Вы находитесь на странице: 1из 56

Multirate Output Sampling Based Load-Frequency Control

A thesis submitted in partial fulfillment of


the requirements for the degree of

Bachelor of Technology

Submitted by
Poonam Chand (2012UEE1378)
Ram Raj (2012UEE1216)
Dharmpal kumar (2012UEE1488)
Shubham Gupta (2012UEE1618)

Supervised by
Dr. Satyanarayana Neeli
Assistant Professor
Department of EE
MNIT Jaipur

Department of Electrical Engineering

Malaviya National Institute of Technology Jaipur


May, 2016

Malaviya National Institute of Technology, Jaipur

Certificate
This is to certify that following students of Department of Electrical Engineering, Malaviya National
Institute of Technology, Jaipur, have undergone a project titled Multirate Output Sampling Based
Load-Frequency Control under my guidance. During this project their performance is found

Poonam Chand (2012UEE1378)


Ram Raj (2012UEE1216)
Dharmpal kumar (2012UEE1488)
Shubham Gupta (2012UEE1618)

satisfactory/unsatisfactory
satisfactory/unsatisfactory
satisfactory/unsatisfactory
satisfactory/unsatisfactory

Dr. Satyanarayana Neeli


Assistant professor
Department of EE
MNIT Jaipur

Malaviya National Institute of Technology, Jaipur

Candidates Declaration
We hereby declare that this project report on Multirate Output Sampling Based LoadFrequency Control which is being submitted in partial fulfillment of the award of degree
Bachelor of Technology in Electrical Engineering, is the result of work carried out by us under
the guidance of Dr. Satyanarayana Neeli, Assistant Professor of Malaviya National Institute of
Technology, Jaipur. We further admit that this project work has not been submitted to MNIT
before or for other purpose.

Poonam Chand
(2012UEE1378)

Ram Raj
(2012UEE1216)

Dharmpal kumar
(2012UEE1488)

Shubham Gupta
(2012UEE1618)

Date:
Place:

Malaviya National Institute of Technology, Jaipur

Acknowledgement
We are highly indebted to our supervisor Dr. Satyanarayana Neeli, Assistant Professor, Department of
Electrical Engineering and would like to express our special gratitude for his guidance and constant
supervision as well as for providing necessary information regarding the project and also for his support
in completing the project.
We would like to thank Dr. Vikas Gupta, Head of Electrical Engineering Department for giving us this
opportunity to do this project work.
We would also like to thank Project Coordinator, Mr. Vinod Sahai Pareek, Associate Professor,
Department of Electrical Engineering and Ms. Nikita Jhanjariya, Associate Professor, Department of
Electrical Engineering for giving us this opportunity to explore new area and enhance our knowledge.
We would also like to thank our institution and faculty members, without whom this project would have
been a distant.

Poonam Chand
(2012UEE1378)

Ram Raj
(2012UEE1216)

Dharmpal kumar
(2012UEE1488)

Shubham Gupta
(2012UEE1618)

ABSTRACT
In an interconnected power system, if a load demand changes randomly, both frequency and tie
line power varies. The main aim of load frequency control is to minimize the transient variations in these
variables and also to make sure that their steady state errors is zero. Many modern control techniques are
used to implement a reliable controller. The objective of these control techniques is to produce and
deliver power reliably by maintaining both voltage and frequency within permissible range. When real
power changes, system frequency gets affected while reactive power is dependent on variation in voltage
value. Thats why real and reactive power are controlled separately. Control of load frequency controls
the active power. The role of automatic generation control (AGC) in power system operations with
reference to tie line power under normal operating conditions is analyzed. Future power systems will
rely on large amounts of distributed generation with large percentage of renewable energy based
sources, what will further increase system uncertainties and thereby induce new requirements to the LFC
system.
Power systems consist of control areas representing a coherent group of generators i.e.
generators which swing in unison characterized by equal frequency deviations. In addition to their own
generations and to eliminate mismatch between generation and demand these control areas are
interconnected through tie-lines for providing contractual exchange of power under normal operating
conditions. One of the control problems in power system operation is to maintain the frequency and
power interchange between the areas at their rated values. Automatic generation control is to provide
control signals to regulate the real power output of various electric generators within a prescribed area in
response to changes in system frequency and tie-line loading so as to maintain the scheduled system
frequency and established interchange with other areas.
The report presents a full state feedback controller for load-frequency control (LFC) in control
areas (CAs) of a power system. As it uses full-state feedback it can be applied for LFC not only in CAs
with thermal power plants but also in CAs with hydro power plants, in spite of their non-minimum phase
behaviours. To enable full-state feedback we have proposed a state estimation method based on fast
sampling of measured output variables, which are frequency, active power flow interchange and
generated power from power plants engaged in LFC in the CA. The same estimation method is also used
forth estimation of external disturbances in the CA, what additionally improves the overall system
behaviour.

Table of Content
1

1. Introduction
1.1 Load Frequency Control
1.2 Need of Maintaining Constant Frequency
1.3 Frequency power characteristic of synchronous generator
1.3.1 Concept of Load-Frequency Control
1.4 Power Swing Equation
1.5 Automatic Load Frequency Control
2. Modeling of ALFC
2.1 Introduction
2.2 Discussion on Speed Governor Model
2.3 Turbine Model
2.4 Power System Model
2.5 State Space Model for single Area System
3. State Space and Multirate Output Sampling
3.1 Introduction to state space
3.1.1 General state-space model
3.2 Solution of Continuous Time State Space Model
3.3 Importance of Discrete Time Controller
3.4 Discretization of the continuous time system
3.5 Numerical Examples
3.5.1 Example 1
3.5.2 Example 2
Discretization of continuous system when input is exponential without disturbance
3.5 Multirate output sampling
4. Controller Design
4.1 Discretization of Continuous Time System with Disturbance
4.1.1 Example of Discretization of the System With Disturbance
4.2 Fast Output Sampling Method
4.3 Design Procedure
4.3.1 State Feedback Controller
4.3.2 Numerical Example
5. Conclusions
Appendix
References

3
3
2
2
4
4
5
7
7
7
11
11
14
15
15
16
17
18
18
19
19
20
20
22
24
24
25
27
29
29
31
33
34
37

1. Introduction
1.1 Load Frequency Control
Electricity is a form of energy and we need it for just about everything. Almost all of our modern
conveniences are electrically powered. Along with the growth in demand for electric power, sustainable
development, environmental issues, and power quality and reliability have become concerns. Electric
utilities are becoming more and more stressed since existing transmission and distribution systems are
facing their operating constraints with growing load.
A quality of power system can be judged by maintain of voltage and frequency at desired values
irrespective of change in loads. It is in fact impossible to maintain both active and reactive power
without control, which would result in variation of voltage and frequency levels. To cancel the effect of

load variation and to keep frequency and voltage level constant, a control system is required. Though the
active and reactive powers have a combined effect on the frequency and voltage, the control problem of
the frequency and voltage can be separated. Frequency is mostly dependent on the active power and
voltage is mostly dependent on the reactive power. The successful operation of interconnected power
systems requires the matching of total active power generation with load demand and associated power
system losses. As the demand deviate from its nominal value with an unpredictable small amount, the
operating point of power system changes, and hence system may experience deviations in system
frequency and scheduled tie line power exchanges, which may yield undesirable affects [14]. Thus, the
issue of controlling power systems can be separated into two independent problems. The active power
and frequency control is called as load frequency control (LFC). It has gained the importance with
change in power system structure and the growth of size and complexity of interconnected systems [15].
The most important task of LFC is to maintain the frequency constant against the varying active
power loads, which is also referred as unknown external disturbance. This is done by minimizing
transient deviations of frequency in addition to tie-line power exchange and also making the steady state
error to zero [8]. It has been shown that for small changes active power is dependent on internal
machine angle and is independent of bus voltage: while bus voltage is dependent on machine
excitation (therefore on reactive generation Q ) and is independent of machine angle . Change in
angle

is caused by momentary change in generator speed. Therefore, load frequency and excitation

voltage controls are non-interactive for small changes and can be modeled and analyzed independently.
Changes in load demand can be identified as: (i) slow varying changes in mean demand, and (ii) fast
random variations around the mean. The regulators must be designed to be insensitive to fast random
changes, otherwise the system will be prone to haunting resulting in excessive wear and tear of rotating
machines and control equipment.

1.2 Need of Maintaining Constant Frequency


The speed of the A.C motors depends on the frequency of the power supply. There are situations
where speed consistency is expected to be of high order. In the case of turbine, if the normal operating
frequency is 50 Hz and the turbines run at speeds corresponding to frequencies less than 49.9 Hz or
above 50.1 Hz, then the blades of the turbines may be damaged. The operation of a transformer below
the rated frequency is not desirable. When frequency goes below rated frequency at constant system
voltage then the flux in the core increases and then the transformer core goes into the saturation region.
Due to that power transformer results in low efficiency and over-heating of the transformer windings.
Change in frequency causes change in speed of the consumers plant affecting production
processes. Further, it is necessary to maintain network frequency constant so that power stations run
satisfactorily in parallel, the various motors operating on the system run at the desired speed [1]. The
electric clocks are driven by the synchronous motors. The accuracy of the clocks are not only dependent on
the frequency but also is an integral of the frequency error The most serious effect of subnormal

frequency operation is observed in the case of Thermal Power Plants. Due to the subnormal frequency
operation the blast of the ID (induced draft) and FD (forced draft) fans in the power stations get reduced
and thereby reduce the generation power in the thermal plants. This phenomenon has a cumulative effect
and in turn is able to make complete shutdown of the power plant if proper steps of load shedding
technique is not engaged. It is pertinent to mention that, in load shedding technique a sizable chunk of
load from the power system is disconnected from the generating units to restore the frequency to the
desired level.

1.3 Frequency power characteristic of synchronous generator


Since synchronous generators are the most common type of machines used in the generation
of electrical power, its characteristics can be used to describe the relationship between frequency and
power during load changes. The most common type of prime mover is a steam turbine, but other types
include diesel engines, gas turbines, water turbines, and even wind turbines. Regardless of the original
power source, all prime movers tend to behave in a similar manner. Any imbalance of power between
generation and consumption the speed of generator will vary corresponding. The decrease in speed is in
general non-linear, but some form of governor mechanism is usually included to make the decrease in
speed linear with an increase in power demand.
The Speed Droop (SD) of a prime mover is defined by
N nl N fl
100
SD =
N fl

Where,
N nl =

(1.1)

No-load prime-mover speed

N fl =

the full-load prime-mover speed.

Most generator prime movers have a speed droop of 2 to 4 percent, as defined in equation (1.1).
In addition, most governors have some type of set point adjustment to allow the no-load speed of the
turbine to be varied. A typical speed-versus-power plot (known as the house curve)

Fig1.1 Speed-power and frequency-power curves (house curves)


Although the house curves are only used for studying the parallel operation of two generators or that of a
single generator connected to a certain network, it helps understanding the variations of electrical
frequency as the power demanded is changed, since the shaft speed is related to the resulting electrical
frequency by the equation (1.2).
N P
f e= m
120
(1.2)

Then the power output of a synchronous generator is related to its frequency and this is clear in fig 1.1.
The relationship between frequency and power can be described quantitatively by the equation
P=S p [ F nl F SYS ]
(1.3)
where, P = power output of the generator
Fnl
= no-load frequency of the generator
F SYS
Sp

= operating frequency of system

= slope of curve, in kW/Hz or MW/ Hz

But this equation is not accurate for multi-area power systems. A similar relationship can be derived for
V
the reactive power Q and terminal voltage T , for which the AVR control loop is used.
1.3.1 Concept of Load-Frequency Control
In the steady state operation of power system, the load demand is increased or decreased in
the form of Kinetic Energy stored in generator prime mover set, which results the variation of speed and

frequency accordingly. Therefore, the control of load frequency is essential to have safe operation of the
power system.
Neglecting resistances
EV
P=
sin
X
(1.4)
Where E is the excitation voltage, V
is the power angle.
If changes to +

, then

P+ P=

is the terminal voltage,

P changes to

is the effective reactance and

P+ P

EV
sin ( + )
X

sin cos + sin cos


EV

Since

(1.5)

is very small.
cos 1 and

P+ P=

sin

EV
EV
sin +
cos .
X
X
P

So

EV
cos .
X

(1.6)
Or

Small power changes mainly depends on

or f .

1.4 Power Swing Equation


Transient stability in power system are done over a very small period of time equal to the time
required for one swing, which approximates to around 1 sec or even less [1]. When the synchronous
generator is fed with a supply from one end and a constant load is applied to the other, there is some
relative angular displacement between the rotor axis and the stator magnetic field, known as the load
angle which is directly proportional to the loading of the machine. The machine at this instance is
considered to be running under stable condition.
Now if we suddenly add or remove load from the machine the rotor decelerates or accelerates
accordingly with respect to the stator magnetic field. The operating condition of the machine now

becomes unstable and the rotor is now said to be swinging with respect to the stator field and the
equation we so obtain giving the relative motion of the load angle with respect to the stator magnetic
field is known as the swing equation for transient stability of power system. Here for the sake of
understanding we consider the case where a synchronous generator is suddenly applied with an
Pe
Pm
increased amount of electromagnetic load, which leads to instability by making
less than
as the rotor undergoes deceleration. Now the increased amount of the accelerating power required to
bring the machine back to stable condition is given by,

Pa=Pm Pe

Accelerating power

Accelerating torque is given by


T a=T mT e
Pa=M .
where,

M =I .

M is angular momentum. I is moment of inertia and is angular displacement .


=S +

d
dt

Angular acceleration is given by


d2 d2
= 2 = 2
d t d t
Now, Swing Equation is written as
M
(1.7)
where, electric power supplied by generator is given by
EV
Pmax sin
Pe =
sin
=
X
(1.8)
Putting the value of

Pe

in equation (1.7)

d2
=Pm Pe
2
d t

d2
=Pm Pmax sin
d2t

(1.9)
where,

Pm

= mechanical power input in MW

Pe

= electrical power output in MW

M = also called the inertia constant in MJ/MVA.


= rotor angular displacement from synchronously rotating reference frame (called torque
angle/power angle)

Fig 1.2 Flow of mechanical and electrical powers in a synchronous machine

1.5 Automatic Load Frequency Control


The main purpose of operating the load frequency control is to keep control the frequency during
the load changes. During the power system operation rotor angle, frequency and power are the subjected
parameters to variable. Changes in real power mainly affect the system frequency; in reactive power
mainly depend on variable of voltage magnitude. Hence, control of frequency and voltage can be
achieved separately Thus, real and reactive powers can be controlled separately.
The Automatic Load Frequency Control (ALFC) controls the real power and the Automatic
Voltage Regulator (AVR) regulates the voltage magnitude and hence the reactive power. The basic
requirement is to take care of megawatt power output of a generator matching with the changing load
and appropriate value of exchange of power linking to the control areas.it aims to facilitate control of the
frequency for larger interconnection.
A power system may be an interconnected system of multiple areas or an isolated system
comprising of single service area. The LFC plays an important role in both types of power systems. A
single area power system is the one, which comprises of a single generator supplying power to a single
service area. The function of LFC in a single area power system is to restore the frequency to the
specified nominal value in case of any fluctuation. However, in case of an interconnected power system,
two or more independently controlled areas are connected together. In such systems, along with
frequency, generation within each area also has to be controlled. This is required to maintain the

scheduled power interchange. So, the main aim of the load frequency control in multi area power
systems is to regulate the frequency to the specified nominal value and to maintain the interchange
power between areas at the scheduled values. However, in case of an interconnected power system, two
or more independently controlled areas are connected together. In such systems, along with frequency,
generation within each area also has to be controlled. This is required to maintain the scheduled power
interchange. So, the main aim of the load frequency control in multi area power systems is to regulate
the frequency to the specified nominal value and to maintain the interchange power between areas at the
scheduled values [5].

2. Modeling of ALFC

2.1 Introduction
The control mechanism needed to maintain the system frequency. The maintaining of system
frequency constant is commonly known as AUTOMATIC LOAD FREQUENCY CONTROL (ALFC). It
has other nomenclatures such as Load Frequency Control, Power Frequency Control, Real Power
Frequency Control and Automatic Generation Control. To maintain the desired megawatt output power
of a generator matching with the changing load and it assist in controlling the frequency of larger
interconnection. In order to keep the net interchange power between pool members, at the predetermined
values. The ALFC loop will maintain control only during small and slow changes in load and frequency.
It will not provide adequate control during emergency when large megawatt imbalances occur.
Figure 2.1 shows schematically the speed governing system of a steam turbine. The system consists:
speed changer, speed governor, hydraulic amplifier, and control valve.

2.2 Discussion on Speed Governor Model


Fly Ball Speed Governor:
This is the heart of the system, which senses the change in speed (frequency). As the speed
increases, the fly balls move outwards and the point B on linkage mechanism moves downwards. The
reverse happens when the speed decreases.
Hydraulic Amplifier:
It comprises a pilot valve and main piston arrangement. Low power level pilot valve movement is
converted into high power level piston valve movement. This is necessary in order to open or close the
steam valve against high-pressure steam.
Linkage Mechanism:
ABC is a rigid link pivoted at B and CDE is another rigid link pivoted at D. This link mechanism
provides a movement to the control valve in proportion to change in speed. It also provides a feedback
form the steam valve movement (link 4).
Speed Changer:
It provides a steady state power output setting for the turbine. Its downward movement opens the
upper pilot valve so that more steam is admitted to the turbine under steady conditions (hence more
steady power output). The reverse happens for the upward movement of speed changer.
Their incremental movements are in vertical direction. In reality these movements are measured in
millimeters; however analysis we shall rather express them as power increments expressed in MW or pu
MW. The movements are assumed positive in the directions of arrows. Corresponding to raise
command, linkage movements will be: A moves downwards; C moves upwards; D moves
upwards; E moves downwards. This allows more steam or water flow into the turbine resulting
incremental increase in generator output power. When the speed drops, linkage point B moves
upwards and again generator output power will increase [2].

Fig. 2.1 Turbine Speed Governing System


Speed Governor:
When the electrical load is suddenly increased then the electrical power exceeds the mechanical
power input. Because of this, the deficiency of power in the load side is extracted from the rotating
energy of the turbine. Due to this reason, the kinetic energy of the turbine i.e. the energy stored in the
machine is reduced and the governor sends a signal to supply more volumes of water or steam or gas to
increase the speed of the prime mover to compensate speed deficiency.

Assume that the system is initially operating under steady conditions-i.e., linkage mechanism
stationary and pilot valve closed, steam valve opened by a definite magnitude, turbine running at
constant speed with turbine power output balancing the generator load. The operating conditions be
characterized by f

P G0 , Y E 0 .

where,
f 0= System frequency (speed)
PG0 = generator output = turbine output (neglecting generator loss)

YE

= steam valve setting

We have to obtain a linear incremental model around these operating conditions.


Let the point A on the linkage mechanism be moved downwards by a small amount

Y A

. It is a

command, which causes the turbine power output to change to mathematically, represented as
Y A =K C PC
(2.1)
Where

PC

is the commanded increase in the power.


PC

The command signal

KC

is the turbine constant.

sets into motion a sequence of events-the pilot valve moves upwards, high

pressure oil flows on to the top of the main piston moving it downwards; the steam valve opening
consequently increases, the turbine generator speed increases. i. e. the frequency goes up. We can model
these events mathematically.
The two factors contribute to the movement of C are
l2
Y A
k Y A
k K PC
(i)
contributes - l1 Y A or - 1
(i.e. upwards) of - 1 C
.

()

(ii) Increase in frequency f

causes the fly balls to move outwards so that B moves downwards by a

'
proportional amount k 2 f . The consequent movement of C with A remaining fixed.

Y A =

l1 +l 2 '
k2 f
l1

( )

k2 f

The net movement of C is


Y C =k 1 K C PC +k 2 f
The movement of D (
YC

and

Y E

is the amount by which the pilot valve opens. It is contributed by

and can be written as

Y D=
=
The movement

Y D,

(2.2)

Y D

l4
l
Y C+ 3 Y E
l 4 +l 3
l 4 +l 3

( )

( )

k3 Y C + K 4 Y E

(2.3)

depending upon its sign opens one of the ports of the pilot valve admitting high-

pressure oil into the cylinder thereby moving the main piston and opening the steam valve by
Certain justifiable simplifying assumptions, which can be made at this stage, are

Y E

(i) Inertial reaction forces of main piston and steam valve are negligible compared to the forces exerted
on the piston by high-pressure oil.
Y D
(ii) Because of (i) above the rate of oil admitted to the cylinder is proportional to port opening
.

Y D

The volume of oil admitted to the cylinder is thus proportional to the time integral of
movement

Y E

. The

is obtained by dividing the oil volume by the area of the cross-section of the-piston.

Thus
t

Y E=k 5 ( Y D ) dt

(2.4)

It can be verified from the schematic diagram that a positive movement


(upward) movement

Y E

Y D

accounting for negative sign used in equation (2.4).

Taking Laplace transform of equations (2.2), (2.3) and (2.4), we get


Y C ( s)=k 1 K C PC ( s)+ k 2 F(s)

(2.5)

Y D ( s ) = k 3 Y C ( s)+k 4 Y E (s)

(2.6)

1
Y E ( s )=k 5 Y D (s )
s
Eliminating

YC

, causes negative

(2.7)

(s) and Y D ( s) , we can write

Y E ( s )=

k 1 k 3 K C PC ( s )k 2 k 3 F( s)

( k + ks )
4

PC ( s )

(2.8)
Where
R=
K sg=

k1 K C
=
k2

speed regulation of the governor (in Hz/Mw)

k1 k3 K C
= gain of speed governor
k4

K sg
1
F (s)
R
1+T sg s

](

T sg=

1
=
time constant of speed governor
k4 k5

We are considering hydraulic valve actuator as a part of Governor model. Therefore, the block diagram
of Governor will be as shown in Fig 2.2

Fig 2.2

GH =

where,

K sg
1+T sg s

2.3 Turbine Model


In normal steady state, the turbine power
gap power

Pg

keeps balance with the electromechanical air-

resulting in zero acceleration and a constant speed and frequency. During transient

state, let the change in turbine power be


Pg

PT

PT

and the corresponding change in generator power be

. The accelerating power in turbine generator unit =

increment

PT

PT P g

depends entirely upon the valve power increment

PV

. The turbine power

and the characteristic of

the turbine. Different type of turbines will have different characteristics. Taking transfer function with
single time constant for the turbine, we can write
PT ( s )=GT PV ( s )=

Where, the turbine time constant

TT

1
PV (s)
1+T T s

is in the range of 0.2 to 2.0 seconds.

The generator power increment


PD

Pg

depends entirely upon the change

PD

in the load

being fed from the generator. The generator always adjusts its output so as to meet the demand

changes

PD

. These adjustments are essentially instantaneous, certainly in comparison with the slow

PT

. The help of previous three equations updates the block diagram developed updated as

changes in

shown in Fig. 2.2. This corresponds to the linear model of primary ALFC loop excluding the power
system response.

Fig. 2.3

2.4 Power System Model


We observed earlier that the loop in Fig. 2.3 is open. We now proceed to close it by finding a
PT
mathematical link between
and f. As our generator is supplying power to a conglomeration of
loads in its service area, it is necessary in our following analysis to make reasonable assumptions about
the lumped area behavior. We make these assumptions: The system is originally running in its normal
state with complete power balance, that is,

Pg =PD +losses . The frequency is at normal value

All rotating equipment represents a total kinetic energy of


load, load demand increases by

PD

W 0kin MW sec. By connecting additional

which we shall refer to as new load. (If load demand is

decreased, new load is negative). Then, generation immediately increases by


load, that is

P g= P D

f0 .

Pg

to match the new

. It will take some time for the control valve in the speed governing system to

act and increase the turbine power. Until the next steady state is reached, the increase in turbine power
Pg
will not be equal to
. Thus, there will be power imbalance in the area that equals i.e.
PT P g

. As a result, the speed and frequency change. This change will be assumed uniform

throughout the area. The above said power imbalance gets absorbed in two ways.(i) By the change in the
total kinetic K.E.(ii) By the change in the load, due to change in frequency.
f 2
W kin=W 0kin ( 0 ) MW Sec.
f
We know that load demand change (
1. Increased generation

Pg

PD

) is met by two changes in the system,

due to opening of steam conditions.

dP D
2. Load decrement due to drop in system frequency ( D= df ).
The old load is a function of voltage magnitude and frequency. Frequency dependency of load can be
written as
dP
D= D
df
PT P D=

d ( W kin )
+D f
dt

(2.9)

By solving kinetic energy equation


d ( W kin ) 2 W 0kin d ( f )
= 0
dt
dt
f
Substituting above value in equation (2.9)
0
2 W kin d ( f )
PT P D= 0
+D f
dt
f
By dividing this equation by the generator rating
H=

Pr

and by introducing per-unit inertia constant

W 0kin
Pr

Where,

is per unit inertia constant,

Pr

is generator rating.

It takes on the form


PT P D=

2 H d ( f )
+D f
dt
f0

pu MW

Laplace transformation of the above equation yields


2H
PT ( s ) PD ( s ) =( 0 s + D) f (s)
f

PT P D
f ( s )=GP ( s ) ]

(2.10)

where,
G P ( s )=

K P=

1
2H
s+ D
0
f

1
D

KP
1+ s T P

T P=

2H
0
f D

Equation (2.10) represents the missing link in the control loop of Fig. 2.3. By adding this, block diagram
of the primary ALFC loop is obtained as shown in Fig. 2.4

P g ( s ) = PT ( s ) PD (s )
X E ( s ) - K f ( s )
K sg
t
1+T sg s
1+T t s

+
PC ( s )

1
R

K ps
1+T ps s
+

Fig 2.4. block diagram of load frequency control (isolated power system)
The transfer function of ALFC system can be calculated by taking
of the system and assuming

PD

PC

as input and f

as output

zero (means no load disturbance).

Therefore, the transfer function of system will be given by


f (s)
=
PC (s )

K sg K t K ps

( 1+T gs s )( 1+ T t s ) ( 1+T Ps s ) +

K sg K t K ps
R

2.5 State Space Model for single Area System


From the above Fig. 2.4, writing equations and rearranging them to obtain the state space
equations of single area as under.
d ( f ) 1
=
[ f + K Ps PGK Ps PD ]
dt
T Ps
(2.11)
d ( X E ) 1
f
=
[ X E + Pc
]
dt
T sg
R
(2.12)
d ( PG ) 1
PG + X E
=
]
dt
Tt [
(2.13)
Where,

f =

Change in frequency

X E =

Change in steam valve position

pG =

Change in generator output power

From above three equations the state space matrix is obtained as under

[ ][ ] [ ] [ ]

1
T sg
XE
x 1
1
p G = x 2 = T
t
f
x 3
0

[ ][]

1
Tt
K Ps
T Ps

1
RT sg
0

1
T Ps

1
0
x1
0
T
x 2 + sg u+ K d
0
Ps
x3
0
T Ps

Pc
In the above matrix u is system input (
) and d

is the disturbance (

PD

) [6].

3. State Space and Multirate Output Sampling


3.1 Introduction to state space
A state-space model is just a structured form or representation of the differential equations for a
system where inputs, outputs and states variables are expressed as vectors. The state-space
representation (also known as the "time-domain approach") provides a convenient and compact way to
model and analyze systems with multiple inputs and outputs. Hence, the use of the state-space
representation is more convenient to systems with multi-input and multi-outputs.
The conventional control theory is completely based on the frequency domain approach while
the modern control system approach is mostly based on time domain approach. Most of these systems
are complex hence it has multiple inputs and multiple outputs. In the modern theory of control system

the stability analysis and time response analysis can be done by analytically method very easily. Now
state space analysis of control system is based on the modern theory which is applicable to all types of
systems like single-input single-output systems (SISO), multiple-input and multiple-output systems
(MIMO), linear and non-linear systems, time-varying and time-invariant systems [3]. Let us consider
few basic terms related to state space analysis of modern theory of control systems.
State space model is a representation of the dynamics of an
differential equation in an

nth

order system as a first order

-vector, which is called the state. It converts the

equation that governs the dynamics into

th

order differential

n first-order differential equations. In a state space

representation, the equation having state variables is called the state equation. The system output is
given in terms of a combination of the current system state, and the current system input, through
the output equation. These two equations form a system of equations known collectively as state-space
equations.
Central to the state-space notation is the idea of a state. A state of a system is the current value of
internal elements of the system, that influence the system behavior completely. In essence, the state of a
system is an explicit account of the values of the internal system components. State variables refers to
t=t 0
t t0
smallest set of variables whose knowledge at
together with the knowledge of input,
gives the complete knowledge of the behavior of the system at any time
defined by x 1(t ) ,

t t0

. State variables are

x 2(t) ......... x n (t) . The internal state variables are the smallest possible subset

of system variables that can represent the entire state of the system at any given time. The minimum
number of state variables required to represent a given system, n , is usually equal to the order of the
system's defining differential equation. If the system is represented in transfer function form, the
minimum number of state variables is equal to the order of the transfer function's denominator after it
has been reduced to a proper fraction. State variables must be linearly independent; that is, no state
variable can be written as a linear combination of the other state variables, or else we would not have
enough information to solve for all other state variables [4].
Suppose there is a requirement of n state variables in order to describe the complete behavior
of the given system, then these n state variables are considered to be n components of a vector

x (t ).

Such a vector is known as state vector. State space refers to the n-dimensional space which has
axis,

x2

axis......

xn

x1

axis. The state space is the vector space that consists of all the possible internal

states of the system. State-space models are useful in many situations: such as Linearization of nonlinear models ,Calculation of time-responses both analytically and numerically, Using simulation

tools: MATLAB, LabVIEW, Octave, and Scribal have simulation functions that assumes state-space
models, Analysis of dynamic systems, e.g. stability analysis, Analysis and design of advanced
controllers and estimators Controllability and observability analysis; Design of LQ optimal controllers,
Model-based predictive control, and Feedback linearization control; Design of state estimators.
3.1.1 General state-space model
In a state-space system representation, we have a system of two equations: an equation for
determining the state of the system, and another equation for determining the output of the system. We
will use the variable y (t) as the output of the system, x (t) as the state of the system, and u(t)
as the input of the system. We use the notation

x (t)

for the first derivative of the state vector of the

system, as dependent on the current state of the system and the current input. The state equation shows
the relationship between the system's current state and its input, and the future state of the system.
The output equation shows the relationship between the system state and its input, and the output. These
equations show that in a given system, the current output is dependent on the current input and the
current state. The future state is also dependent on the current state and the current input The most
general state-space representation of a linear system with p inputs, q outputs and n state
variables is given by the following two equations [7]x (t)= Ax( t)+Bu( t)
y (t)=Cx(t)+ Du(t )
The first equation is called the state equation; the second equation is called the output equation. The
vectors,

(t) R

u(t) R

,and

y (t) R

are state, input and output of the system

respectively. The constant matrices, A, B, and C, and D are n n , n p , q n , q p respectively.


is the system matrix, and relates how the current state affects the state change x .
If the state change is not dependent on the current state, A will be the zero matrix. The exponential of
Matrix

At
the state matrix, e

is called the state transition matrix. Matrix B

is the control matrix, and

determines how the system input affects the state change. If the state change is not dependent on the
system input, then B will be the zero matrix. Matrix C is the output matrix, and determines the
relationship between the system state and the system output. Matrix D is the feed-forward matrix,
and allows for the system input to affect the system output directly. A basic feedback system like those
we have previously considered do not have a feed-forward element, and therefore for most of the
systems we have already considered, the D matrix is the zero matrix.

Fig 3.1 Block diagram representation of the linear state-space


equations

3.2 Solution of Continuous Time State Space Model


Continuous time system with state space model given by
x ( t )= Ax ( t )+ B u(t )

(3.1)
y (t )=Cx (t)

(3.2)
where,
x (t 0)

u(t)

x (t )

is the control vector,

is the state vector,

y (t )

is the measurements vector and

is the initial value of the state vector, which usually is assumed to be known . A

matrix, B

is the input matrix, C

is the state

is the output matrix.

At
Obtaining solutions of this state space equation-multiplying both sides of the equation (3.1) by e

e At x ( t ) eAt Ax ( t )=e At B u(t )


e At x ( t ) eAt Ax ( t )=

since,

d At
{e x ( t ) }
dt

(3.3)
(3.4)

Using equation (3.4), equation (3.3) can be rewritten as,


t

dtd {e At x ( t ) } d= e A B u( ) d
t
t
0

Here

t 0, t

are time interval during which we are interested to find the response of the system.

Substituting the values of the limits on the left side of the equation
t

eAt x ( t )| t = e A B u( ) d
t0 t
0

Obtaining

e At x ( t )

in terms of

x (t 0)

and input u()


t

e At x ( t ) =eA t Ax ( t 0 ) + eA Bu ( ) d
0

t0

Now removing

At

from left half of the equation and obtaining expression for

x(t )

in terms of

x (t 0)
and u()
t

x ( t )=e

A (tt 0 )

x ( t 0 ) + e A (t ) B u( )d

(3.5)

t0

This is the solution of the state space equation. As we see, the solution consists of two parts. The
first part represents the autonomous response (homogenous solution) driven only by initial values
different from zero. The second term represents the in homogenous solution driven by the control
variable, u(t) . In order to compute the first term, we have to compute the matrix exponential
e

A (t t 0)

. This matrix exponential is defined as the transition matrix, because it defines the transition of

the state from the initial value,

x (t 0)

x ( t )= Ax ( t ) with known initial state

, to the internal state


x (t 0)

x (t )

in an autonomous system

3.3 Importance of Discrete Time Controller


Digital controllers operate only on numbers. Decision-making is one of their important functions.
In most of the control system, it is not only to stabilize the system but also involved in the optimal
overall operation of industrial plants. Digital controllers are extremely versatile, they can handle
nonlinear control equation involving complicated computation or logic operations. A very much wider
class of control laws can be used in digital controllers. Also, in the digital controller, by merely issuing a
new program the operations being performed can be changed completely. Digital controllers are capable
of performing complex computational accuracy at relatively little increase in cost at present because of
inexpensive microcomputers, digital controllers are being used in many large and small-scale control
systems.
In the digital controllers digital components, such as sample-and-hold circuits, analog to digital
(A/D) and digital to analog (D/A) converters, and digital transducers, are rugged in construction, highly
reliable, and often compact and lightweight. Moreover, digital components have high sensitivity, are
often cheaper and are less sensitive to noise signals and, digital controllers are flexible in allowing
programming changes

3.4 Discretization of the continuous time system


Consider the continuous time state space system given in the equations (3.1) and (3.2).
Now apply an input that changes only at discrete (equal) sampling intervals. It would be nice if we could
G
H ,
find matrices and
independent of t and k so that we could obtain a discrete time
model of the system,
x ( ( k +1 ) )=G x ( k ) + H u ( k )

(3.6)

y ( k )=Cx ( k )+ Du (k)

(3.7)

We will now determine the values of the matrices

and

. It will turn out that while they are

constant for a particular sampling interval, they depend on the value of the sampling interval, so for that
G
H
reason we have written them as and
in (3.6) above.
t o=0

Substituting,

and t=( k +1 )

in equation (3.5) we get

(k +1)

x ( (k +1) )=e

A (k+1)

x ( 0) +

e A {( k+1) t } Bu ( t ) dt

(3.8)

At

t o=0, t =k ,

we have
k

x ( k )=e

Ak

x (0)+ e

Ak

e At Bu ( t ) dt

(3.9)

In order to write
e

A ( K +1 )

x ( ( K +1 ) ) in terms of ( k ) , multiply all terms of (3.9) by e A and solve for

x ( 0 ) , obtaining
k

A ( K +1)

x ( 0 ) =e x ( k )e

A ( K +1 )

e At Bu ( t ) dt

(3.10)

A ( K +1)
x ( 0)
Substituting for e

in (3.5), we obtain

(k+1 )

x ( ( k +1 ) )=e A x ( k )+ e A (K +1 )

e At Bu ( t ) dt eAt Bu ( t ) dt

Which, by linearity of integration, is equivalent to

(3.11)

( k+1 )
A

x ( ( k +1 ) )=e x ( k )+ e

A ( K +1 )

e At Bu ( t ) dt

(3.12)

Next, we notice that within the interval from k

to (k +1) , u ( )=u ( k ) is constant, as is the

matrix B , so we can take them out of the integral to obtain


( k+1 )
A

x ( ( k +1 ) )=e x ( k )+ e

A ( K +1 )

e At Bu ( t ) dt

t [ k ,(k +1) ]

(3.13)

( k+1 )

x ( ( k +1 ) )=e x ( k )+

e A { (k+1 ) t } Bu ( t ) dt

(3.14)

Now we see that as

ranges from

the exponent of e ranges from


t

. Then d

= d t

and

(k +1)

to

(the lower to the upper limit of integration)

to 0. Accordingly, lets define a new variable

ranges from

to 0 as

ranges from

(k +1) -

to

(k +1) .

Thus we have

x ( ( k +1 ) )=e x ( k )+ e A Bu ( k ) d
A

(3.15)

We see that in (3.15) we have written the state update equation exactly in the form of (3.6) where
A
A
1
G =e ;
H =(e I ) A B
(3.16)
K

k1

K j1

x ( k )=( G ( ) ) x ( 0 ) + ( G ( ) )
j=0

and we can see that at the sampling instants

H ( ) u ( j ) ,k =1,2,3

(3.17)

k , this has exactly the same value as is obtained using

(3.1). Specifically,
k

K
( G ( ) ) =( e A ) =e Ak

(3.18)

These equations and derivation will follow when input and output sampling period are same (i.e.
sampling period= ).
3.5 Numerical Examples
In this section two different examples are proposed to illustrate the validity of discretization of
continuous time system.

3.5.1 Example 1
Let us now apply the results developed in the previous sections for discretization of the continuous
time system. Consider a continuous time system represented in state-space model as
x = A x + B u
where,

1 0
0 3

[]
1
1

and the input is step signal.

If the system is sampled with a sampling time =0.1 sec, we obtain following descretized model

x ( k + 1 )=G x ( k ) + H u ( k )
where,

G =

0.905
0
0
0.741

H =

[ ]
0.095
0.086

1
x 1(t)

0.6
0.4
0.2

0.6
0.4
0.2

50

time,t

100

150

0.4

0.4

0.3

0.3
x 2(t)

0.2

amplitude

amplitude

0.1
0

x 1(k)

0.8

amplitude

amplitude

0.8

20
40
sampling time,k

60

x 2(k)

0.2
0.1

50

time,t

100

150

20
40
sampling time,k

60

Fig.3.2 State response of continuous time system and discrete time system.
3.5.2 Example 2
Discretization of continuous system when input is exponential without disturbance
Let us now apply the results of previous sections to discretize a continuous time system system
with a disturbance. Consider a continuous time system represented by

Ax

Bu

where,

A= 1 2
1 1

[]

B= 1
2

u=e0.1t

If the system is sampled with a sampling time =0.1 sec, we obtain following descretized model
x ( k + 1 )=G x ( k ) + H u(k )
where,

G = 1.094 0.220
.110 1.094

H = 0.1262
0.2043

60

150
x 1(t)

100
amplitude

amplitude

40
20
0
-20

100

200
300
time,t

400

-50

500

20
40
sampling time,k

60

150
x 2(t)

x 2(k)

100
amplitude

20
amplitude

50
0

30

10
0
-10

x 1(k)

50
0

100

200
300
time,t

400

500

-50

20
40
sampling time,k

60

Fig.3.4 State response of continuous time system and discrete time system.

3.5 Multirate output sampling


A multirate sampling (MR) system is defined as a hybrid system composed of continuous time
elements, usually the plant, and some discrete time components, usually the controllers or the filters,
where two or more variables are sampled or updated at different frequencies. It can be also considered

that the discrete actions are not equally spaced on time and/or delayed. Moreover, in a great number of
computer control applications the approximation of a regular pattern of sampled signals is assumed. In
multirate sampling, the output is sampled more rapidly than the input ie the sampling frequency of the
output is more which can be written like this:
=
where,
Value of
For

is the input sampling period and is the output-sampling period.


N

is always greater than 1 for multirate sampling (integer value i.e.- 2, 3, 4.).

=1 both sampling periods are same, it is called single rate sampling.

As we know
y ( k )=Cx (k )
Then

(3.19)

y ( k + )=Cx( k +)

In general, this equation can be written as


y ( k +(N 1) )=Cx ( k+( N 1) )
But we know that during the period

t=k

to

(3.20)
t=(k +1) value of

x (t )

equal to the sampled

value i.e
x ( t )=x (k )
Now equation 2 can be written as
x ( k + ) =e A x ( k )+(e A I ) A1 B u(k )
x ( k + ) =G x ( k ) + H u(k )
where,

G=e A

Now putting the value of

H =(e A I ) A1 B

x ( k + )

in output equation (3.19)

y ( k + )=C G x ( k ) +C H u(k)
x ( k + 2 )=G x ( k + )+ H u (k + )
since,

(3.21)

u ( k + )=u( k)

Putting the values from equation, this equation becomes

(3.22)

G H +
2
x ( k + 2 )=G x ( k ) +

H u(k )

In general
x ( k+(N 1) )=G N1 x ( k ) +

N 2

Gi H u(k )

i=0

and output equation is given by


y ( k +(N 1) )=C G N 1 x ( k ) +C
With input sampling period

N 2
i=0

G i H u(k )

and output sampling period

, we obtain following discrete-time

system
x ( ( k +1 ) )=G x ( k ) + H u(k)

(3.23)

y k+1 =C0 x ( k )+ D 0 u( k )

where,

y (k )
y (k + )
y k+1 =

y (k + ( N 1 ) )

(3.24)

]
(3.25)

[ ]

C
C G
C0 =

C G N 1

The matrices of

system and

G =GN

D 0=

[ ]
0
C H

N 2

C G H
i

i=0

system have the following relation

4. Controller Design
Constant frequency electricity is very important and necessary now-a-days. Load frequency
control (LFC) is an important tool to insure the stability and reliability of power systems. The goals of
the LFC are to maintain zero steady state errors in a multi-area interconnected power system and to
fulfill the requested dispatch conditions. Now-a-days power systems are very complex and
interconnected. Because of this, to maintain frequency constant, controller is required. So appropriate
control should be designed for the concerned power system.

4.1 Discretization of Continuous Time System with Disturbance


A general continuous-time linear system with added disturbance is described with the following
equations:
x ( t )= Ax ( t )+ Bu ( t ) + Dd (t)
(4.1)
y (t )=Cx (t)
(4.2)

Let us assume that control signal u from (4.1) is able to change its value only every
where

seconds,

is a sampling period.

In order to design discrete-time estimator, system (4.1) is discretized using the Zero-Order-Hold (ZOH)
discretization method, with sampling period . That results in the following discrete-time system:
x ( ( k +1 ) )=G x ( k ) + H u ( k ) +W d (k )

(4.3)

y ( k )=Cx (k )

(4.4)

Consider the system given by the equation (4.1) and (4.2).


G
H
We will now determine the values of the matrices
and
. It will turn out that while they are
constant for a particular sampling interval, they depend on the value of the sampling interval, so for that
G
H
reason we have written them as and
in (4.2) above.
Substituting,

t o=k

and t=( k +1)

in equation (3.5) we get


(k +1)

x ( (k +1) )=e

A (k+1)

x (0)+

e A {( k+1) t } Bu ( t ) dt

( k +1)

e A {(k +1) t } Dd ( t ) dt

At

t o=0, t=k ,

we have
k

x ( k )=e

Ak

x (0)+ e

Ak

At

Bu ( t ) dt+ e

In order to write
solve for e

A ( K +1)

x ( ( K +1 ) ) in terms of

Ak

e At Dd ( t ) dt
0

( k ) , multiply all terms of above equation by e A

x ( 0 ) , obtaining
k

e A ( K +1) x ( 0 ) =e A x ( k )e A ( K +1) e At Bu ( t ) dte A ( K +1) e At Dd ( t ) dt


0

and

Substituting for e

A ( K +1)

x ( 0)

in (3.5), we obtain

(k+1 )

x ( ( k +1 ) )=e x ( k )+ e

A ( K +1 )

k
At

+e

A ( K +1)

Bu ( t ) dt eAt Bu ( t ) dt
0

(k +1)

k
At

Dd ( t ) dt eAt Dd ( t ) dt
0

Which, by linearity of integration, is equivalent to


( k+1 )
A

x ( ( k +1 ) )=e x ( k )+ e

A ( K +1 )

e At Bu ( t ) dt

( k +1)

+e

A ( K +1)

eAt Dd ( t ) dt

u(t)=u(k )

Through zero-order hold (ZOH),

(k +1) , where

d (t)=d (k )

and

over the time interval

to

B , is constant so we can take them out of the

is sampling period. The matrix

integral to obtain
( k+1)

x ( ( k +1 ) )=e A x ( k ) + e A ( K +1 )

eAt Bu ( t ) dt

( k +1)

+e

A ( K +1)

eAt Dd ( t ) dt

Now we see that as

the exponent of e ranges from


t

. Then d

ranges from

= d t

and

to

(k +1)

[ k ,(k +1) ]

(the lower to the upper limit of integration)

to 0. Accordingly, lets define a new variable

ranges from

to 0 as

ranges from

(k +1) -

to

(k +1) .

Thus we have

x ( ( k +1 ) )=e x ( k ) + e Bu ( k ) d+ e A Dd ( k ) d ;
0
0
A

(4.5)

We see that in (4.5) we have written the state update equation exactly in the form of (4.3), where

G =e

H =(e I) A B

W =( e I ) A D

(4.6)
These matrices can also be written as
G =e A ,

H = e At Bdt
0

(4.7)

W = e At Ddt
0

4.1.1 Example of Discretization of the System With Disturbance


Let us now apply the results of previous sections to discretize a continuous time system system with a
disturbance. Consider a continuous-time system with disturbance is represented by
x = Ax+ B u+ Dd

where,

A=

1 0
0 3

B=

[]
1
1

D=

[]
5
4

If the system is sampled with a sampling time =0.1 sec, we obtain following descretized model
x ( k + 1 )=G x ( k ) + H u ( k ) +W (k )
where,

0 ,
G = 0.905
0
0.741

H = 0.0952
0.0864

[ ]

W = 0.476
0.346

6
5

x 1(t)

amplitude

amplitude

3
2
1

4
3
2
1

50

time,t

100

150

1.5

1.5
x 1(t)

0.5

amplitude

amplitude

x 1(k)

20
40
sampling time, k

60

x 2(k)

0.5

50

time,t

100

150

20
40
sampling time, k

60

Fig.3.3 State response of continuous time system and discrete time system
As we can analyse from the plots that the states response of both continuous-time system and
discrete-time system are almost same so we can say that discretization does not change the states hence
we can discretise the continuous time system to design controller.

4.2 Fast Output Sampling Method


Fast output sampling (FOS) is an estimation technique appropriate for continuous time system
controlled with discrete-time control signal, where the output signal can be sampled several times during

one period of the control signal [10]. FOS shows better performance than standard estimation
techniques, because it reduces the estimation error to zero after just one sampling period [11]. Standard
estimators need at least v sampling periods to achieve errorless estimation, where v is the
observability index of the system [12]. To use FOS estimation technique, it must be satisfied

N >v

[10].

Fig.4.1 The usage of the FOS estimation method in system control.


The principle of using FOS estimation technique in system control is shown in Fig. 4.1. Firstly,
the last N subsamples of the output signal y (t), measured in the most recent sampling period
, are used to estimate the system state. Then, that estimated state is used to compute the control signal
for the next sampling period.
x ( ( k +1 ) )=G x ( k ) + H u ( k )+W d ( k )
y ( k )=Cx (k )

(4.8)
(4.9)

Let us also assume that only system output is measurable, and only at certain time instances, y (k ) ,
where

is a subsampling period:
=

(4.10)
where

N N . Those samples can be used as input signals of the appropriate estimator for

unmeasured state and disturbance signals in (4.1). LFC applied nowadays in real power systems is an

example of a system with multiple sampling periods. In LFC, control signal is sent to the power plants in
discrete-time. In UCTE interconnection that period is 15 s [9]. Additionally, during one sampling
period several measurements of frequency f (k ) and tie-line power Ptie (k ) signals are
gathered. Besides those subsamples, which are inputs to classical PI controller, subsamples of generated
power Pg ( k ) are also gathered for monitoring purposes. Those samples could also be used as
inputs to the estimator. Because a substitute power plant is used in modeling a CA and also in controller
synthesis, all other state and disturbance signals, that cannot be measured in the real system, must
therefore be estimated.
The systems N consecutive subsamples, taken during the sampling period , can now be
calculated as:
y ( k )=Cx (k )

Then

y ( k + )=Cx(k +)

In general

y ( k +(N 1) )=Cx ( k +(N 1) )

But we know that during the period t=k

to

x (t )

t=( k +1) value of

equal to the sampled

value i.e
x ( t )=x (k )
Now equation (4.5) can be written as
x ( k + ) =e A x ( k )+ ( e A I ) A1 Bu ( k )+ ( e A I ) A1 Dd ( k )
x ( k + ) =G x ( k )+ H u ( k ) +W d ( k )
G=e A

where,

H =(e A I ) A1 B

(4.11)
Now putting the value of

x ( k + )

in output equation

y ( k + )=C G x ( k ) +C H u ( k )+C W d (k )
Now
since,

x ( k + 2 )=G x ( k + ) + H u ( k + ) +W d ( k + )
u ( k + )=u(k )

Putting the values from equation, this equation becomes

W =(e A I ) A1 D

G H +
2
x ( k + 2 )=G x ( k ) +

In general

x ( k +( N 1) )=G

N1

G W +
H u ( k ) +

N2

x ( k )+

i=0

W d (k )

Gi [ H u ( k ) +W d ( k ) ]

(4.12)

And output equation is given by


y ( k +(N 1) )=C GN 1 x ( k ) +C
With input sampling period

N 2
i=0

Gi [ H u ( k ) +W d( k )]

and output sampling period

x ( ( k +1 ) )=G x ( k ) + H u ( k )+W d (k )

(4.14)

y k+1 =C0 x ( k )+ D 0 u ( k )+W 0 d ( k )

where,

(4.13)

y (k )
y (k + )
y k+1 =

y (k + ( N 1 ) )

(4.15)

[ ]

C
C G
C0 =

C G N 1

(4.16)

D 0=

[ ]
0
C H

N 2

C G H
i

i=0

W 0=

[ ]
0
CW

N 2

G =GN

C G i W
i=0

4.3 Design Procedure


4.3.1 State Feedback Controller
The state of a dynamical system is a collection of variables that permits prediction of the future
development of a system. We now explore the idea of designing the dynamics a system through
feedback of the state. State feedback, or pole placement, is a method employed in feedback control
system theory to place the closed-loop poles of a plant in pre-determined locations in the s-plane.
Full state feedback is utilized by commanding the input vector u . Consider an input proportional (in
the matrix sense) to the state vector [13],

u ( t ) =Kx (t)

(4.17)

Substituting into the state space equations above,


x ( t )=(ABK ) x (t)
(4.18)
y (t )=(CDK ) x(t)

The roots of the state feedback system are given by the characteristic
equation, det [ sI(AB K )] . Comparing the terms of this equation with those of the desired
characteristic equation yields the values of the feedback matrix

which force the closed-loop

eigenvalues to the pole locations specified by the desired characteristic equation. This works only for
Single-Input systems. Multiple input systems will have a K matrix that is not unique. Choosing,
therefore, the best K

values is not trivial.

One should note that although state feedback control is very attractive because of precise
computation of the gain matrix K , implementation of a state feedback controller is possible only
when all state variables are directly measurable with help of some kind of sensors. Due to the excess
number of required sensors or unavailability of states for measurement, in most of the practical
situations this requirement is not met. Only a subset of state variables or their combinations may be
available for measurements. Sometimes only output y is available for measurement. Hence the need for
an estimator or observer is obvious which estimates all state variables while observing input and output.
To enable full state feedback we have to estimate the state of the systems [16]. With increase of
complexity and optimization of the performance, the discrete controllers are more suitable. Hence need
to design a discrete state feedback controller. In the next section, the design of discrete time controller is
given.
We consider a continuous time system with added disturbance as
x ( t )= Ax ( t )+ Bu ( t ) + Dd (t)

(4.19)
y (t )=Cx (t)

(4.20)
To design discrete time controller, we need to discretize the above system. Assume sampling time period
seconds. With sampling period , the discretized representation of the above system..
x ( ( k +1 ) )=G x ( k ) + H u ( k )+W d (k )

(4.21)

y ( k )=Cx (k )
where, matrices involved in these equations is given by equation (4.6) and (4.16). The system controlled
by a state feedback controller of the form
u ( k )=Kx (k )
K

Where, the matrix

(4.22)

can be obtained by pole-placement method. The biggest challenge in most of


x (k ) . Following will illustrate the design of controller of

the systems is finding the full state vector


the form (4.22).
N

We choose a suitable value of

such that

N >n . With output sampled at

times

faster than the input, the stacked output will be


y k+1 =C0 x ( k )+ D 0 u ( k )+W 0 d ( k )
Where, matrices

C0

D0

and

y k+1 =[ C0 W 0 ]

W0

(4.23)

are given by (4.16). From (4.23), we have

[ ]

x (k )
+D 0 u ( k )
d (k )

(4.24)

[]
C

Let inverse of matrix

[C0

W 0]

Multiply (4.24) with

[]

and rearranging the equation, we will get

is given by matrix

[ ][]

x(k ) C
= [ y k+1D 0 u ( k ) ]
D
d (k )

(4.25)

By substituting controller (4.22) into (4.21), we get


x ( ( k +1 ) )=G x ( k )H Kx ( k ) +W d (k )
[ G H K ] x ( k ) +W d (k )

(4.26)

This can be rewritten as


x ( ( k +1 ) )=[ (G H K ) W ]

[ ]
x(k )
d (k )

(4.27)

Substituting (4.25) into (4.27), the close loop system will be


C
x ( ( k +1 ) )=[ (G H K ) W ]
y D 0 u ( k ) ]
[ k+1
D

[]

[ ( G H K ) C+W
D ] [ y k+1D 0 u ( k ) ]

(4.28)
From equation (4.28) it is evident that the close loop system is free from the disturbance. Hence, the
controller will nullify the effect of disturbance on the system.

4.3.2 Controller Algorithm


Step 1: Consider the continuous time system of the form (4.19) and
(4.20).
Step 2: Choose sampling period

G ,

Step 3: Obtain discrete time system matrices

and

using (4.7).
K

Step 4: Obtain state feedback matrix

based on pole-placement

design.
Step 5: Choose

N > order of the system (n) .

Step 6: Obtain the matrices

Step 7: Obtain

[]
C

C0 , D 0

, is the inverse of

and

[C0

W0

using (4.16).

W 0] .

Step 8: The state and disturbance vectors are obtained using (4.25).
Step 9: Implement controller using (4.28).

4.3.3 Numerical Example


Example of Discretization of the System With Disturbance by the use of multirate
sampling:

Let us now apply the results of previous sections to discretize a continuous time system system with a
disturbance. Consider a continuous-time system with disturbance is represented by
x = Ax+ B u+ Dd

where,

A= 1 0
0 3

[]

B= 1
1

[]

D= 5
4

If the system is sampled with a sampling time =0.2

sec, we obtain following descretized model

x ( k + 1 )=G x ( k ) + H u ( k ) +W (k )
where,
Now taking

G =

N=2

] [ ]

0.819
0
0.181
, H =
0
0.549
0.150

W=

[ ]
0.906
0.602

and obtaining multirate output sampling system output equation, as per described

in above section, is given by


y k+1 =C0 x ( k )+ D 0 u ( k )+W 0 d ( k )

where,

C0 = 1.0000 1.0000
0.9408 0.7408

D 0=

0
0.1816

W 0=

0
0.8214

comparision of true state and its multirate output estimate

x 1(k)

x 2(k)
MOS of x 1(k)
MOS of x 2(k)

4
3
2
1
0
-1
-2

10

15

20

25

30

35

40

45

50

Fig.4.2 Comparison of true state and its multirate output sampled estimate

Error between true states and its MOS estimate

1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2

10

15

20

25

30

35

Fig.4.3 error between the true states and their MOS estimate

40

45

50

5. Conclusions
From this we can conclude that the discretization of the system does not change its original or
true states.so discretization is valid for controller design purposes as we know that mostly controllers are
of discrete type. Also we dont know the system states in most of the control systems or system may not
be a full system but we need system states for state feedback controller. For this we have used multirate
output sampling estimation of the system states as we have output variable, which we have sampled,
using MOS. From graphs we can also conclude that the state estimation done using MOS is very well
accurate as the error between the true state reduces to zero after just one sampling period and also
estimate state FOS shows better performance than standard estimation technique

Appendix
3.5.1 Discretization of continuous system without disturbances when
input is unit step

clear all
clc
a=[-1 0; 0 -3];
b=[1; 1];
c=[1 0];
d=0;
% for continous time system
sys=ss(a,b,c,d);
[y,t,x]=step(sys,10);
% for discrete time system
[ad, bd]=c2d(a,b,0.1);
m(:,1)=[0;0];
u=1;
for k=1:51
m(:,k+1)=ad*m(:,k)+bd*u;
end
k=1:length(m);
subplot(2,2,1)
plot(x(:,1),'-r')
hold on
subplot(2,2,2)
plot(k,m(1,:),'-b');
hold on
subplot(2,2,3)
plot(x(:,2),'+b');
hold on
subplot(2,2,4)
plot(k,m(2,:),'--g');
hold off;

4.1.1 Discretization of the continuous System With Disturbance when


input is unit step:
clear all
clc
a=[-1 0; 0 -3];

b=[1; 1];
c=[1 0];
d=0;
p=[5;4];
[ad,bd]=c2d(a,b,0.1)
[dummy, dd]=c2d(a,p,0.1);
sys=ss(a,b+p,c,d);
[y,t,x]=step(sys,10);
m(:,1)=[0;0];
u=1;
for k=1:51
m(:,k+1)=ad*m(:,k)+bd*u+dd;
end
k=1:length(m);
subplot(2,2,1);
plot(x(:,1),'--r');
subplot(2,2,2);
plot(k,m(1,:),'+b');
hold on
subplot(2,2,3)
plot(x(:,2),'g');
subplot(2,2,4)
plot(k,m(2,:),'.r');
hold off
4.3.2 Discretization of the System with Disturbance by the use of
multirate sampling:
clear all
clc
%------continuous time system--------a=[-1 0;0 -3]; b=[1;1];c=[1 1];d=[5;4];
%multirate sampling matrices
T=0.2;
[aT bT]=c2d(a,b,T);
[dummy cT]=c2d(a,d,T);
x(:,1)=[-2;0];u=1;g=1;
for k=1:50
x(:,k+1)=aT*x(:,k)+bT*u+cT*g;
end
%---one output is given for second order system--c=[1 1];
y1=c*x(:,1);
%observer design parameters
J_tan=place(aT',c',[0.3;0.4]);

J=J_tan';
xhat(:,1)=[-1;1];
for k1=1:50
xhat(:,k1+1)=(aT-J*c)*xhat(:,k1)+bT*u+J*y1;
y1=c*x(:,k1+1);
end
k=0:length(x)-1;
plot(k,x(1,:),'r:',k,x(2,:),'k-','LineWidth',1.5)
hold on
k1=0:length(xhat)-1;
plot(k1,xhat(1,:),'b.',k1,xhat(2,:),'g-','LineWidth',1.5)
hold off
title('comparision of state and its estimate:stable system')
hold on
legend('x_{1}(k)','x_{2}(k)','estimate of x_{1}(k)','estimate of
x_{2}(k)')
%---plot of error between actual state and estimated state-plot(k,x(1,:)-xhat(1,:),k,x(2,:)-xhat(2,:))
%------Multirate output sampling--------[aD bD]=c2d(a,b,T/2);
[dd cD]=c2d(a,d,T/2);
co=[c;c*aD];do=[0;c*bD];wo=[0;c*cD];
yo(:,1)=[-1;0];
for k2=1:50
xmos(:,k2)=aT*inv(co)*yo(:,k2)+(bT-(aT*inv(co)*do))*u+(cT(aT*inv(co)*wo))*g;
yo(:,k2+1)=co*x(:,k2)+do*u+wo*g;
end
k2=0:length(xmos)-1;
k3=0:length(x)-2;
%---plot to compare true state and its estimate based on MOS--%plot(k2,xmos(1,:),'r:',k2,xmos(2,:),'k',k3,x(1,1:50),'b.',k3,x(2,1:50
),'g+','LineWidth',2)
% title('comparision of true state and its multirate output estimate')
%legend('x_{1}(k)','x_{2}(k)','MOS of x_{1}(k)','MOS of x_{2}(k)')
% ---Error between true state and its MOS estimate---plot(k2,xmos(1,:)-x(1,1:50),'r:',k2,xmos(2,:)x(2,1:50),'k:','LineWidth',2)
title('Error between true states and its MOS estimate')

References
1. Dr. B.R. GUPTA,Power System Analysis and Design, Sixth Edition, S.Chand & Company Ltd.
2. D.P. Kothari, Modern Power System Analysis,3rd Edition Tata McGraw Hill, India.
3. B.C. Kuo, Automatic Control Systems, 3rd Edition Prentice-Hall Inc., New Jersey.
4. Norman S. Nise,Control System Engineering, Fourth Edition, John Wiley & Sons, Inc.
5. Swati Sondhi, Yogesh V. Hote, Fractional order PID controller for load frequency control ,
Science Direct Journal of Energy Conversion and Management Volume 85, September 2014,
Pages 343353.
6. Nilaykumar N. Shah, Chetan D. Kotwal, The State Space Modeling of Single, Two and Three
ALFC of Power System Using Integral Control and Optimal LQR Control Method, IOSR
Journal of Engineering, Mar. 2012, Vol. 2(3) pp: 501-510.
7. Brogan, William L. (1974). Modern Control Theory (1st ed.). Quantum Publishers, Inc. p. 172.
8. Elgerd Ol. Electric energy systems theory- an introduction, 2nd ed.Tata McGrawHill:2000.
9. Y. Rebours, D. Kirschen, M. Trotignon, S. Rossignol, A survey of frequency and voltage control
ancillary servicesPart I: Technical features, IEEE Transactions on Power Systems 22 (1)
(2007) 350357.
10. M. Saaj, B. Bandyopadhyay, H. Unbehauen, A new algorithm for discrete-time sliding-mode
control using fast output sampling feedback, IEEE Transactions on Industrial Electronics 49 (3)
(2002) 518523.
11. S. Janardhanan, Multirate output feedback based discrete-time sliding mode control strategies,
Ph.D. thesis, Indian Institute of Technology, Bombay, India, 2005.
12. V. Bandal, Power system stabilizer design based on multirate output feedback sliding mode
control strategies, Ph.D. thesis, Indian Institute of Technology, Bombay, India, 2005.
13. S. Mishra, G. Mallesham, and P. Shekhar, Biography based optimal state feedback controller for
frequency regulation of a smart microgrid, IEEE Transactions on S mart Grid, Vol. 4, no. 1, pp.
628-637,2013.
14. O. Elgerd and C. Fosha, Optimum megawatt frequency control of multi-area electric energy
systems, IEEE Trans Power Appl. Syst., Vol.89, no. 4, pp 556-563, 1970.

15. Ibraheem, P. Kumar, and D.P.Kothari,Recent philosophies of automatic generation control


strategies in power systems, IEEE Trans Power Syst., Vol.20, no.1, pp. 346-357, 2005.
16. T. Hagiwara ; Dept. of Electr. Eng., Kyoto Univ., Japan; M. Araki, Design of a stable state
feedback controller based on the multirate sampling of the plant output, IEEE Transactions on
Automatic Control, Vol. 33, no. 9, pp 812-819, 2002.
17. H. Warner, Fast output sampling regulators with integral action, European Journal of Control 7
(6) (2001) 618-624.
18. Luenberger D.G. 1967. Canonical forms for linear multivariable systems, IEEE Transactions
on Automatic Control, 12(3):290-293.

Вам также может понравиться