Sensorless Control of PMSM for
Electric Vehicle Application
Author
Mohamed Hassan Abou El Ella
Supervisors:
Prof. Dr. Osama Mahgoub
Prof. Dr. Abdelatif El Shafei
Ass. Prof. Dr. Sherif Zaid
Cairo University, Egypt
6/3/2014
1
Contents
Introduction
Sensorless Control Techniques
PMSM Mathematical Modeling
Control basics of PMSM
Field Weakening Control of PMSM
Introduction to MRAS based estimators
MRAS speed Estimator for PMSM
ANFIS based MRAS speed estimator for PMSM
Conclusion
6/3/2014
2
6/3/2014
3
Introduction
Introduction
Permanent Magnet Motors have received great attention in research in recent time due to:
High Power Density
High Efficiency
Maintenance free operation.
High reliability
High torque to inertia ratio
6/3/2014
4
Introduction Cont
Permanent Magnet Motor Types
6/3/2014
5
PMAC motor PMDC motor
Named as permanent magnet Synchronous motor PMSM.
Sinusoidal back EMF. Sinusoidal stator
current.
Named as brushless DC motor BLDC.
Trapezoidal back EMF. Square wave stator
currents.
Introduction Cont
PMSM Types
6/3/2014
6
Surface Mounted Interior Magnet
Magnets are mounted on the rotor Surface.
D, Q axis inductances are equal.
Zero reluctance torque
Magnets are buried inside the rotor
D, Q axis inductances are not equal.
Reluctance torque exists
Introduction Cont
PMSM Control Techniques Open Loop Control
Direct Torque Control
Vector Control
Sensorless Control
6/3/2014
7
6/3/2014
8
PMSM Mathematical Modeling
PMSM Mathematical Modeling
6/3/2014
9
Voltage Equations:
Where;
Cross section in the PMSM
PMSM Mathematical Modeling Cont
Voltage Equations in the frame
6/3/2014
10
Voltage Equation
Back EMF
Current Model
PMSM Mathematical Modeling Cont
Voltage Equations in the DQ frame
6/3/2014
11
Voltage Equation
Current Model
6/3/2014
12
Control Basic of PMSM
Control Basics of PMSM Control Techniques
6/3/2014
13
A. Open Loop Control:
Maximum torque is achieved by maintaining V/f ratio constant.
Valid for high speed values near to the base speed where the stator resistance voltage may be neglected.
The speed command should be from reference speed curve to avoid losing of synchronization.
6/3/2014
14
Control Basics of PMSM Control Techniques
June 3, 2014 14
B. Direct Torque Control: The basic idea is to control the torque and flux linkage
by selecting the voltage space vectors properly using a pre-defined switching table.
The selection of the voltage vector is based on hysteresis controller for both the stator flux linkage and Torque.
Torque is controlled by changing angle between stator and rotor flux when stator flux is kept constant.
IPM
SPM
6/3/2014
15
Control Basics of PMSM Control Techniques
June 3, 2014 15
B. Direct Torque Control: Block Diagram For the conventional DTC.
6/3/2014
16
Control Basics of PMSM Control Techniques
June 3, 2014 16
B. Direct Torque Control: The Voltage vector is always maintained within a defined
hysteresis band according to values of and .
6/3/2014
17
Control Basics of PMSM Control Techniques
June 3, 2014 17
C. Current Vector Control (FOC): The basic idea is to control the torque and flux linkage
independently by comparison of the motor currents and reference values in the rotor reference frame.
Reference values of the currents are obtained from Torque or Speed command.
Hysteresis or space vector based switching may be applied.
Torque is controlled by changing Iq while keeping Id constant.
IPM
SPM
6/3/2014
18
Control Basics of PMSM Control Techniques
C. Current Vector Control: Feed back required data:
Stator Current is measured and transformed to the rotating DQ reference frame using Parks transformation.
Decoupling between Id and Iq for independent control for torque and flux is achieved by Feed Forward compensation.
Rotor position is obtained by the aid of hall effect sensors, resolvers and optical shaft encoders.
6/3/2014
19
Control Basics of PMSM Control Techniques
C. Current Vector Control: Vector Control Block Diagram:
6/3/2014
20
Control Basics of PMSM Control Techniques
D- Sensorless Control:
Position data is always required in PMSM vector control.
Position Sensors used may be: Resolvers
Hall Effect Sensors.
Shaft Encoders
D- Sensorless Control: Elimination of Sensors is recommended for the
following: Reduction of the overall drive cost Increasing the system reliability Reduction of system complexity.
All the previous mentioned reasons caused great motivation for research in the Sensorless control of PMSM
6/3/2014
21
Control Basics of PMSM Control Techniques
Sensorless Control of PMSM Cont
Techniques Used in Sensorless Control:
1. Back EMF Estimation
Has great performance at high speed
Suffers at low speed and standstill due to low back EMF values.
Affected by the value of stator resistance.
6/3/2014
22
Sensorless Control of PMSM Cont
Techniques Used in Sensorless Control:
2. Saliency based methods
Depends on inductance variation due to saliency.
Shows great performance at low speed range and standstill.
Causes torque ripples at high speed ranges
6/3/2014
23
Sensorless Control of PMSM Cont
Techniques Used in Sensorless Control:
3. Extended Kalman Filters
Suffers from parameter sensitivity, complex computations
Requires initial conditions.
6/3/2014
24
Sensorless Control of PMSM Cont
Techniques Used in Sensorless Control:
4. Sliding Mode Observers
Great immunity against parameter variation.
Suffers from chattering problems
Requires great computational power.
6/3/2014
25
Sensorless Control of PMSM Cont
Techniques Used in Sensorless Control:
5. Model Reference Adaptive System (MRAS).
Great dynamic performance
High immunity against parameter variations.
We Will discuss the MRAS based Sensorless Control in this literature
6/3/2014
26
Control Basics of PMSM Regions Of Operation
6/3/2014
27
n
T
Constant Torque MTPA
Below base speed the operation is in the Constant torque With the voltage and power increasing as the speed increases.
Control Basics of PMSM Regions Of Operation
6/3/2014
28
n
nb
P T
Constant Torque MTPA
Constant Power
FW
Above the rated Speed Voltage is limited to its maximum value and the operation is in the constant Power speed range with flux weakening.
6/3/2014
29
Control Basics of PMSM Regions Of Operation
Phasor Diagram
Below Rated Speed
Phasor Diagram
Above Rated Speed
6/3/2014
30
Control Basics of PMSM Voltage constraint
All PMSM operation should be within the specified voltage and current limits.
The voltage constraint: From the voltage equations of the motors.
6/3/2014
31
Control Basics of PMSM Voltage constraint
Ellipse Equation Centre at (-m/Ld,0)
The voltage constraint: Referring the voltage constraint equation to the id and iq
axis
Id
Iq
-m/Ld
6/3/2014
32
Control Basics of PMSM Voltage constraint
The voltage constraint: For Surface Mounted Motors. Ld=Lq=Ls Voltage constraint will be equation of a circle.
Id
Iq
-m/Ls
6/3/2014
33
The current constraint: Equation of a circle with radius Ismax
Id
Iq
Ismax
Current Constraint
Voltage Constraint
Control Basics of PMSM Current constraint
6/3/2014
34
Control Basics of PMSM Voltage and Current constraint
PMSM are divided into two main types from the constraints point of View
Type I:
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4-10
-8
-6
-4
-2
0
2
4
6
8
10
iq(A
)
id (A)
Voltage limit circles
w1
w2
w3
w4
Voltage Limit Circle
current
Limit
Circle
Extending Speed is Limited by Current
Constraint
6/3/2014
35
Control Basics of PMSM Voltage and Current constraint
PMSM are divided into two main types from the constraints point of View
Type II:
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4-10
-8
-6
-4
-2
0
2
4
6
8
10
iq(A
)
id (A)
Voltage / Current limit circles
w1
w2
w3
Voltage Limit Circle
current
Limit
Circle
w4
w5
Speed Can be increased to infinity
6/3/2014
36
Control Basics of PMSM Control Regions
1. Maximum Torque Per Ampere (MTPA) Current Control:
Applied when speed
6/3/2014
37
Control Basics of PMSM Control Regions
1. Maximum Torque Per Ampere (MTPA) Current Control:
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4-10
-8
-6
-4
-2
0
2
4
6
8
10
voltage limit
circle
iq(A
)
id (A)
Voltage / current limit circles
current
limit
circle
MTPA trajectory
W1
6/3/2014
38
Control Basics of PMSM Control Regions
2. Flux Weakening Control: Applied at speed > b. Flux weakening can be achieved by applying negative armature
reaction by (Id ve). Voltage is limited to its rated value. Current vector should be maintained within its limit Relation between Id and Iq during FW control
6/3/2014
39
Control Basics of PMSM Control Regions
2. Flux Weakening Control: For Type I motor
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4-10
-8
-6
-4
-2
0
2
4
6
8
10
voltage limit
circle
iq(A
)
id (A)
Voltage / current limit circles
current
limit
circle
W1
W2
MTPA trajectoryFW trajectory
W3
W4
6/3/2014
40
Control Basics of PMSM Control Regions
2. Flux Weakening Control: For Type II motor
-5 -4 -3 -2 -1 0 1 2 3 4 5-5
-4
-3
-2
-1
0
1
2
3
4
5
voltage limit
circle
iq(A
)
id (A)
Voltage / current limit circles
Current
limit
CircleW4
W2
W3
MTPA
FW
LVMTW1
0
A
B
C
6/3/2014
41
Control Basics of PMSM Control Regions
2. Flux Weakening Control: Region 1: MTPA will be applied where Region 2: FW will be applied where Region 3 (For type2 motors only): , LVMT will be applied where
Motor Used in this Thesis a Type Motor
6/3/2014
42
Field Weakening Control Of PMSM
6/3/2014
43
Field Weakening Control
The main aim in FW control is selection of a proper Id value that meets the motor constraints Id reference Value is calculated from the torque demand of the speed loop.
+-
r*
PI +-
FW +-
PI
PI
Vq* V
V
Space
Vector
PWM
d-q
-
PMSM
Motor
d-q
abc
Iq
Iq*
Id* Vd*
Id
Vabc
Iabc
Position / Speed sensor
6/3/2014
44
Field Weakening Control
Reference Id is affected by the transient response in torque demand and the motor speed.
The proposed trajectory depends on setting a
predefined value depending on the desired speed
The optimum current trajectory at a certain speed is calculated from the voltage and current constraints
6/3/2014
45
Field Weakening Control
The Graph representing this trajectory
-5 -4 -3 -2 -1 0 1 2 3 4 5-5
-4
-3
-2
-1
0
1
2
3
4
5
iq(A
)
id (A)
Voltage / current limit circles
Voltage Limit
B
Te 3
Te 2
Te 1
A
Current Limit
6/3/2014
46
Field Weakening Control
The Offline Reference values of Id is
-1 0 1 2 3 4 5-5
-4
-3
-2
-1
0
1
2
iq(A)
id (
A)
Id reference values
id
1200 RPM
1100 RPM
1300 RPM
1400 RPM
1500 RPM
1600 RPM
1700 RPM
1800 RPM
6/3/2014
47
Field Weakening Control
A look up table will be used containing the reference Id value according to each speed demand.
The reference value represents the average of each curve The trajectory is described by the following figure
-5 -4 -3 -2 -1 0 1 2 3 4 5-5
-4
-3
-2
-1
0
1
2
3
4
5
iq(A
)
id (A)
Voltage / current limit circles
Voltage Limit
Te 3
Te 2
Te 1
Current Limit
A
A'
BB'
6/3/2014
48
Field Weakening Control
A smooth transition will take place after transient response to the optimum value of the current.
The transition will only take place if : Id(Optimum) < Id (LUT) This is done to increase the efficiency of the motor. For the case of: Id(optimum) > Id (LUT) No transition will take place which will only require a
small increase in the DC bus voltage The proposed trajectory will result in greater efficiency
of operation and better dynamic response
6/3/2014
49
Field Weakening Control
Graphical Representation of Proposed Trajectory
-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0
-3
-2
-1
0
1
2
3
iq(A
)
id (A)
Voltage / current limit circles
Current Limit
Voltage
Limit BB'
AA'
C C' Te
6/3/2014
50
Field Weakening Control
Graphical Representation of Proposed Trajectory
-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0
-3
-2
-1
0
1
2
3
iq(A
)
id (A)
Voltage / current limit circles
Current Limit
Voltage
Limit BB'
A'
C C' Te
A
1200 RPM
1500 RPM
6/3/2014
51
Field Weakening Control * < base
MTPA
Yes
No
Vs < Vsm
Yes
No
Id*(LUT)
|| *- || < error
Get Id** (iq*)
No
Yes
Id** (iq*) < Id*(LUT)
FW id**(iq*)
Yes
No
6/3/2014
52
Field Weakening Control Simulation Results
Case I : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 0.5 NM
0 5 10 15-5
-4
-3
-2
-1
0
1
2
time (Sec)
id V
s.
id*
(A
)
Actual Vs Desired Id
id
id*
6/3/2014
53
Field Weakening Control Simulation Results
Case I : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 0.5 NM
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
Id /
Iq
(A
)
Id and Iq
id
iq
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
sta
tor
Cu
rre
nts
(A
)
Ia and Ib
ia
6/3/2014
54
Field Weakening Control Simulation Results
Case I : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 0.5 NM
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
IS
(A)
Magnitude of current vector
IS
0 5 10 150
10
20
30
40
50
60
70
80
90
100
time (Sec)
Vs
(V)
Magnitude of Voltage vector
Vs
6/3/2014
55
Field Weakening Control Simulation Results
Case II : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 2 NM
0 5 10 150
500
1000
1500
2000
2500
time (Sec)
nsp
vs.
nfb
(R
PM
)
Actual Vs Desired Speed
Sp
Fb
0 5 10 15-5
-4
-3
-2
-1
0
1
2
time (Sec)
id V
s.
id*
(A
)
Actual Vs Desired Id
id
id*
6/3/2014
56
Field Weakening Control Simulation Results
Case II : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 2 NM
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
Id /
Iq
(A
)
Id and Iq
id
iq
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
sta
tor
Cu
rre
nts
(A
)
Ia and Ib
ia
6/3/2014
57
Field Weakening Control Simulation Results
Case II : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 2 NM
0 5 10 15-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
IS
(A)
Magnitude of current vector
IS
0 5 10 150
10
20
30
40
50
60
70
80
90
100
time (Sec)
Vs
(V)
Magnitude of Voltage vector
Vs
6/3/2014
58
Field Weakening Control Simulation Results
Case III : Speed Demand of 1200 , 1500 , 1800 RPM Tl = 0.5 NM and 2 NM Comparing two techniques
0 5 10 150
500
1000
1500
2000
2500
time (Sec)
nsp
vs.
nfb
(R
PM
)
Actual Vs Desired Speed
Sp
Fb
0 5 10 150
500
1000
1500
2000
2500
time (Sec)
nsp
vs.
nfb
(R
PM
)
Actual Vs Desired Speed
Desired
LUT
direct
6/3/2014
59
Field Weakening Control Experimental Results
Case I : Speed Demand of 500, 800, 1200 , 1500 RPM Tl = 2 NM
0 2 4 6 8 10 12 14 160
500
1000
1500
2000
2500
time (Sec)
nsp
vs.
nfb
(R
PM
)
Actual Vs Desired Speed
Sp
Fb
0 2 4 6 8 10 12 14 16-5
-4
-3
-2
-1
0
1
2
time (Sec)
id V
s.
id*
(A
)
Actual Vs Desired Id
id
id*
6/3/2014
60
Field Weakening Control Experimental Results
Case I : Speed Demand of 500, 800, 1200 , 1500 RPM Tl = 2 NM
7.5 8 8.5 9 9.5
-2.4
-2.2
-2
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
time (Sec)
id V
s.
id*
(A
)
Actual Vs Desired Id
id
id*
11.5 12 12.5 13 13.5
-3.4
-3.2
-3
-2.8
-2.6
-2.4
-2.2
-2
time (Sec)
id V
s.
id*
(A
)
Actual Vs Desired Id
id
id*
6/3/2014
61
Field Weakening Control Experimental Results
Case I : Speed Demand of 500, 800, 1200 , 1500 RPM Tl = 2 NM
0 2 4 6 8 10 12 14 16-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
Iq*
vs.
Iq
(A)
Actual vs. Desired Iq
iq
iq*
0 2 4 6 8 10 12 14 160
1
2
3
4
5
6
7
8
9
10
time (sec)
Is
(A)
Current Vector Is
Is
6/3/2014
62
Field Weakening Control Experimental Results
Case I : Comparing Two Techniques
8 9 10 11 12 13 14 15
-200
0
200
400
600
800
1000
1200
1400
1600
1800
time (Sec)
nsp
vs.
nfb
(R
PM
)
Actual Vs Desired Speed
Sp
FW with transition
FW without Transition
6/3/2014
63
MRAS Based Estimators
MRAS based Estimators MRAS is based on the comparison of two
models: The reference Model Adjustable Model Each Model Has its input
6/3/2014
64
Reference Model
Adjustable Model In 2
In 1
MRAS based Estimators
6/3/2014
65
Reference Model
Adjustable Model error
The error between the two model is processed to a certain adaptation mechanism
The output of the adaptation mechanism is used to tune the adjustable model.
In 2
In 1
Adaptation Mechanism
MRAS based Estimators Cont
The adaptation mechanism results in minimizing the error between both models
At minimum error both models will be the same
6/3/2014
66
Reference Model
Adjustable Model
error
In 2
In 1
Adaptation Mechanism
=
6/3/2014
67
MRAS based on Popov Theory
Speed Estimators
MRAS based Speed Estimator Block Diagram
6/3/2014
68
Reference Mode (PMSM Motor)
Adjustable Model (Current Model)
Vq
Vd
+ -
Adaption Mechanism
For Hybrid Learning
+ -
Id
ed
6/3/2014
69
MRAS based Speed Estimator Current Model
Current Model in
DQ Frame
where
6/3/2014
70
MRAS based Speed Estimator Current Model
Replacing Actual Values with the estimated values, the adjustable model used is
Adjustable Model
6/3/2014
71
MRAS based Speed Estimator Adaptation Mechanism
The generalized error equation is
where
6/3/2014
72
MRAS based Speed Estimator Adaptation Mechanism
According to Popov hyperstability Theory the following conditions should be satisfied:
Positive Definite
From these two conditions the speed can be calculated from the following Equation
Position
6/3/2014
73
MRAS based Speed Estimator Simulation Results
+-
r*
PI +-
FW +-
PI
PI
Vq* V
V
Space
Vector
PWM
d-q
-
PMSM
Motor
d-q
abc
Adjustable
Model
Vd
Vq
Iq
+-
Iq*
Id* Vd*
Id
Vabc
Iabc
r
r
id
iq
+-
r
6/3/2014
74
MRAS based Speed Estimator Simulation Results
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
100
200
300
400
500
600
700
800
900
1000
time (Sec)
ns
p v
s.
nfb
(R
PM
)
Desire, Actual, Estimated Speed (RPM)
Sp
Fb
Estimate
Case I: speed demand (100-800) RPM, Tl=(1-2-0.5 )NM , Nominal Parameters
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-20
-15
-10
-5
0
5
10
15
20
time (Sec)
err
or(
RP
M)
speed error
error
6/3/2014
75
MRAS based Speed Estimator Simulation Results
Case I: speed demand (100-800) RPM, Tl=(1-2-0.5 )NM , Nominal Parameters
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
1
2
3
4
5
6
7
8
Time (Sec)
Th
eta
m
(Ra
d)
Actual vs. Estimated position
Act
Est
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-5
-4
-3
-2
-1
0
1
2
3
4
5
time (Sec)
Id /
Iq
(A
)
Id and Iq
id
iq
6/3/2014
76
MRAS based Speed Estimator Simulation Results
Case II: speed demand (300-800-1200-1500-1800) RPM, Tl=(2 )NM , Nominal
Parameters
0 1 2 3 4 5 6 7 8 9 100
200
400
600
800
1000
1200
1400
1600
1800
2000
time (Sec)
nsp
vs.
nfb
(R
PM
)
Desired, Actual, Estimated Speed (RPM)
Sp
Fb
Estimate
0 1 2 3 4 5 6 7 8 9 10-20
-15
-10
-5
0
5
10
15
20
time (Sec)
err
or(
RP
M)
speed error
error
6/3/2014
77
MRAS based Speed Estimator Simulation Results
Case II: speed demand (300-800-1200-1500-1800) RPM, Tl=(2 )NM , Nominal
Parameters
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
Time (Sec)
Th
etam
(R
ad)
Actual vs. Estimated position
Act
Est
6/3/2014
78
MRAS based Speed Estimator Simulation Results
Case III: speed demand (300-800-1200) RPM, Tl=(2 )NM , 30% increase inductance
0 1 2 3 4 5 60
200
400
600
800
1000
1200
1400
1600
1800
2000
time (Sec)
nsp
vs.
nfb
(R
PM
)
Desired, Actual, Estimated Speed (RPM)
Sp
Fb
Estimate
0 1 2 3 4 5 6-20
-15
-10
-5
0
5
10
15
20
time (Sec)
err
or(
RP
M)
speed error
error
6/3/2014
79
MRAS based Speed Estimator Simulation Results
Case III: speed demand (300-800-1200) RPM, Tl=(2 )NM , 30% increase inductance
0 1 2 3 4 5 60
1
2
3
4
5
6
7
8
Time (Sec)
Th
eta
m (R
ad
)
Actual vs. Estimated position
Act
Est
6/3/2014
80
MRAS based Speed Estimator Experimental Results
Case I: speed demand (500-1200-1500) RPM, Tl=(2 )NM, Nominal parameters
1 2 3 4 5 6 7 8 9 10 11 120
200
400
600
800
1000
1200
1400
1600
1800
2000
time (Sec)
Sp
ee
d
(RP
M)
Actual Vs. Estimated Speed
Act speed
Estimated Speed
1 2 3 4 5 6 7 8 9 10 11 12-200
-150
-100
-50
0
50
100
150
200
time (Sec)
Sp
ee
d E
rro
r (
RP
M)
Speed Error
Speed Error
6/3/2014
81
ANFIS based MRAS Speed Estimators
ANFIS based MRAS Speed Estimator
6/3/2014
82
Reference Mode (PMSM Motor)
Adjustable Model (Current Model)
Vq
Vd
+ -
ANFIS Adaption
Mechanism
For Hybrid Learning
6/3/2014
83
ANFIS Architecture
ANFIS Introduction
ANFIS is Adaptive Neuro Fuzzy Inference System Advantages: Combines advantages of both
1. Artificial Neural Network (ANN) Learning ability
2. Fuzzy Logic Controllers Linguistic representation
Can handle the non linearities of a system such as the PMSM
Have high convergence rate due to hybrid learning algorithm
6/3/2014
84
ANFIS Architecture
ANFIS consists of 5 network layer with two inputs DQ currents and one Output, electrical angular speed.
Layer 1 (Fuzzification layer):
Each node is an adaptive node.
Node is square represented.
Three membership functions are assigned to each input.
6/3/2014
85
ANFIS Architecture Cont
Layer 1 (Fuzzification layer):
The mathematical formula for each node is
6/3/2014
86
Member Ship Functions
ANFIS Architecture Cont
Layer 1 (Fuzzification layer):
Parameters of this layer are called Premise parameters .
Premise parameters will be tuned through back propagation technique
6/3/2014
87
ANFIS Architecture Cont
6/3/2014
88
Layer 1
ANFIS Architecture Cont
Layer 2 ( Multiplication layer):
Multiplies the incoming signals.
Each node is circle represented symbolized with .
Mathematical formula
6/3/2014
89
ANFIS Architecture Cont
6/3/2014
90
Layer 1 Layer 2
ANFIS Architecture Cont
Layer 3 ( Normalization layer):
Calculates the normalized firing strength of each rule
6/3/2014
91
ANFIS Architecture Cont
6/3/2014
92
Layer 1
Layer 2
Layer 3
ANFIS Architecture Cont
Layer 4 Consists of adaptive nodes Multiplies the normalized firing strength with
the Input Parameters are known as consequent
parameters. Consequent parameters will be tuned by
Recursive Least Square Estimate(RLS)
6/3/2014
93
ANFIS Architecture Cont
6/3/2014
94
Layer 1
Layer 2
Layer 3
Layer 4
ANFIS Architecture Cont
Layer 5 ( output layer)
Calculates the overall output of the system.
The output r is used to tune the adjustable model.
The mathematical formula is
6/3/2014
95
ANFIS Architecture Cont
6/3/2014
96
Layer 1
Layer 2
Layer 3
Layer 4
Layer 5
6/3/2014
97
ANFIS Training Algorithm
ANFIS Training Algorithm
6/3/2014
98
Hybrid Learning Algorithm
Premise Parameter
Back Propagation
Consequent Parameter
RLS
ANFIS Training Algorithm Cont
6/3/2014
99
Back Propagation Method
The error used will be the quadrature current error
The updating formula will be
ANFIS Training Algorithm Cont
6/3/2014
100
RLS Method
The updating formula will be
6/3/2014
101
Simulation Results
Simulation Results
Simulation is carried out by Matlab/Simulink
A simple FOC is used to control the speed of the PMSM
The drive system used will be
6/3/2014
102
Simulation Results Cont
Case 1: Speed variation at No Load and nominal parameters
6/3/2014
103
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-500
0
500
1000
1500
2000
time (Sec)
Est
imate
d v
s. A
ctu
al
speed
(R
PM
)
Act
Est
Simulation Results Cont
Case 1: Speed variation at No Load at nominal parameters
6/3/2014
104
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
1
2
3
4
5
6
7
time (Sec)
Est
imate
d v
s. A
ctu
al
po
siti
on
Rad
Act
Est
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-200
0
200
400
600
800
1000
time (Sec)
Sp
eed
Err
or
(RP
M)
speed error
Simulation Results Cont
Case 2: Speed variation at with load and nominal parameters
6/3/2014
105
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-200
0
200
400
600
800
1000
1200
1400
1600
1800
time (Sec)
Est
imate
d v
s. A
ctu
al
speed
(R
PM
)
Act
Est
Simulation Results Cont
Case 2: Speed variation at with load and nominal parameters
6/3/2014
106
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
1
2
3
4
5
6
7
time (Sec)
Est
imate
d v
s. A
ctu
al
po
siti
on
Rad
Act
Est
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-200
0
200
400
600
800
1000
time (Sec)
Sp
eed
Err
or
(RP
M)
speed error
Simulation Results Cont
Case 3: Speed variation with 20% resistance increase , with and without torque.
6/3/2014
107
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
200
400
600
800
1000
1200
1400
1600
1800
2000
time (Sec)
Est
imate
d v
s. A
ctu
al
speed
(R
PM
)
Act
Est
Simulation Results Cont
Case 4: Speed variation with 20% resistance and inductance increase , with constant torque.
6/3/2014
108
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-500
0
500
1000
1500
2000
time (Sec)
Est
imate
d v
s. A
ctu
al
speed
(R
PM
)
Act
Est
Simulation Results Cont
Case 4: Speed variation with 20% resistance and inductance increase , with constant torque.
6/3/2014
109
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
1
2
3
4
5
6
7
time (Sec)
Est
imate
d v
s. A
ctu
al
po
siti
on
Rad
Act
Est
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-200
0
200
400
600
800
1000
time (Sec)
Sp
eed
Err
or
(RP
M)
speed error
6/3/2014
110
Thank You
Questions