CHAPTER ONE

INTRODUCTION

1.1 Background Information

Presently, the research trend in the electric motor industry follows the direction where-by new

products have to struggle to remain in competition by addressing several design constraints

which include Power consumption reduction, cost reduction, power factor correction and

reduced radiations. These mentioned challenges can be addressed through the development of

advance control algorithms (Bhardwaj M, 2013). The use of high performance electric motor

drives is very essential for the running of modern industrial applications. Direct current motors

were widely used in the past; however weight, size, frequent maintenance and undesired

sparking due to the Commutation made them to be expensive (Zaky M.S, 2007). Attempts were

then made to use Induction motors instead.

The Induction motor is an alternating current motor in which the electromagnetic energy is

transferred by inductive coupling from the stator to either a wound rotor or a squirrel cage rotor,

with the two windings being separated by an airgap (Bird J, 2017). The three phase induction

motors are the frequently encountered in the industry because they are simple, rugged, low

priced and easy to maintain (Wildi T, 2014). They can also be manufactured with characteristics

to suit most industrial requirements (Jaber A I, 2015). The motors find use in both industrial and

commercial sectors ranging from applications of electric vehicles, pumps, conveyors, fans and

home appliances (Hussein M, 2014). However, it also have disadvantages, one of which is the

necessity of a sophisticated control which is much complex compared to other type of motors

1
(Karlovsky P, 2017). The reasons are due to the high nonlinear properties that result from the

interaction of Stator and Rotor fields (Ali N H, 2017).

Choosing the type of electric motor type is the first step in the realization of a robust and

efficient electric drive, after that controlling it with the best performance is really an important

objective. In this research, the three phase induction motor is controlled through field orientation

(Vector control) technique. Field oriented control is simply based on projections which transform

a three phase time variant system to a two coordinate time invariant system, thereby allowing for

independent control of flux and torque like in a separately excited dc motor. and the control

requires knowledge of the rotor speed for the decoupling to occur (Surya P, 2019). This can be

obtained either by measurement using a speed sensor mounted on the shaft, or it can be estimated

using state equations and machine parameters (Manohar M, 2015). The speed sensor happened to

be undesirable because of the increase in cost and reliability problems, besides there is also need

for shaft extension (Bose BK, 2006). Sensor-less electric drives are therefore more suitable with

control algorithms especially in uncertain systems where the dynamics and parameters of the

plant changes during operation (Vinh VQ, 2019). Numerous methods of estimating speed include

Rotor slot harmonics, Kalman filter, Luenbergern observer, Saliency techniques and model

reference adaptive system. The Model reference adaptive system observers are the well-known

estimation methods due to their simplicity and direct physical interpretation (Maheswari NV,

2018)

One of the popular and growing adaptive control method used in practical applications is the

model reference adaptive system with Proportional-Integral controller as the adaptive mechanism

(Sivaraman E, 2015). It was used at a first time by Schauder (Schauder C, 1992). Since then, it

has been a topic of many publications. The Model reference adaptive system is important since it

2
leads to relatively easy to implement system with high speed of adaptation for a wide range of

applications (Bekakra Y, 2011). The preliminary idea of the Model reference adaptive system as

proposed by Landau (Landau Y, 1979) is to adjust the state variables of a system under study

using two different sub models and find out an adaptation mechanism to minimize the error

between the two models, in order to estimate a desired physical quantity. Depending on the

choice of output quantities that form the error vector (flux, stator current, back emf, reactive

power, etc.), several model reference adaptive system structures are possible (Sedhuraman k,

2012). The selection of reactive power as a function for Model reference adaptive system based

speed estimator deduced simpler system model equation independent of flux, which is easier to

design and implement and become advantageous in real time applications (Jevremovic V.R,

2010).

Particle swarm optimization (PSO) is a population based Stochastic optimization technique

developed by Dr. Eberhart and Dr. Kennedy in 1995(Kennedy J, 1995). The algorithm was

devised to simulate behavior of social animals such as ants foraging, fish schooling, bird

flocking…in which an iterative procedure is used to improve candidate solutions (called

particles) to attain a prescribed degree of quality (Yung C.2018). The algorithm is a one way to

achieve optimum control in that it allows rapid convergence, low operational cost and is suitable

for a range of different conditions (Yung C.2018). Moreover, it is shown to be effective in

optimizing difficult multidimensional discontinuous problems in a variety of fields. Similar to

other population-based optimization method such as the Genetic algorithm, the Particle swarm

optimization algorithm starts with random initialization of a population of individuals in the

search space. Each particle in the search space is adjusted by its own flying experience and the

3
other particles flying experience to find the global best solution at each generation (Lin FJ,

2009).

1.2 Problem Statement:

Parameter sensitivity and high computational effort are the main shortcomings of sensor-less

control. The Sensor-less drives provide good dynamic performance in medium and high speed

regions, but find it challenging in low and zero speed of operation (Holtz J, 2002). At low speed,

when the stator frequency becomes close to zero, most Sensor-less drives fail which makes them

improper for applications such as cranes and tractions, where the drive is required to maintain

full load torque down to zero speed (Tisla B, 2012). Therefore the issue of extending the region

of operation around low speed became an active area of research. The MRAS technique for

sensorless control employing reactive power as state variable automatically makes the system

immune to stator resistance variation; moreover it eliminates the requirement for any flux

estimation in the process of computation. However the problem of instability in the regenerative

mode of operation still persists (Rashed, 2004; Maiti et al., 2008; Wang and Deng, 2012). The

most important part of the model reference adaptive system estimation process is the design of

the adaptation mechanism, which at the end will determine the stability of the system. This

adaptation mechanism keeps track of the process output and the model output, and then calculate

a suitable parameter setting such that the output difference tend to zero (Ahmad RH, 2006). The

adaptation mechanism is mostly implemented using a Proportional-integral (PI) controller which

ensures that error as a result of comparing the outputs of reference and adaptive models

asymptotically converges to zero. The classical heuristic tuning methods of the PI controller

involves trial and error, but can provide a satisfactory performance over a wide range of ideal

operation (George E, 2012). However, a satisfactory performance may not be achieved with the

4
use of the PI controller under sudden change in speed and load disturbance conditions. With each

parametric variation in the system, the MRAS estimator will be perturbed; as a result the

controller may not provide good tuning and tends to produce overshoots. Moreover, a precise

mathematical model, continuous tuning and accurate gain values of the parameters are required

in order to achieve high performance drives, which is difficult using the fixed gain values of the

PI controller. Optimization of controller gains by an artificial intelligence search technique like

the particle swarm optimization in this work can allow the estimator to adapt to each parametric

variation, thereby increasing the robustness of the sensorless field oriented control even at low

speed.

1.3 Aim and Objectives

The aim of this research is to apply Particle swarm optimization for tuning the adaptation

mechanism of a Reactive-power based model reference adaptive system estimator.

The objectives are:

I. Model a reactive power model reference adaptive system estimator scheme that can be

used to provide a real time adaptive estimation of the rotor speed.

II. Apply particle swarm optimization for finding the optimal parameter gains in the

adaptation mechanism of the estimator.

III. Simulate a field oriented control scheme for the three phase induction motor using step 1

and 2.

1.4 Research Methodology:

i. Modeling a “reactive-power model reference adaptive system estimator

ii. Application of Particle swarm optimization algorithm with formulated objective

function for tuning the adaptive mechanism of the observer Scheme.

5
iii. Simulating step (iii) in the MATLAB/SIMULINK environment.

iv. Represent the obtained result in graphical, bar-chart and tabular forms.

v. Comparative analysis with the lead paper.

6
 CHAPTER TWO

LITERATURE REVIEW

2.1 Introduction

The literature review comprises of the overview of fundamental concepts as well as the review of

similar works. In the review of fundamental concepts, most of the pertinent works and the

fundamental theories that are used for the success of this research are reviewed, after which

review of similar works followed

2.2 Review of Fundamental concepts:

In this subsection, concepts that are fundamental and pertinent to understanding the rotor speed

estimation process of a three phase induction motor are reviewed.

2.2.1 Three-Phase Induction Motor

Induction motors derive their name from the way the rotor magnetic field is created. The

rotating stator magnetic field induces currents in the short circuited rotor. These currents produce

the rotor magnetic field, which interacts with the stator magnetic field, and produces torque,

which is the useful mechanical output of the machine. The three phase squirrel cage induction

motors are commonly used in industry applications. It has three main parts; rotor, stator and

enclosure. The stator and the rotor do the work and the enclosure protects the stator and rotor.

Most induction motors are rotary type with basically a stationary stator and a rotating rotor. The

stator has a cylindrical magnetic core that is housed inside a metal frame. The stator magnetic

core is formed by stacking thin electrical steel laminations with uniformly spaced slots stamped

in the inner circumference to accommodate the three distributed stator windings. The stator

windings are formed by connecting coils of copper or aluminum conductors that are insulated

7
from the slot walls. The rotor consists of a cylindrical laminated iron core with uniformly spaced

peripheral slots to accommodate the rotor windings (Workagen T., 2017)

Fig 2.1 Sectional view of a three phase induction motor (Burham J, 2013)

The principle of operation of the induction motor is based on generating a rotating magnetic

field. When three phase induction motor is connected to the appropriate voltage source it

produces stator current and each of three coils for each phase is fed with an alternating current.

The current induced in each phase generate magnetic field intensity and this cause a flow of

magnetic flux. The rotating magnetic field interacts with a set of conductors arranged on the

rotor and short circuited at the ends with two rings. This interaction between the magnetic field

and the conductor induces a current in the bars. A current flows in the conductors of the rotor

through the short circuiting rings at the end. This current in turn produces a magnetic field. There

is a difference in between revolving field speed and rotor speed then the revolving field induces a

voltage in the rotor winding. The difference between the rotor and the revolving field speeds is

8
called the slip speed. The induced voltage results in a rotor current that generates a flux in the

counter direction to the flux generated by the stator windings. The interaction between the rotor

magnetic field and the squirrel cage bars induces torque and causes rotation (Workagen T, 2017)

2.2.2 Field oriented control:

The closed loop control of the induction motor drive is normally required to satisfy the steady

state and transient performance specifications of ac drives. The control strategy can be

implemented using scalar control and vector control. The scalar control is characterized by poor

dynamic and static performance (Ayman Y, 2015).

The most popular vector control method was the one invented by Hasse (1970) called field

oriented control. The aim of the control is to find the rotor flux angle that is used to convert a

three phase system from stationary reference frame into a synchronous reference frame and vice

versa. The calculating rotor flux angle is achieved by measuring the rotor speed and calculating

the slip speed (Ali NH, 2017).

Field oriented control can be carried out by system-coordinate transformations of the basic

equations of the motor. After applying the transformations, the alternating and sinusoidal

quantities became non alternating quantities. Due to uncoupling, the currents can be controlled,

and then after back transformation it is possible to modify the output of the inverter with three

phase quantities. In this way the magnitude and phase of the supply voltage can be modified

(Kohlruszg, 2011).

Sensorless vector controlled induction motor drives essentially mean vector control without any

speed sensor. An incremental shaft mounted speed encoder, usually an optical type is required

for closed loop speed speed or position control in both vector controlled and scalar controlled

drives (Santosh k, 2014).

9
The equivalent circuit of the induction motor is generally used to analyze the machine and

calculate its different quantities at different loading conditions. However this representation

cannot be used to study the transient and steady state operation when the machine is fed by a

non-sinusoidal source (Chaozheng M, 2003). Based on the two axis theory of the electrical

machines and space vector theory, a dynamic model of the induction motor can be derived to

allow the analysis of the machine performance at any operating condition and any type of power

supply. The 3-phase motor quantities (such as voltages, currents, magnetic flux, etc.) can be

expressed in terms of complex space vectors (Ahmed I., 2015). Such model is valid for any

instantaneous variation of voltage and current and adequately describes the performance of the

machine under both steady-state and transient operation. Complex space vectors can be

described using only two orthogonal axes. The motor can be considered as a 2-phase machine.

The utilization of the 2-phase motor model reduces the number of equations and simplifies the

control design (Binder A, 2009)

Any three time varying quantities, which always sum to zero and are spatially separated by 120°

can be expressed as space vector. The space vector can be defined by considering the

instantaneous values, ua , ub ,u c. A three phase system defined by ua ( t ) , ub ( t ) ,u c (t ) can be

represented uniquely by a rotating vector. The space vector u may represent the motor variables

(voltage, current and flux).

10
 Fig 2.2 State vector representation for three phase variables

Assume thati sa, i sb∧i sc are the instantaneous balanced 3-phase stator currents, which are

sinusoidal and shifted by 120 degrees in the time domain. Then:

i sa +i sb+ i sc=0……………………………………. (2.1)

The stator current space vector can then be defined as (Vas P, 1998)

2 2
i s= (i a +α i b +α i c )…………………………….. (2.2)
3

2π 4Π
Where; α =e j 3 and α 2=e j 3 are spatial operators………….. (2.3)

To find the 2 axis components of the stator current vector i s, equation (2.3) is substituted into

equation (2.2) which leads to the following equations:

2 1 1
i sα = i sa − i sb − i c ……………………………………….. (2.4)
3 3 3

1 1
i sβ = i sb− i …………………………………………….(2.5)
√3 √ 3 sc

11
This current space vector, which represents the three phase sinusoidal system, could be

transformed into a two time invariant coordinate system by using the field oriented control. The

field oriented control consists of two transformations namely, Clarke transformation, and Park

transformation.

2.2.2.1 The (a-b-c to α-β) Clarke transformation:

This transformation output a two coordinate time variant system from a three-phase system. The

transformation matrix is given below

Figure 2.3 Stator current space vector in stationary reference frame

12
Here the space vector can be expressed in a reference frame with only two orthogonal axes called

(α,β) It is assumed that the axis "a" and the axis "α" are in the same direction as shown in Figure

2.3 above.

Since in a three-phase balanced system, the sum of three phase currents i a +i b +i c = o.

The projection that modifies the (a-b-c) three-phase system into the (α-β) two dimensional

orthogonal systems is given as:

i sα =i a………………………………………………… (2.3)

1 2
i sβ =( )i a +( )i b………………………………….. (2.4)
√3 √3

If the three phase currents are assumed as

i a=Isin ( ωt ) … … … … … … … … … … … … … … … … … ..(2.5)

(
i b=Isin ωt+
2π
3 )
… … … … … … … … … … … … … … … .(2.6)

(
i c =Isin ωt −
2π
3 )
… … … … … … … … … … … … … … … .(2.7)

Then the output( α , β ) currents are

i α =Isin ( ωt ) … … … … … … … … … … … … … … … … … …(2.8)

(
i β =Isin ωt+
π
2)… … … … … … … … … … … … … … … …(2.9)

2.2.2.2 Park transformation:

Park transformation is used to rotate the two axis coordinate system so that it is aligned with the

rotating motor and this projection modifies a two phase orthogonal a-b system in the d-q rotating

13
reference frame. It changes the two-phase orthogonal system (α-β) into (d-q-0) rotating reference

frame. The transformation matrix is given below:

Figure 2.4 Stator current space vector in rotating reference frame

The flux and torque components of the current vector are determined by the following equations

i d =i α cosθ +i β sinθ…………………………………. (2.10)

i q=−i α sinθ+i βcosθ………………………………… (2.11)

2.2.3 The induction motor mathematical model:

Ideally the motor model is symmetrical, with a linear magnetic circuit characteristic. The system

model defined in the stationary α −β coordinate system attached to the stator is expressed by the

following equations which are based on Kirchhoff’s law

The stator voltage equations are given below:

14
 d Lm d ψ sα
V sα =R s i sα +σ Ls i sα + ……………………… (2.7)
dt Lr dt

d Lm d ψ r β
V sβ =Rs i sβ +σ Ls i sβ + ……………………… (2.8)
dt Lr dt

The Rotor voltage equations are given below:

d
V rα =0=Rr i sα + ψ +ω ψ ………………………….. (2.9)
dt rα r rβ

d
V rβ=0=R r i rβ + ψ −ω r ψ rα … … … … … … … … …...(2.10)
dt rβ

Where ω r is the angular velocity of the rotor

The flux linkage equations are given as:

ψ sα =Ls i sα + Lm i rα ……………………………………… (2.11)

ψ sβ =Ls i sβ + L m i rβ … … … … … … … … … … … … … … ….. (2.12)

ψ rα =Lm i sα + Lr i rα ……………………………………. (2.13)

ψ rβ=Lm i sβ + Lr i rβ … … … … … … … … … … … … … … … .(2.14)

Where Ls =Lis + Lm

Lr =Lir + Lm

2.2.4 Model Reference Adaptive System Observer:

A simple definition for adaptive control is “a system that adapts itself to changes in the process”.

It is clear from this definition that process state variations gives rise to adaptation of the system.

The aim of reacting to states changes is to maintain a high system performance, even if the

process states are unknown or varying (Eray O, 2005). Model reference adaptive system

(MRAS) is one of the famous speed estimation usually used for sensorless speed control of

induction motor drive. It is one of many promising techniques employed in adaptive control.

Among various types of adaptive system configuration, MRAS is important since it leads to

15
relatively easy to implement systems with high speed of adaptation for a wide range of

applications (Workagegn T, 2017). One of the most noted advantage of this type of adaptive

system is its high speed of adaptation. This is due to the fact that a measurement of the difference

between the outputs of the reference model and adjustable model is obtained directly by the

comparison of the states of the reference model with those of the adjustable system. The block

reference model represents demanded dynamics of actual control loop. The block adjustable

model has the same structure as the reference one, but with adjustable parameters instead of the

unknown ones as shown in the Figure below.

Figure 2.4 General structure of MRAS (Mohammad A., 2017)

The error between measured and estimated state variables is then used to drive an adaptation

mechanism which generates the estimated speed, for the adjustable model as shown in the

diagram above. It should be noted that, speed estimation methods using MRAS can be classified

into various types according to the state variables. The most commonly used are the rotor flux

based MRAS, reactive power MRAS, back-emf based MRAS, and stator current based MRAS.

16
In rotor flux based MRAS, the presence of an open integration in the stator leads to problems

with initial conditions and drift. A low pass filter may be used instead of the pure integration;

however, it has a degrading effect on speed estimation at low speeds and introduces time delay.

The model reference adaptive approach based on back-emf rather than the rotor flux offers an

alternative to avoid the problem of pure integration. But still there is another source of

inaccuracy which is a possible mismatch of reference model parameters.

The reactive power based MRAS present a speed identification system with low sensitivity to

parameter variations, the stator resistance in particular which affects applications of low speed

drives. It is based on the comparison between the outputs of two observers. The observers are

used to calculate the instantaneous reactive power maintaining the magnetizing current

(Chaozheng M, 2003). The input data required for this model is the stator voltages and stator

currents in the (α, β) stationary reference frame. Two sets of equations are developed to compare

reactive power of the induction machine in the reference model and the adaptive model. The

reference model does not involve the rotor speed, while the adaptive model needs the estimated

rotor speed to adjust the computed reactive power to that computed from the reference model.

The error between two models is used to drive a suitable proportional integrator (PI) controller,

^ r.
which generates an estimated rotor speedw

2.2.4.1 Reference Model: The continuous time representation for back-emf of induction motor

expressed in the stationary reference frame is derived from equations (2.7), (2.8), (2.11) and

(2.12) expressed by:

L m d ψ rα d
e mα = =V −Rs i sα −σ Ls i sα …………………….. (2.15)
Lr dt sα dt

Lm d ψ rβ d
e mβ= =V −R s i sβ −σ Ls i sβ…………………….. (2.16)
Lr dt sβ dt

17
 e m=emα + j e mβ………………………………………………. (2.17)

2
Lm
Whereσ =1− is the leakage coefficient
Lr Ls

The reactive power q, which represents the instantaneous reactive power that maintains the

magnetizing current, can be computed from the following equation:

q m=i s × em……………………………………………………. (2.18)

The continuous time representation is further derived as

d d
q m=i sα v sβ −i sβ v sα −σ Ls (i sα i sβ −i sβ i s α ) …………… (2.19)
dt dt

Before being used in digital systems, it has to be in discrete form; therefore the discrete form of

the reactive power equation is obtained by backward approximation of equation (3.5) with the

sampling period T, then

i sα ( k ) i sβ ( k ) −i sβ ( k−1 ) i ( k ) −i sα ( k −1 )
q ( k )=i sα ( k ) v sβ ( k ) −i sβ ( k ) v sα ( k )−σ L s [ −i sβ (k ) sα ]…. (2.20)
T T

Further simplified as:

L
q ( k )=i sα ( k ) v sβ ( k ) −i sβ ( k ) v sα ( k )−σ {i ( K ) i sα ( k −1 )−i sα ( k ) i sα ( k−1 ) }…… (2.21)
T sβ

The per unit discrete time representation is obtained by dividing equation (3.7) with base power

of v b i b

q ( k )=i sα ( k ) {v sβ ( k )−k 1 i sβ ( k ) [ v sα ( k ) + k 1 i sα ( k−1 ) ] } pu…………………………. (2.22)

σ Ls I b
Where k 1=
T Vb

V b is thebase voltage

I b is base current

2.2.4.2 Adaptive model:

18
The active power in the adaptive model is computed from the cross product of stator current and

the estimated counter electromotive force (emf). The continuous time representation of the

estimated back-emf in the adaptive model is derived from equations (2.9) to (2.14), expressed as
2 2
L m i mα Lm
e^ mα = d ¿ (−τ r w
^ r i mβ−i mα +i sα )……………………….. . (2.23)
Lr dt Lr τ r

2 2
Lm i mβ Lm
e^ mβ= d ¿ (−τ r w
^ r i mα −i mβ +i sβ )…………………………. (2.24)
Lr dt Lr τ r

Lr
Where τ r= , is rotor time constant
rr

The currents i mα , i mβcan be computed as follows

di mα
^ r i mβ− 1 i mα + 1 i sα ………………………………………… (2.25)
=− w
dt τr τr

di mβ
^ r i mα− 1 i mβ + 1 i sβ………………………………………. … (2.26)
=w
dt τr τr

The estimated reactive power in the continuous time is written as:

q^ =i s × e^ m=i sα e^ mβ−i sβ e^ mα … … … … … … … … … … … … … … … … … .. ( 2.27 )

¿ is ×
(
Lm L m
Lr T r
1
i s − ψ^ r + ω
Tr )
^ r ψ^ r … … … … … … … … … … … … … … … … ..(2.28)

Which is been discretized as:

q^ ( k )=i s ( k ) × e^ m ( k ) =i sα ( k ) e^ mβ ( k ) −i sβ ( k ) e^ mα ( k ) … … … … … … … … (2.29)

The discrete time representation of the back-emf equations is given by

2 2
Lm d i mα Lm
e^ mα (k )=
Lr dt
¿
Lr τ r
{−τ r ^wr ( k ) imβ ( k )−imα ( k ) +isα ( k ) } … … … … … .(2.30)

2 2
Lm d i mβ Lm
e^ mβ (k)=
Lr dt
¿
Lr τ r
{−τ r ^
wr ( k )r i mα ( k )−i mβ ( k ) +i sβ ( k ) } … … … … …(2.31)

19
The per unit representation of the back-emf is obtained by dividing equations (2.30) and (2.31)

with base voltage V b , hence

e^ mα ( k )=k 2 {−k 3 ω
^ r ( k ) i mβ ( k ) −imα ( k ) +i sα ( k ) } pu … … … … … … … … …(2.32)

e^ mβ ( k ) =k 3 {−k 3 ω
^ r ( k ) i mα ( k ) −i mβ ( k ) +i sβ ( k ) } pu … … … … … … … … …(2.33)

2
Lm I b
Where k 2=
Lr τ r V b

Lr ω b
k 3=
τr

The discrete time representation of equations (3.11) and (3.12) is given as

{ } ( ) ( ) [ ]
2 2 2 2 2
−T 2 T T T T T T
i mα ( k )=i mα ( k −1 ) ^ r ( k ) +1− + 2 −i mβ ( k −1 ) w
w ^r (k ) T − +i sα ( k ) − 2 −i sβ ( k ) ^
wr ( k ) ……………
2 τr τr τr τr 2 τr 2 τr

{ } ( ) ( ) [ ]
2 2 2 2 2
−T 2 T T T T T T
i mβ ( k )=imβ ( k−1 ) ^ r ( k )+1− + 2 −i mα ( k−1 ) w
w ^ r (k ) T − + i sβ ( k ) − 2 −i sα ( k ) w
^ r (k ) … …………
2 τr τr τr τr 2 τr 2τr

The per unit representation is obtained by dividing equations (2.34) and (2.35) by the base

current I b

i mα ( k )=i mα ( k −1 ) ¿(k)+k 5 ¿−i mβ ( k−1 ) w

^ r ( k ) k 6 +i sα ( k ) k 7−i sβ ( k )^
wr ( k)k 8 ………. (2.36)

i mβ ( k )=imβ ( k−1 ) ¿(k)+k 5 ¿−i mα ( k −1 ) w

^ r ( k ) k 6+ i sβ ( k ) k 7−i sβ ( k )^
w r ( k)k 8

……….. (2.37)
2 2 2
ω T T T
Wherek 4= b , k 5=1− +
2 τ τr

−T T 2 ωb T
2
k 6=ω b ¿ ), k 7= + 2 k 8=
τr τ 2

2.2.4.3 Adaptation Mechanism:

20
The error/speed tuning signal ε can be minimized by applying Proportional integral controller,

and therefore it can be considered as the adaptation mechanism of the observer. The PI controller

monitors the error between a desired set point and a process variable.

The continuous time equation of the PI controller is written as

^r (t)=k p ε ( t )+ k i∫ ε ( t ) dt = y 1 ( t ) + y 2 ( t ) … … … … … … … … … … .(2.38)
ω

Applying Laplace transform gives

ki
^r (t)=k p ε (s)+
ω ε ( s ) =Y 1(s) +Y 2(s)…………………………….. (2.39)
s

(
^r ¿ k p +
ω
s)
ki
ε … … … … … … … … … … … … … … … … … … … … … … … …(2.40)

Where the error of the reactive power being the speed tuning signal is given by

ε =q m−^
qm

= i s × ( e m− e^m ) … … … … … … … … … … … … … … … … … … … … … … ..(2.41)

^r such that the reactive power q^ generated by

The PI controller tunes the estimated rotor speed ω

the adaptive model matches the q generated by reference model. From equations (2.22), (2.29)

and (2.40), it is evident that the speed estimation process of figure 2. Is robust to stator resistance

and thus not affected by stator resistance thermal variations

2.2.4.4 Implementation of MRAS

To implement the proposed MRAS scheme, some parameters are needed which include

 Number of poles

 Rotor resistance

 Stator leakage inductance

 Rotor leakage inductance

 Magnetizing inductance
21
Base quantities are

 Base current

 Base voltage

 Base angular velocity

 Sampling period

2.2.4. Particle Swarm Optimization:

Particle swarm optimization is a population based heuristic search algorithm devised to simulate

social behavior in which an iterative procedure is used to improve candidate solutions (called

particles) to attain a prescribed degree of quality (Yung C, 2018). Here, instead of using genetic

operators, individuals called particles are “evolved”. A particle represents a potential solution to

a problem. Each particle adjusts its flying according to its own flying experience and its

companion’s flying experience. Each particle is treated as a point in a D-dimensional space the i th

particle is represented as X i . The best previous position giving the minimum fitness value of any

particle is called particle’s best represented as pbest . The index of the best particle among all

particles in the population is called global best represented by the symbol gbest .

In this work, the inertia weight method was used because it promotes the search for an optimum

solution in the initial computation stage and improves convergence quality in the follow up

computation The inertia weight PSO algorithm applies the concept of progressive deceleration.

In the inertia weight PSO algorithm, the particle is configured with large-scale fast search

velocity in the initial stage, and when the iteration is gradually increased, the searching velocity

is gradually decreased to improve convergence (Yung C, 2018)

22
Advantages of PSO:

The advantages of the Particle Swarm Optimization include; it is an auto tuning method, it does

not require detailed mathematical description of the process and it finds the optimal parameters

based on the performance index provided for algorithm convergence (Sivaraman E, 2015)

Fig 2.5 Two dimensional relationship between velocity and position search space for a particle

(Yung C, 2019)

The iteration formulas are given below:

v i ( k +1 )=w . v i ( k )+ c1 . r 1 ( p best −x i ) +c 2 . r 2 ( gbest −xi )……… .(3.27)

x i ( k +1 )=x i ( k ) +v i ( k +1 ) … … … … … . … … … … … … … … … … …(3.28)

Where:

i denote each particle

k is the discrete time index

th
v i ( k )∧v i ( k +1 ) are the current∧next velocities of thei particle

th
x i ( k )∧x i ( k +1 ) are the currrent∧next positions of the i particle

w is the weighting factor

23
 '
pbest i is the particle s bestposition

gbest i is the optimum position solution of the swarm particles

c 1∧c 2 are acceleration constants

r 1∧r 2 are uniform distribution random generated numbers with a range(0 ,1)

w . v i ( k ) stands for improvement effect of convergence

the major steps in the procedure of PSO algorithms are described below:

Step 1 : Generate a population of particles with random positions and velocities in the search

space

Step 2: Evaluate the objective function of each particle in the swarm

Step 3 : Modify the searching point of each particle

Step 4: Check the loop terminating condition

24
 Fig 2.6 PSO algorithm

2.5 Review of similar works:

In order to have adequate knowledge of what is obtainable presently in the area of this research,

it is very important to carry out a review of some related works done by different authors.

Author and year Findings Limitations

Prashant M et Speed estimation using artificial Rotor flux as variables needs flux

al.,2020 neural estimator

Network

Vo Q et al.,2019 Speed estimation using fuzzy logic Intensive computational analysis

controller

Maheswari N et Performance analysis based on Parameter variations

al.,2018 conventional PI-controller

Aymen F.,2017 Estimation using Particle swarm Analysis limited to flux weakening

optimization region

Acikgoz H et Speed estimation based on Genetic Bulky computations

al,2015 algorithm and fuzzy controller

Sedhuraman k et Estimation based on neural learning Problems have to be translated into

al,2013 adaptive system numerical values

Barut M et Optimization via differential Complex procedure and poor search

25
 al,2013 evolution algorithm ability of the algorithm

Gadoue S et Artificial neural network based Computationally expensive and

al.,2009 estimation hardware dependence

Table 2.1: Outlook on literature review

CHAPTER THREE
DESIGN AND METHODOLOGY
3.1 Introduction
In this chapter, an elaborate procedure for the actualization of the work is discussed. The
materials employed and the parameters setting used for the PSO algorithm. Also, a
comprehensive methodology adopted for this work is given
3.2 Material:
The material employed for the research is as follows
3.2.1 Computer Specifications:
Simulation analysis were carried out using HP laptop with the following specifications
i. Intel Core i5 3317U
ii. 1.70 GHz Processor
iii. 8 GB RAM
iv. 64 bits Windows10 Operating System
3.2.2 MATLAB 2016a Software
3.2.3 Induction Motor Parameters
3.3 METHODOLOGY:
The objectives as outlined are achieved in line with the following procedures:
3.3.1 Modeling a Reactive-Power Model Reference Adaptive System Estimator

v sαβ
Reference Model
i sαβ Equation (2.2) + Qref

Adjustable Model
26
Equation (2.9)
 _ Qest ε (speed tuning signal)

^ r−est
w
PSO-PI

Adaptation

Fig 3.1 Block diagram of the proposed Reactive power MRAS model

3.3.2 Formulation of objective function:

The performance of the three phase induction motor varies according to the adaptation

mechanism gains. The PI controller minimizes the error function e ( t ) in the MRAS observer,

which reduces the value of the specified objective function. The integral time absolute error

(ITAE) performance index is chosen as objective function, because it has the benefit of

producing less oscillations and smaller overshoot (Rashmita, 2014). The purpose of Particle

swarm optimization is to minimize the objective function. The objective function determines

wither a particle is the best solution.

∞

The ITAE index is expressed as, ITAE = ∫|e( t)|tdt … … … … … … … … … … .(3.25)

0

Where e (t) is the error signal

The objective of the QMRAS is to minimize the ITAE of the following total objective function

e q(t) = q^ −q

Therefore objective function = ∫|q^ −q|t dt … … … … … … … … … … .... (3.26)

0

Where the fitness value is reciprocal of the magnitude of objective function

The inertia weight particle swarm optimization algorithm was used in this work. The particle is

configured with large scale fast search velocity in the initial stage, and when the iteration is

gradually increased, the searching velocity is gradually decreased to improve convergence.

27
The algorithm steps are given below

Step 1: initialize positions and velocities of particles (k p∧k i ) randomly in the problem space.

Step 2: Calculate the fitness function of each particle in the swarm using equation (2)

Step 3: Find the individual best ( pbest ¿ and the global best position in the swarm

Step 4: Start iteration

Step 5: Update particles positions and velocities using equations (3.27) and (3.28)

Step 6: Calculate the fitness function at their corresponding positions

Step 7: Update pbest and gbest based on the fitness function

Step 8: Iteration = iteration + 1

Step 9: If iteration¿ max iteration, then go to step5 otherwise continue

Step 10: Obtained gbest is the optimum value of PI parameter gains

Step 11: Apply the optimum values to the system

3.3.3 PSO parameters selection

For the purpose of this research, the initialization and user defined parameters for the PSO are

presented in table

Table 3.3: PSO Variable Value Description parameters

n 25 Size of the swarm

step 10 Maximum number of iterations

dim 2 Dimension of the search space

C1 1.5 Velocity constant

C2 0.15 Velocity constant

w 0.9 Inertia weight

Min k p 10 Minimum value of k p

Max k p 60 Maximum value ofk i

28
Min k i 0 Minimum value of k i

Maxk i 5 Maximum value of k i

29
 start

Select parameters of PSO

Initialize particles with random

velocities and positions

Run the IM in FOC

Update position and Calculate the objective function for each

velocity of particles particle using equation (2

Update Particle's best and Global best of

Iter + 1 the swarm using eqn (2

Maximum
number of
iterations
reached

Optimal PI parameter
gains

End

Figure 3.2: Proposed flowchart of the optimization process

30