 0 rpm at ide* = 0.1 A • Speed change from 0 to 6 (0.4 Hz ele) to 0 rpm at ide* = 11 A Sensorless Torque Control at Low Speed • Speed change from 0 to 6 rpm (0.4Hz ele) at ide* = 0 A Slide: 37

38. Sensorless Position Initialization • Start from zero speed, rated current • Initial position intentionally set wrong Slide: 38

39. Sensorless Speed Control • Position error signal polarity correction function Slide: 39

40. Sensorless Speed Control •  30rpm speed transients Slide: 40

41. Conclusions • The Principle of the Zero Vector Current Derivative Technique was shown • Back EMF signal is strong enough for sensorless control providing the possibility of implementing this algorithm in any PMSM, even if no saliency is available • Saliency component allows operation of the drive even down to zero speed • Limitation of the method in the very low speed region in two of the four quadrants of operation • dide/dt error signal polarity needs to be adjusted for four operation quadrant operation • Torque and speed control of the sensorless PMSM were shown Slide: 41

42. 3. Use of PWM Harmonics for IM Rotor Position Detection Slide: 42

43. Objectives • Rotor position detection of an IM using the PWM Harmonics, without Additional Hf Signal Injection • To define suitable position signals • Analysis and Decoupling of Corrupting harmonics • Rotor position reconstruction • Sensorless operation Slide: 43

44. PWM Carrier Harmonics • Typical PWM waveforms • Frequency spectrum of 1 PWM cycle • 2nd PWM harmonic shows highest amplitude • Can be regarded as ‘injection signal’ • Floating FFT spectrum Slide: 44

45. 2nd PWM Voltage Harmonic (fPWM2) • fPWM2 harmonic pulsates at 2fPWM and rotates at we • Amplitude and direction cannot be controlled without compromising the fundamental PWM scheme • Amplitude is variable, reflecting fundamental conditions Slide: 45

46. Definition of Position Signals • To overcome the variable hf excitation, define an equivalent impedance tensor zPWM2 as follows: • In IMs, |i| does not drop to zero due to magnetizing current • Demodulation scheme: Slide: 46

47. Test Rig • Off-the-shelf MEZ induction machine • Skewed rotor with semi open slots • Parameters: Slide: 47

48. Examining the position signals • Rotor geometry modulation quite visible • Additional modulation, apart from saturation saliency effect, depending on position of iPMW2 in ab frame. Assumed to be inverter nonlinearity effect. Slide: 48

49. Decoupling Saturation Saliency Modulation • Saturation saliency modulation depends on: • imposed stator currents is • relative position of iPWM2 to saturation saliency axis • Latter dependency makes compensation more challenging as relative position is speed dependent • SMP referenced by isq and Slide: 49

50. Equivalent impedance compensation as functions of isq and at a fixed stator current angle Decoupling Saturation Saliency Modulation • Additional dimension required to decouple inverter nonlinearity effects Slide: 50