University of Nottingham

School of Electrical and Electronic Engineering

Sensorless Control of AC Motor Drives at low and zero frequency

Professor Greg Asher

Power Electronics, Machines and Control Group

OverviewOverview

General principles

Problems of sensorless control at low speed

How to get best performance from standard approaches

Principle of sensorless control by signal injection

Practical problems

Illustrative results for SM Permanent Magnet Machine

Illustrative results for standard cage Induction Machine

Review and Conclusions

General Principles Vector Control

•• Flux angle tracking Flux angle tracking –– SM Permanent Magnet MachineSM Permanent Magnet Machine

isd = 0

isq = stator current controlling motor torqueisq ΨR

λr

Knowledge of λ r at any time means:

Can control stator currents (mag and phase) so that stator current distribution gives correct iS for required isq

Now λ r = rotor position θr

General Principles Vector Control

•• Flux angle tracking Flux angle tracking –– Induction MotorInduction Motor

Knowledge of λ r at any time means:

Can control stator currents (mag and phase) so that stator current distribution gives correct iS for required isq and isd

But must know λ r at all times as it rotates in space

ΨR

iS

λr

isq

isd

isd = stator current controlling motor field

isq = stator current controlling motor torque

Need for sensorless controlNeed for sensorless control

isd* , isq*

Is

PWM

Coordinate

Trans

Coordinate

TransPI

θr (PM)

isd , isq

λr

Speed

Position

control

ωr (IM or PM)

Flux angle

calculator

Is

VsMachine

model

• Expensive on small drives• Fragile• Mounting problems

• Power supply required• Isolation• Noise and glitches

Flux angle

calculator

ModelModel--based Sensorless Drivebased Sensorless Drive

PWM

Machine

model

Coordinate

Trans

Coordinate

TransPI

θr (PM)

Isisd , isq

isd* , isq*

Is

Vs

λr

• λr incorrect if model parameters are wrong– lose flux and torque control

• Fails asymptotically if ωe→ 0

Sensorless performance of 4-pole 5kW IM under full load(Adaptive Luenburger Observer)

- 2 0 0

- 1 5 0

- 1 0 0

- 5 0

0

5 0

0 5 0 1 0 0L o a d to r q u e [ % ]

Nrre

f [rpm

]

U n s t a b l e r e g io n

• ωe = 0 (theoretical failure) occurs in re-generative region - red line

• ωe→ 0 region of instability (grey region) for a given scheme

• ωe→ 0 region of instability widens if parameters in model are in error

Sensorless performance of 5kW IM under full load(Adaptive Luenburger Observer)

-300

0

300

600

900

1200

1500

0 4 8 12 16 20 24 28 32 36 40Time [s]

Nrre

f , Nr [

rpm

]

Holds full load at zero speed for 30s followed by accelaration to 1200rpm

-1500

-1200

-900

-600

-300

0

300

600

900

1200

1500

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Time [s]

Nrre

f , Nr [

rpm

]1200 to –1200 rpm Quickly through zero

-1500

-1200

-900

-600

-300

0

300

600

900

1200

1500

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Time [s]

Nrre

f , Nr [

rpm

]1200 to –1200 rpm Slowly through zero

Going to zero in 30rpm stepsRs

est = 0.8Rs and 50 % load torque

-60

-30

0

30

60

90

120

150

180

0 1 2 3 4 5 6 7 8 9 10Time [s]

Nrre

f , Nr

[rpm

]

How to get best performance from model-based PWM drive

1. When integrating to find flux e.g.

- Do NOT use DC blocking filter to eliminate integrator drift

- Use non-linear feedback integrator exploiting fact that flux is constant

( )∫ −= ssss iRVΨ

2. Use linear (!) power amplifier to get sinusoidal V, I as ωe→ 0

- best possible dead-time compensation (ie. hardware sgn(i) detection)

- must compensate for IGBT & diode voltage drops

- avoid high switching frequencies (VTA distortion)

3. Exploit best Rs identifier you can find

4. Avoid heavily saturated machines and under flux slightly

**** BUT IT WILL STILL FAIL AS ωe→ 0 ****

How to get best performance frommodel-based PWM drive - 2 0 0

- 1 5 0

- 1 0 0

- 5 0

0

5 0

0 5 0 1 0 0L o a d t o r q u e [ % ]

Nrre

f [rpm

]

U n s t a b l e r e g i o n

Step reduction of torque to zero speed – motoring quadrant

From J. Holtz, “Drift and parameter compensated flux estimator for persistent zero frequency operation of sensorless induction motors”, IAS 2002

University of Nottingham

School of Electrical and Electronic Engineering

Sensorless Control of AC Motors Drives at low and zero frequency

Using Signal Injection

Principle of high frequency injectionPrinciple of high frequency injection

Saturation

No saturation

• Apply high frequency signal e.g. 20V at 1kHz• Impedance seen by hf varies around machine due to

saturation effects• Measured hf currents will be amplitude modulated and will

contain information of d-axis (ψr) position

A axis

Fit hall probe in air gap aligned to Phase A coil

Principle of high frequency injectionPrinciple of high frequency injection

A or α axis

pα

• Demodulate to get pα

β axis

pβ• Apply quadrature hf to β coilDemodulate to get pβ

• Have two “resolver” signals

ResolverResolver signals psignals pαβαβ in practice in practice -- SM PM machineSM PM machine

Harmonics exist on resolver signals

0 4 8 12 16 20 24 28 32 36 40 44 480

0.2

0.4

0.6

0.8

1Before Compensation

Frequency [Hz]

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

0 40

0.2

0.4

0.6

0.8

1

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

desired undesired

• Non-sinusoidal distribution of saturation

• Inverter effects – dead time & device voltage drop

• Dirty resolver signal must be cleaned up

Harmonic Compensation using machine signatureHarmonic Compensation using machine signatureof PM Machine of PM Machine

0 4 8 12 16 20 24 28 32 36 40 44 480

0.2

0.4

0.6

0.8

1Before Compensation

Frequency [Hz]

Amplit

ude [%

of Sal

iency

Positio

n Harm

onic]

0 4 8 12 16 20 24 28 32 36 40 44 480

0.2

0.4

0.6

0.8

1After Compensation

Frequency [Hz]

Amplit

ude [%

of Sal

iency

Positio

n Harm

onic]

Sensorless rotor position control structure for SMSensorless rotor position control structure for SM--PM machine using PM machine using αβ injection

Is

Is

isd , isq

isd* , isq*

PWM

Coordinate

Trans

Coordinate

TransPI

λr

Speed

Position

control

BPFDemodulationSignal

Cleaningatan

λr*

Encoder for monitoring

vαβ_hf

5kW Surface mount PM machine5kW Surface mount PM machineSensorless Position Control Sensorless Position Control –– 0% load0% load

• Response to 180° position demand

5kW Surface mount PM machine5kW Surface mount PM machineSensorless Position Control Sensorless Position Control –– 100% load100% load

• Response to 180° position demand- no integrator in control loop (incremental position only)- isq (torque current) limited to 1.3 x rated

PM Machine PM Machine Hybrid model / hf for wide speed range operationHybrid model / hf for wide speed range operation

injru Ψ= ˆmodˆ ru Ψ=

i αβv *

αβ

Rotor Flux model

modˆ rΨ

injrΨ̂

injrλ̂

rλ̂tan-1

12% 20%

)ˆsin(

)ˆcos(ˆ

injr

injrm

j λ

λ

+

Ψ

rω̂ to controller processingfrom hf

processing

• Under 12% speed use angle from hf processing

• Over 20% speed use angle from flux model

• Fuzzy interpolation between these these values

• Injection turned off e.g. 22% speed

PM machine Sensorless Hybrid control Hybrid control 1500 to 1500 to ––1500 rpm under 100% load1500 rpm under 100% load

0.6 0.8 1 1.2 1.4 1.6 1.8-2000

-1000

0

1000

2000

Rot

or sp

eed

[rpm

]

0.6 0.8 1 1.2 1.4 1.6 1.8-2000

-1000

0

1000

2000

Time [s]

Estim

ated

spee

d [r

pm]

Time [s]0.9 1 1.1 1.2 1.3 1.40

90

180

270

360

Inje

ctio

n θ r

[oel

ec]

0.9 1 1.1 1.2 1.3 1.40

90

180

270

360

θ rand

hyb

rid θ

r[o

elec

]^

^

Top: real speed

Bottom: estimated speed

Top: final rotor position

Bottom: hf estimate injrλ̂

rλ̂

PM machine Sensorless Hybrid controlHybrid controlStep position through 16 revolutions under 100% loadStep position through 16 revolutions under 100% load

0 1 2 3 4 5 6 7 8 9 10

0

1080

2160

3240

4320

5400

Rot

or p

ositi

on [

oel

ec]

0 1 2 3 4 5 6 7 8 9 10

0

1080

2160

3240

4320

5400

Time [s]

Est.

roto

r pos

ition

[o

elec

]

Real and estimated speed

0 1 2 3 4 5 6 7 8 9 10-2000

-1000-600

6001000

2000

Rot

or sp

eed

[rpm

]

0 1 2 3 4 5 6 7 8 9 10-2000

-1000-600

6001000

2000

Estim

ated

spee

d [r

pm]

Time [s]

Real and estimated position

• Below 400rpm – position estimated exclusively from hf injection

• Above 600rpm - position estimated exclusively from machine model

5kW Surface mount PM machine5kW Surface mount PM machineSensorless Position Control Sensorless Position Control –– 100% load100% load

1°

1°

1°

Position holding ((θθr r (ref)(ref) -- θθr r ))

at three different demand positions 45° apart

(illustrates cogging effect)

Actual Actual ResolverResolver signals psignals pαβCage Induction Cage Induction MachineMachine αβ

• Clean up to leave slotting harmonic to give rotor position tracking

• Better to have un-skewed, open slot, exhibiting good slot harmonic

Case 1 – closed slot, skewed machine with low rotor slotting

• Same effects as PM machine but higher undesired harmonics

• Clean up to get fundamental to give flux tracking

0 4 8 12 16 20 24 28 32 36 40 44 480

0.2

0.4

0.6

0.8

1Before Compensation

Frequency [Hz]

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

0 40

0.2

0.4

0.6

0.8

1

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

desired undesired

2fe

0 4 8 12 16 20 24 28 32 36 40 44 480

0.2

0.4

0.6

0.8

1Before Compensation

Frequency [Hz]

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

0 40

0.2

0.4

0.6

0.8

1

Amplit

ude [

% of

Salien

cy Po

sition

Harm

onic]

undesired desired

Case 2 – closed skewed machine with high rotor slotting

nfr2fe

Test signal injection for IM machinesTest signal injection for IM machines

Case 1 – closed slot, skewed machine with low rotor slotting

• Only flux position obtained

• Good for torque and flux control

• Need auxiliary structure for speed control e.g. model

Case 2 – open slot, exhibiting rotor slotting effects

• Rotor position obtained

• Good for torque, flux, speed and position control

Control mode & performance is machine dependent

Inverter and (selected machine) sold together

Technology most suitable for INTEGRATED DRIVES

Real Integrated Induction Motor DriveReal Integrated Induction Motor Drive

+ =

Integrated Motor Drive(Power Electronics housed in a redesigned End Plate)

Matrix ConverterInduction Motor

Power Electronics housed in the motor end plate

IGBTs, diodes and filter capacitors

Redesigned end plate with extra fins to cool the devices

Using the zeroUsing the zero--sequence current in an sequence current in an Integrated IM Drive Integrated IM Drive

Given an integrated drive

• Machine terminal connections at inverter

• Have access to machine phase currents in ∆ machine so can measure the zero sequence current, Iz

• Gives best method of tracking rotor or rotor flux position

• Applying test vector combinations to PWM, can show that λr = f(dIz/dt)

• dIz/dt measured using single non-integrating Rogowski coil

• Diagram shows torque control (rotor flux tracking) since have closed-slot machine

Machine terminals at inverter of integrated drive

Induction Machine Zero sequence currentsInduction Machine Zero sequence currentsSensorless Torque Control Sensorless Torque Control ωωrr= = ±±30 rpm at 30 rpm at 100% torque100% torque

Sensorless IM dynamometerDrive machine

ω = -30rpm

ω = 30rpm T = 100%

IA

Drive speed

IM flux angle

IA ∝ Driving torque

Induction Machine Zero sequence currentsInduction Machine Zero sequence currentsSensorless Torque Control Sensorless Torque Control 100% torque at 100% torque at ωωrr=0=0

Drive speed

IM flux angle

IM torque demand

IA ∝ Driving torque

IA

Sensorless IM dynamometerDrive machine

T = -100%ω = zero

T = 100%

Induction Machine Zero sequence currentsInduction Machine Zero sequence currentsSensorless Torque Control Sensorless Torque Control 100% torque at zero frequency100% torque at zero frequency

ω = -40rpm

Drive speed

IM flux angle

IA ∝ Driving torque

IA

T = 100%

Sensorless IM dynamometerDrive machine

Observer based Sensorless Observer based Sensorless DrivesConclusions DrivesConclusions

• Asymptotic failure as fe→ 0

• For best performance should focus on:

- obtaining linear behaviour for switching converter

- good integrator and resistance estimators

• Estimator/observer type not so important

• Expect failure as fe→ 0 under regeneration

• Performance will depend on machine to a certain extent

- sensorless drives assume validity of 2-axis dq theory

Signal injection based Signal injection based Sensorless PM DrivesConclusions Sensorless PM DrivesConclusions

• Position tracking effective down to and at fe→ 0 under all loads• No dependence on machine parameters

• Works for Surface mount PM machines (saturation saliency)

- very effective for buried magnet machines (salient machine)

• Hybrid technique appropriate for wide-speed range drive

• Harmonic compensation (signal cleaning) advisable- position accuracy down to 0.1° mechanical- sensorless closed loop position bandwidth 3 - 5Hz

• Improvements in performance if:- machine designed to minimise cogging- obtain linear behaviour for switching converter

Signal injection based Signal injection based Sensorless IM DrivesConclusions Sensorless IM DrivesConclusions

• Rotor Flux/ Rotor Position tracking effective down to and at fe→ 0 under all loads

• No dependence on machine parameters

• Control mode heavily machine dependent

- skewed, closed slot machines yield torque and flux control

- open-slot machines of given slot combination for rotor position control

• Commercial penetration likely to be as integrated drive

• Hybrid technique appropriate for wide-speed range drive

• Harmonic compensation (signal cleaning) essential