Development of a Neuro Fuzzy Technique for Position Sensor Elimination in a SRM L. O. Henriques,...

Development of a Neuro Fuzzy Technique for Position Sensor Elimination in a SRM

L. O. Henriques, L.G. Rolim, W. I. Suemitsu,

P.J. Costa Branco

Presentation

• Introduction

– Operating modes

• System Training and Operation– Estimator types

• Results– Simulated and experimental

• Conclusions

Introduction

• Necessity for a perfect operation:– Correct excitation of the phases in synchronism with

the rotor position

• Reasons to remove the position sensor:– Cost, reliability, size, weight and inertia

• Basic voltage equation:

n

kkjjj dt

dRiv

1

Operating Modes in Sensorless Control

ba s e

5- Veryh igh speed

4- H ighspeed

3- Lowspeed

2- Very lowspeed

1- S till

Operating Modes in Sensorless Control

•Region 1 and 2: – Counter EMF is smaller than the DC bus voltage.

– Can operate in CR-PWM (hysteresis)

–Drawback: produces significant braking torque

–Voltage pulse injection to measure the inductance

base

5- Veryhigh speed

4- Highspeed

3- Lowspeed

2- Very lowspeed

1- Still

Operating Modes in Sensorless Control

•Region 3– Limitations (precision and accurancy) using diagnosis signal techniques

•Region 4– Operation is single pulse

– Region named “Constant power region”

•Region 5– Requires efficient real time algorithms

System Training and Operation

• Conventional estimator

dt

di

iiRv

dt

d...

1

System Training and Operation

• Initial proposed estimator

Voltage Profile Training Data

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 -150

-100

-50

0

50

100

150

Voltage (V)

-8

-6

-4

-2

0

2

4

Filtered Voltage

(V)

time (s) 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

System Training and Operation

• Final proposed estimator

sensor

+

-

Training phase

1/s

Operation phase

Training and Operation Phase

PIController

+

+

Converter+

Motor+

-

NeuroFuzzy

Compensator

NeuroFuzzy

Estimator

filter

Vphases

iref

iref

ipi

i

re f

Simulated Result

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

10

20

30

40

50

60

70

80

90

Time(s)

Po

siti

on

(deg

ree)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

10

20

30

40

50

60

70

Time(s)

spee

d(r

pm

)

estimated real

Experimental Results

0 500 1000 1500-5

0

5 Filtered Phase Voltage

phas

e 1

(vol

t)

0 500 1000 1500-5

0

5

0 500 1000 1500-5

0

5

Sample point #

phas

e 2

(vol

t)ph

ase

3(v

olt)

Experimental Result

0 500 1000 15000

0.5

1

1.5

2Reference current

Sample point #

Am

pere

(A

)

Experimental Result

0 100 200 300 400 50090

95

100

105

110Real Speed

Sample point #

(rpm

)

Experimental Result

Estimated Speed

0 100 200 300 400 50085

90

95

100

105

110

Sample point #

(rpm

)

Experimental Result

0 100 200 300 400 50085

90

95

100

105

110Real and Estimated Speed

Sample point #

(rpm

)

Conclusions

• A review of the torque-speed operating modes is shown

• A new technique of position sensor elimination is presented, based on intelligent techniques with neural nets and fuzzy logic

• Simulated and experimental results demonstrated the feasibility of the technique.

Basic Voltage Equation

n

k

kjdt

djRijv

1

dt

di.

idt

d.jRijv

dt

di.

idt

d.

dt

d)i,(f

dt

di.

ii.Rv.

dt

d 1