investigation of saturation impact on an ipmsm saliency

June, 2016 I Property of Valeo. Duplicati

June, 2016

Property of Valeo. Duplication prohibited

Confidential

Investigation of saturation impact on an IPMSM saliency-based sensorless control for automotive

applications

Wided ZINE – PhD Researcher (VALEO-SATIE)

Eric Monmasson, SATIE Lahoucine Idkhajine, SATIE Antoine Bruyere, VALEO Bruno Condamin, VALEO Pierre-Alexandre Chauvenet, VALEO

Journée GDR-SEEDS


Outline

2

3.1. Magnetic saturation impact on differential IPMSM saliency

3.2. Experimental evidence of IPMSM saliency behaviour

1. HF signal injection-based sensorless control: Objectives & Challenges

2. Description of the HF injection-based sensorless control system

4. Investigation of cross-coupling effect on estimation accuracy

3. Investigation of iron saturation effects on sensorless feasibility

4.1. Evaluation of cross-saturation impact on estimation accuracy

4.2. Experimental evidence of cross-coupling effect

5. Conclusion


June, 2016


Confidential

HF signal injection-based sensorless control: Objectives & Challenges


Context : Dedicated application & challenges

PMSM Sensorless control applied to the EV traction context is aimed at :

Provide the position/speed information necessary for the Torque vector control in the traction chain

Establish a digital information redundancy with optimized efficiency/cost/volume compromise

Ensure system & people safety : guarantee the service continuity when a fault on the position sensor appears

4

The main challenge is to cover the complete Torque/Speed operating area.


Sensorless techniques over the zero and low speed area

5

Sensorless techniques for a Permanent Magnet

Synchronous Machine

High speeds Low speeds

& standstill

Anisotropy/saliency-based

techniques

Pulses injection

INFORM method

HF signal injection method


June, 2016


Confidential

Description of the HF injection-based sensorless control system


Structure of the sensorless control system

1 • Torque

vector control layer

(ACL)

2

• Sensorless algorithm

layer (SAL)

7


HF signal injection algorithm optimization

HF Pulsating signal injection

HF signal injection

module Signal processing module P.I Observer module

Iαβ Δθmin

1 2( ) ( , ,........., )nMin f x x x

0

In order to indentify the minimization function elements, the sub-systems need to be

detailed

11 1 12 2 1 1

21 1 22 2 2 2

1 1 2 2

......... ( , , )

......... ( , , )

.

.

......... ( , , )

n n

n n

m m mn n m

a x a x a x b

a x a x a x b

a x a x a x b

where

[x1, x2, ....xn] : Control variables

aij, bi : parameters TBD

8


2.1. High frequency signal injection module

In this module, key control variables are fh anf Vh, the frequency and the amplitude of the HF signal.

fh et Vh [x1, x2, ....xn] ϵ

10

pwm

h

ff h ef f

h cemf f /

2

h h s

sh

f jL R

Rf

jL

For x1=fh

10pwmf KHz0 120eHz f Hz

TBD It is comprised between 5 to 10

Hz for studied machines

9


h rippleV V

For x2=Vh

ON OFFh DT dc

s

DT T TV V V

T

2.1. High frequency signal injection module

- For a frequency comprised between 500 et 1000 Hz, the audible noise

starts from some dBs, and the noise discrimination zone starts from 80

dB.

In this case, Vh is tuned from 5 to 30 V and the corresponding gain does

not exceed 35 dB. For this criterion, the induced noise is then

considered acceptable.

- On another hand, Vh should be tuned based on the induced torque

ripple => For example

Torque ripple < 20% Torque reference

VDT = 1 to 2V foe a Dead-Time DT =2μs.

11


2.4. Expression of the minimization function

Δθ= Min (fh , Vh, fBP, flp, fbp)

Constraints: b=[fpwm, fe, fCEM, Vripple, VDT, fh, 2fh, γmax]

12


IPMSMs are good candidates for the VE traction chain Field Oriented Control (FOC)

PMSM saturation impact on sensorless control perfomances

(+) High efficiency and power density

(+) Reduced cost and size

(+) Saliency ratio

(-) Magnetic saturation

The magnetic saturation emerges as a twofold phenomenon :

Simple iron saturation

Cross-saturation (cross-coupling)

13


June, 2016


Confidential

Investigation of iron saturation effects on sensorless feasibility


Magnetic saturation impact on differential IPMSM saliency

- Obviously, d-axis inductance hardly varies as function of d-axis and q-axis currents (20%)

whereas q-axis inductance considerably varies mainly as function of q-axis current (60%).

- Saliency ratio varies by 42% when Torque target varies from 0 N.m to 100 N.m

Hence, differential inductance (Lq-Ld) is load-dependent : it decreases with the applied

torque : It varies by 42% when Torque target varies from 0 N.m to 100 N.m

Sensorless feasibility is affected

14


Experimental evidence of IPMSM saliency behaviour

The power stage is mainly composed of four different elements:

Electrical source : Electrical grid/ Battery

Machine : PMSM1

Converters : 6-leg Inverter

Load : Asynchronous load machine

Maximum

power [KW]

Maximum

torque

[N.m]

Maximum

speed

[tr/min]

Ld (pu) Lq (pu) Rs (Ω)

Pole

pairs

(pp)

PMSM-1 180 250 15000 0.15 0.74 0.043 4

Parameters of the test machine

15

6-leg inverter


Investigation of (d,q) HF current loci (Rotating carrier)

16

Rotating carrier

injection

V αβ *

Vαβh

Idq*

Idqf

Iαβ

Iαβh

PMSM


Torque = 50 N.m Torque = 75 N.m

Torque = 100 N.m Torque = 200 N.m

PMSM HF output current : FFT analysis

Position

observer input

Iαβh

17

Iαβ Iαβ

Iαβ Iαβ


Sensorless feasible region in (Id, Iq) plane

The sensorless feasible region can be defined as function of saturation effects limitation => Bounded by Ldif=0

18


June, 2016


Confidential

Investigation of cross-coupling effect on estimation accuracy


Evaluation of cross-saturation impact on estimation accuracy

q

qd

d

KI

? ddq

q

KI

?

, ( ) ( , )

, ( ) ( , )

( ) .

( ) .

d d q

q d q

d d d q d I d m dq I I

q q d q q I q qd I I

f I I L I

f I I L I

, ( , ) ( , )

, ( , ) ( , )

( ) .

( ) .

d q d q

d q d q

d d d q d I I d m dq I I

q q d q q I I q qd I I

f I I L I

f I I L I

Generalized flux model

Suggested flux model

, ( )

, ( )

( ) . .

( ) . .

d

q

d d d q d I d m dq q

q q d q q I q qd d

f I I L I K I

f I I L I K I

20


Compensation procedure

Off-line algorithm

,

,

( ) ( )

( ) ( )

dq d d q d d

qd q d q q q

I I I

I I I

dq

dq

q

kI

PS: For simplifying reasons, Kdq is considered to be equal to Kqd

qd

qd

d

kI

3- Plot a LUT for kdq and induced θsat values

2sat

dq

d q

Arctgk

L L

1-

2-

Available Data

,

,

( )

( )

d d d q

q q d q

I I

I I

( )

( )

d d d

q q q

I

I

and

, ( )

, ( )

( ) . .

( ) . .

d

q

d d d q d I d m dq q

q q d q q I q qd d

f I I L I K I

f I I L I K I

21


Solution : Refer to the original flux linkage cartography as

function of currents (Model (2))

,

,

( )

( )

d d d q

q q d q

I I

I I

Model (2)

Flux linkages measurement (VALEO)

A and B are two points in (d,q) plane :

, *

,

( )

( )

d q

A B

d q

A I II I

B I I

Principle:

1- Impose Id and Iq current references

(A(Id,Iq)) and measure the corresponding

Vd and Vq (VdA and VqA)

.

.

dA s d q

qA s q d

V R I

V R I

2- Impose Id and -Iq current references

(B(Id,-Iq)) and measure the corresponding

Vd and Vq (VdB and VqB)

.

.

dB s d q

qB s q d

V R I

V R I

1( ) .

2

1( ) .

2

dA dB q

qA qB d

V V

V V


Cross-coupling terms : Predicted position error

Kdq cartography as function of (d,q) currents θsat cartography as function of (d,q) currents

- Cross-coupling term Kdq is varying mainly as function of Iq which confirms the

experimental measurements previously shown.

- Resulting calculated position offset θsat also corresponds to the one previously measured on

bench

23


Position error obtained by experiments

Estimated versus real position.(a)T=20N.m,(b).T=50N.m,(c).T=100N.m (d).T=150N.m

Operating speed : 300 rpm

24


June, 2016


Confidential

Conclusion


This paper evaluates the performance and the feasiblity of saliency-based sensorless control implemented algorithms along the specified operating range for an automotive application.

It has been highlighted that :

Iron saturation strongly affects the observer input signal quality. => Sensorless feasibility may be affected in saturated operating mode.

The estimated position is impacted by a load-dependent phase delay

Thus, the challenge is first to preserve a proper signal to-noise ratio along

the operating area, and then to correctly compensate the angular

displacement of the position estimate while implementing the control

algorithm.

26

investigation of saturation impact on an ipmsm saliency

Documents