Aadilllllllllllllllllllllllllaaaa

Application of Supervisory Adaptive Predictive Control Scheme in
Electrical Power Systems
Abstract
Fast acting energy storage devices can effectively damp electromechanical oscillations in
a power system, because they provide storage capacity in addition to the kinetic energy of the
generator rotor, which can share the sudden changes in power requirement. The present project
shows the effectiveness of energy storage device Supercapacitor Energy Storage System
(SCESS) unit for this application and suggests means of best utilization of the small energy
storage capacity of such unit to improve the load frequency dynamics of large power areas. The
power station considered here is a thermal power system of reheat type
Supercapacitors, also known as ultra capacitors or electrochemical capacitors, utilize high
surface area electrode materials and thin electrolytic dielectrics to achieve capacitances several
orders of magnitude larger than conventional capacitors. In doing so, supercapacitors are able to
attain greater energy densities while still maintaining the characteristic high power density of
conventional capacitors.
This project focuses on the development of a supervisory control scheme for improved
and profitable operation of a small-rating supercapacitor energy storage system (SCESS) in load-
frequency-control applications. The proposed scheme that is based on adaptive generalized
predictive control (AGPC) effectively handles the operational constraints of the SCESS while
generating an appropriate power reference for the primary (inner) control loop of SCESS. The
primary control loop in turn reacts by forcing the SCESS to follow the power command received
from the supervisory (outer) control loop in response to a load perturbation. The conventional
tracking controller associated with the inner control loop is tuned offline by employing a genetic
algorithm. The overall primary control loop is designed to behave similar to a reference system
whose model is utilized for calculating dynamic inequality constraints in the optimization
process of AGPC. Various modelling and design aspects are addressed, and simulation
experiments are finally carried out on a two area power system to demonstrate the effectiveness
of the proposed scheme.
Department Of Electrical Engineering Page 1

Chapter 1
Load Frequency Control Problem
The flows of active and reactive power in a transmission network are fairly independent of each
other and are influenced by different control actions. Hence, they may be studied separately for a
large class of problems. Active power control is closely related to frequency control, and reactive
power control is closely related to voltage control. As constancy of frequency and voltage are
important factors in determining the quality of power supplied, the control of active and reactive
power is vital to the satisfactory performance of power systems.
1.1 Active Power and Frequency Control

For satisfactory operation of a power system, the frequency should remain nearly constant.
Relatively close control of frequency ensures constancy of speed of induction and synchronous
motors. Constancy of speed of motor drives is particularly important for satisfactory
performance of generating units as they are highly dependent on the performance of all the
auxiliary drives associated with the fuel, the feed-water and the combustion air supply systems.
In a network, considerable drop in frequency could result in high magnetizing currents in
induction motors and transformers. The extensive use of electric clocks and the use of frequency
for other timing purposes require accurate maintenance of synchronous time which is
proportional to integral of frequency. As a consequence, it is necessary to regulate not only the
frequency itself but also its integral.
The frequency of a system is dependent on active power balance. As frequency is a common
factor throughout the system, a change in active power demand at one point is reflected
throughout the system by a change in frequency. Because there are many generators supplying
power into the system, some means must be provided to allocate change in demand to the
generators. A speed governor on each generating unit provides the primary speed control
function, while supplementary control originating at a central control centre allocates generation.
In an interconnected system with two or more independently controlled areas in addition to
control of frequency, the generation within each area has to be controlled so as to maintain
scheduled power interchange. The control of generation and frequency is commonly referred to
as load-frequency control (LFC).

Figure 1.1. AGC and AVR

1.2 Generator Response to Load Change
The basic concepts of speed governing are best illustrated by considering an isolated generating
unit supplying a local load as shown in Figure 1.2.
Figure 1.2. Generator Supplying Isolated Load
When there is a load change, it is reflected instantaneously as a change in the electrical torque
output Te of the generator. This causes a mismatch between the mechanical torque Tm and the
electrical torque Te which in turn results in speed variations as determined by the equation of
motion which leads to variations in system frequency.
1.3 Speed Governing System
Figure 1.3 shows schematically the speed governing system of a steam turbine. The system
consists of the following components:
a) 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
the linkage mechanism moves downwards and vice-versa.
b) Hydraulic amplifier : It comprises of a pilot valve and a 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.

c) 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 the change in speed. It also provides a feedback from the steam valve
movement (link 4).
d) 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-state conditions. The reverse happens for upward movement of the
speed changer.
Figure 1.3. Turbine Speed Governing System

1.3.1 Model of Speed Governing System
Assume that the system is initially operating under steady conditions-the 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. Let the operating
conditions be characterized by:
0
f = system frequency
0
PG = generator output = turbine output (neglecting generator loss)
y 0E = steam valve setting
We shall obtain a linear incremental model around these operating conditions. Let the point A on
the linkage mechanism be moved downwards by a small amount We shall obtain a linear
incremental model around these operating conditions. Let the point A on the linkage mechanism
y
be moved downwards by a small amount E . It is a command which causes the turbine
power output to change and can therefore be written as. It is a command which causes the turbine
power output to change and can therefore be written as
y A =K C PC

Where Pc is the commanded increase in power.
The command signal Pc sets into motion a sequence of events the pilot valve moves upwards,
high pressure oil flows on the top of the main piston moving it downwards; the steam valve
opening consequently increases, the turbine generator speed increases, that is, the frequency goes
up.
Two factors contribute to the motion of C:
l
a)
yA
contributes ( 2 ) ) y A or -
K1

yA
(i.e., upwards) or -
K1 KC

l1
PC
.

b) Increase in 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 at
l 1 +l 2
is
yA
is ( K '2 f = K 2 f (i.e., downwards).
l1
The net movement of C is therefore
yC K1 KC PC K2
=- + f (1.1)
yD
The movement of D, is the amount by which the pilot valve opens. It is contributed by
yC yE
and and can be written as
i4 i3
yD y y K3 yC K4 yE
= i 4 +i 3 ) C + i 3 +i 4 ) E = +

(1.2)
yD
The movement , 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
y
steam valve by E . Certain justifiable simplifying assumptions, which can be made at this
stage are:
a) Inertial reaction forces of the main piston and steam valve are negligible compared to
forces exerted on the piston by high pressure oil.
b) Because of (i), the rate of oil admitted to the cylinder is proportional to port opening
yD
.
yD
The volume of oil admitted to the cylinder is thus proportional to the time integral of .
yE
The movement of is obtained by dividing the oil volume by the area of cross-section of
the piston. Thus,


yE
= K 5 ( y D ) dt (1.3)
0
Taking Laplace transform of equations 1.1, 1.2 and 1.3 and eliminating y D (s ) and
y C ( s ) and rearranging:
1 K sg
y E ( s ) = [ P C ( s )( )
]
R
F ( s ) +(
1+sT sg
) (1.4)
K1 K c K2
Where R= / speed regulation of the governor
K sg=K 1 K 3 K C K4
/ = gain of the speed governor
T sg K4 K5
= 1/ = time constant of speed governor
Equation 1.4 is represented in the form of a block diagram in figure 1.4

Figure 1.4. Speed Governor
1.3.2 Speed Governor Dead-Band
The effect of speed governor dead-band is that for a given position of the governor
control valves, an increase/decrease in speed can occur before the position of the valve changes.
The governor dead-band can materially affect the system response. In AGC studies, the dead-
band effect indeed can be significant since relatively small signals are under consideration.
Mechanical friction and backlash and also valve overlaps in hydraulic relays cause the
governor dead band. Due to this, though the input signal increases, the speed governor may not
immediately react until the input reaches a particular value. Similar action takes place when the
input signal decreases. Thus, the governor deadband is defined as the total magnitude of
sustained speed change within which there is no change in valve position. The limiting value of
dead-band is specified as 0.06%. One of the effects of the governor dead-band is to increase the
apparent steady-state speed regulation, R.

Figure 1.5. Governor with dead-band
It has been found that the backlash non-linearity tends to produce a continuous sinusoidal
oscillation with a natural period of about 2 seconds. An approximate equation for governor dead-
band has been developed as follows
BX N1
D = X + ( N 2 /0 X
N 1= N 2=
where DB is the dead-band. The constants, 0.8 and 0.2
1.4 Turbine Model

Let us now relate the dynamic response of a steam turbine in terms of changes in power
y
output to changes in steam valve opening E . Figure 1.6(a) shows a two stage steam turbine
with a reheat unit. The dynamic response is largely influenced by two factors,
a) entrained steam between the inlet steam valve and the first stage of the turbine,
b) the storage action in the reheater which causes the output of the low pressure stage to lag
behind that of the high pressure stage. Thus, the turbine transfer function is characterized
by two time constants. For ease of analysis it will be assumed here that the turbine can be

modelled to have a single equivalent time constant. Figure 1.6(b) shows the transfer
function model of a steam turbine.
Tt
Typically the time constant lies in the range of 0.2 to 2.5 s.
Figure 1.6
1.4.1 Generation Rate Constraints (GRC)
In power systems having steam plants, power generation can change only at a specified
maximum rate. The generation rate (from safety considerations of the equipment) for reheat units
is quite low. Most of the reheat units have a generation rate around 3% per minute. Some have a
generation rate between 5% per minute to 10% per minute. If these constraints are not
considered, the system is likely to chase large momentary disturbances. This results in undue
wear and tear of the controller.
Several methods have been proposed to consider the effect of GRCs for the design of
automatic generation controllers. When GRC is considered, the system dynamic model becomes
non-linear and linear control techniques cannot be applied for the optimisation of the controller

setting. In figure 1.7, the maximum rate of valve opening/closing speed is restricted by the
T sg gmax
limiters. Here is the power rate limit imposed by the valve or gate control.
Y E g max
The banded values imposed by the limiters are selected to restrict the generation rate to
10% per minute
Figure 1.7. Governor Model with GRC
1.5 Generator Load Model

The increment in power input to the generator load system is
PG P D
PG = Pt
where , incremental turbine power output assuming generator incremental loss to
PD
be negligible and is the load increment.

This increment in power input to the system is accounted for in two ways:
a) Rate of increase of stored kinetic energy in the generator rotor. At scheduled frequency
f 0 , the stored energy is
W 0ke = H
Pr
kWs(kJ)
Pr
where is the kW rating of the turbo-alternator and H is defined as its inertia
constant.
The K.E. being proportional to square of speed (frequency), the K.E. at a frequency of (
f 0+ f ) is given by
f 0 + f 2 f
2 H P r (1+
0
W ke )
= W ke ( f
0 ) f
0
Rate of change of K.E. is therefore,
d 2 H Pr d
( W ke )= 0
( f )
dt f dt
b)
As the frequency changes, the motor load changes being sensitive to speed, the rate of
PD
change of load w.r.t. frequency, i.e., f can be regarded as nearly constant for small
changes in frequency f and can be expressed as

PD
f =B f
f
where the constant B can be determined empirically.
Writing the power balance equation, we have:
2 H Pr d
PG P D= ( f )+B f
f 0 dt
Taking the Laplace transform and rearranging (in p.u), we get:
K ps
F ( s ) =[ PG ( s ) P D ( s ) ] ( 1+ s T ps )
where
2H
T ps= =
Bf 0 power system time constant
1
K ps = =
B power system gain
The above equation can be represented in the block diagram form as:
Figure 1.8. Generator Load Model
The complete block diagram representation of load frequency control of an isolated thermal
power station is shown in figure 1.9

Figure 1.9. LFC of an Isolated Thermal Power System
1.6 MATLAB Modelling and Simulations

The system represented by the block diagram in figure 1.9 is developed in Simulink and
the simulation results are obtained.

Figure 1.10
Model Parameters

Parameter Value
R 1.1
N1 0.8
N2 -0.2
0 rad/s
TG 1 0.2 sec
T R1 0.3 sec
KR 1 11.11
KT1 10 sec
KP1 120
T P1 20 sec
PD 1 0.01 p.u
Pr 75 MW
Table 1.1
Simulation Results

Figure 1.11. Dynamic Response of Change in Frequency for Step Change in Load
It can be seen from the above plot that in the primary LFC, the steady state error in
frequency, when the system in subjected to a step change in load, is not reduced to zero. This
steady state error can be effectively reduced to zero by the employment of a PI controller in the
LFC loop.
1.7 Effect of PI Controller on the LFC Loop

System frequency specifications are rather stringent and, therefore, so much change (as
depicted in Fig. 1.11) in frequency cannot be tolerated. In fact, it is expected that the steady
change in frequency will be zero. While steady state frequency can be brought back to the
scheduled value by adjusting speed changer setting, the system could undergo intolerable
dynamic frequency changes with changes in load. It leads to the natural suggestion that the speed
changer setting be adjusted automatically by monitoring the frequency changes. For this purpose,
a signal from f is fed through an integrator to the speed changer resulting in the block diagram
configuration shown in Fig. 1.12. The system now modifies to a proportional plus integral
controller, which, as is well known from control theory, gives zero steady state error, i.e f |
steady state = 0.

Figure 1.12. Proportional Plus Integral Load Frequency Control
1.8 MATLAB Modelling and Simulations (with PI Controller)
The block diagram in Fig. 1.12 is developed in SIMULINK.

Figure 1.13. Simulink Model for Block Diagram in Fig.1.12
Simulation Results

Figure 1.14 Dynamic Response of LFC with and without PI Controller
Here we can observe that the steady state error in frequency has been reduced to zero by
the use of a PI controller.

Figure 1.15 Dynamic Response of LFC for different values of KI
Chapter 2

Two-Area Load Frequency Control
An extended power system can be divided into a number of load frequency control areas
interconnected by means of tie lines. Without loss of generality we shall consider a two-area case
connected by a single tie line as illustrated in Fig. 2.1.
Figure 2.1
The control objective now is to regulate the frequency of each area and to simultaneously
regulate the tie line power as per inter-area power contracts. As in the case of frequency,
proportional plus integral controller will be installed so as to give zero steady state error in tie
line power flow as compared to the contracted power.
It is conveniently assumed that each control area can be represented by an equivalent
turbine, generator and governor system. Symbols used with suffix 1 refer to area 1 and those
with suffix 2 refer to area 2.
2.1 Development of Model for Two-Area Control
PG P D
In an isolated control area case the incremental power ( ) was accounted
for by the rate of increase of stored kinetic energy and increase in area load caused by increase in
frequency. Since a tie line transports power in or out of an area, this fact must be accounted for in
the incremental power balance equation of each area.
Power transported out of area 1 is given by

V 1 V 2
Ptie ,1
= X 12 sin( 01 02 )

where 01 , 02 = power angles of equivalent machines of the two areas.
For incremental changes in 01 and

02 the incremental tie-line power can be
expressed as
Ptie ,1 T 12 1 2
(pu) = ( - )
V 1 V 2 cos
0 0
T 12 Pr 1 X 12
Where = ( 1 2 ) = synchronising

coefficient
Since incremental power angles are integrals of incremental frequencies, we can write
Ptie ,1
(pu) = 2
T 12
( f 1 dt - f 2 dt ) (2.1)
Ptie ,2
(pu) = 2
T 21
( f 1 dt - f 2 dt ) (2.2)
Pr1
T 21 T a T
= ( Pr 2 ) 12 = 12 12
The incremental power balance equation for area 1 can be written as

2 H1 d
PG 1 PD 1 f1 B1 f 1 Ptie ,1
= P2 dt ( )+ + (2.3)
Taking the Laplace transform of (2.3) and rearranging:
K ps1
F 1( s) = [ PG 1 ( s ) P D 1 ( s ) Ptie ,1 (s) ] +( ) (2.4)
1+sT ps 1

Taking the Laplace transform of (2.1) and (2.2):
2 T 12
Ptie ,1
(s) = s [ F 1 ( s ) F2 (s) ]
2 a12 T 12
Ptie ,2
(s) = s [ F 1 ( s ) F2 (s) ]
In central load frequency control of a given control area, the change (error) in frequency
is known as Area Control Error (ACE). Let us now turn our attention to ACE in the presence of a
tie line. In the case of an isolated control area, ACE is the change in area frequency which when
used in integral control loop forced the steady state frequency error to zero. In order that the
steady state tie line power error in a two-area control be made zero another integral control loop
(one for each area) must be introduced to integrate the incremental tie line power signal and feed
it back to the speed changer. This is accomplished by a single integrating block by redefining
ACE as a linear combination of incremental frequency and tie line power. Thus, for control area
1.
ACE 1= P tie1 +b 1 f 1
b1
where constant is called area frequency bias.
In Laplace domain:
ACE 1 (s)= Ptie 1 ( s)+b1 F 1(s)
and
ACE 2 (s)= P tie2 (s)+b2 F 2 (s)
Combining the block diagram of Fig.1.12 and the equations developed above, we can
easily obtain the block diagram of Fig. 2.2.

Model Parameters
Parameter Value
b1 0.425
b2 0.425
K I ,1 0.20
K I ,2 0.20
a12 -1.0
T 12 0.0707
Table 2.1

Figure 2.2 Block Diagram of Two-Area LFC
2.2 MATLAB Modelling and Simulations
Fig. 2.3 shows the SIMULINK model for Fig. 2.2.

Figure 2.3. SIMULINK Model for Block Diagram in Fig. 2.2
Simulation Results

Figure 2.4. Change in Frequency in Area 1 and Area 2
Figure 2.5. Change in Tie-Line Powers for Area 1 and Area 2
Chapter 3

LFC with Supercapacitor Energy Storage System

(SCESS)
A new technology, the supercapacitor, has emerged with the potential to enable major
advances in energy storage. Supercapacitors are governed by the same fundamental equations as
conventional capacitors, but utilize higher surface area electrodes and thinner dielectrics to
achieve greater capacitances. This allows for energy densities greater than those of conventional
capacitors and power densities greater than those of batteries. As a result, supercapacitors may
become an attractive power solution for an increasing number of applications. This brief
overview focuses on the different types of supercapacitors, the relevant quantitative modelling
areas, and the future of supercapacitor research and development.
In response to the changing global landscape, energy has become a primary focus of the
major world powers and scientific community. There has been great interest in developing and
refining more efficient energy storage devices. One such device, the supercapacitor, has matured
significantly over the last decade and emerged with the potential to facilitate major advances in
energy storage.
Supercapacitors, also known as ultra capacitors or electrochemical capacitors, utilize high
surface area electrode materials and thin electrolytic dielectrics to achieve capacitances several
orders of magnitude larger than conventional capacitors. In doing so, supercapacitors are able to
attain greater energy densities while still maintaining the characteristic high power density of
conventional capacitors.
Figure 3.1
3.1 FEATURES OF MAXWELLS BOOSTCAP SUPERCAPACITOR

1. Temperature effects and performance changes over time

The performance of Maxwell Technologies ultra capacitors is very stable over a wide
operating temperature due to the chemistry and physical make up of the products. An advantage
of the ultra capacitors organic based electrolyte is its low freezing point. This enables the ultra
capacitors to be utilized over a wide range of temperatures, with relatively unaffected
performance. Plots of capacitance and resistance change as a function of temperature are
available on request.
2. Lifetime
Ultracapacitors life is predominantly affected by a combination of operating voltage and
operating temperature. The ultracapacitors has an unlimited shelf life when stored in a
discharged state. When referring to ultracapacitors life the data sheets reflect the change in
performance, typically decrease in capacitance and increase in resistance. The ultracapacitors
does not experience a true end of life rather the performance continually degrades over the life of
the use of the product.
The typical degradation behaviour of the ultra capacitor resembles that of an exponential
decay. The majority of the performance change occurs during the initial use of the ultra capacitor
and this performance change then levels off over time. The most dramatic effect of the life
degradation is on the internal resistance of the device.
In many applications, such as UPS, the ultra capacitors will be maintained at working
voltage until needed. Plots are available on request that show the degradation in rated
capacitance for ultra capacitors held at typical working voltages for long periods of time, and at
different temperatures. To give one example, a 15% reduction in rated capacitance and a 40%
increase in rated resistance may occur for an ultra capacitors held at 2.5 V after 88,000 hrs at 25
o C. The plots, along with the fact that the influence of temperature has a doubling effect for
every 10 o C, can be used to predict the expected performance change for a variety of conditions.
3. Cycling
From cycle testing performed on the products, under typical conditions the product is
expected to provide in excess of 1 million duty cycles with an approximate 20% reduction in
rated capacitance. More details of the testing and plots of capacitance versus cycles are available
on request.
4. Frequency response

Ultra capacitors have a typical time constant of approximately one second. One time constant
reflects the time necessary to charge a capacitor 63.2% of full charge or discharge to 36.8% of
full charge.
The time constant of an ultra capacitor is much higher than that of an electrolytic capacitor.
Therefore, it is not possible to expose ultra capacitors to a continuous ripple current as
overheating may result. The ultra capacitor can respond to short pulse power demands, but due to
the time constant the efficiency or available energy is reduced.
5. Voltage
Ultra capacitors are capable of operating between their rated voltage and zero volts.
Occasional spikes above the rated voltage will not immediately affect the capacitor. Depending
on the frequency and duration of voltage spikes the life will be reduced.
Efficient utilization of the available energy and power storage is achieved with the widest
operating voltage range use. Most electronics have a minimum voltage threshold for utilization,
limiting the effective utilization voltage of the capacitor although there is no limitation in the
capacitor itself. Since the energy in the capacitor is proportional to the voltage squared according
to E = C V 2 .
It is possible to utilize approximately 75% of the available energy if the application utilizes
from the rated voltage to rated voltage of the capacitor.
6. Polarity
Unlike many batteries the anode and cathode of an ultra capacitor are comprised of the same
material. If the positive and negative terminal and casing are also comprised of similar materials,
then theoretically the ultra capacitor has no true polarity.
For manufacturing and consistency purposes the terminals are marked with polarity. It is
recommended practice to maintain the polarity although catastrophic failure will not occur if the
ultra capacitor is reversed charged for some reason. If the ultra capacitor has been conditioned
for charge in a certain direction and then is changed, the life can be reduced due to this
conditioning. For the PC5, PC10 and PC55 products the case is comprised of stainless steel. Due
to the corrosion potential it is required to maintain the polarity indicated on the products, and
reverse polarity will cause accelerated life reduction.
7. Charging
Since the energy storage mechanism of the ultra capacitor is not a chemical reaction,
charging/discharging of the ultracapacitors can occur at the same rate. Therefore, the rated
current for the ultra capacitor applies for both charge and discharge. The efficiency of charge and
discharge are in practical terms the same. A variety of methods are possible for charging of the

ultracapacitors. This may be either through constant current or constant power charging via a dc
source or through ac charging methods. A separate application note is available discussing
different methodologies for ultra capacitor charging.
8. Series connection and balancing

Since the individual ultra capacitor cell voltage is relatively limited compared to the majority
of application requirements, it is necessary to series connect the ultra capacitors to achieve the
voltage required. Because each ultra capacitor will have a slight tolerance in capacitance and
resistance it is necessary to balance, or prevent, individual ultra capacitors from exceeding its
rated voltage.
Balancing can be achieved through two different methods, active balancing or passive
balancing:
1
Active balancing schemes are varied. Maxwell has adopted a balancing methodology based
on a linear voltage balancing scheme. This methodology will always attempt to balance two
adjoining ultra capacitors based on the voltage mismatch between the two ultra capacitors. The
maximum current during balancing varies by product. Refer to the product data sheet or product
manual for more information.
Passive balancing implies no variation in the voltage regulation as a function of the ultra
capacitor condition. The most typical method of passive balancing utilizes resistors in parallel
with the ultra capacitors.
A variety of interconnect methods are employed with the various product offerings. They
range from bus bar interconnecting to soldering. In general the larger the cell capacitance the
more critical the cell interconnects becomes. The larger capacitance devices have internal
resistances on the order of a few hundred micro ohms. A poor interconnection can have more
resistance than the internal resistance of the device itself. Larger devices will generally be
required to carry larger currents, thus necessitating reliable interconnects.
9. Efficiency
Unlike batteries, the ultra capacitor has the same efficiency during charge or discharge. This
enables the ultra capacitor to be recharged quickly without current limiting as long as the current
is within the rated current for the device.
The only efficiency losses associated with ultra capacitors are due to internal resistance of the
device resulting in IR drop during cycling. For most uses the ultra capacitor efficiency is in
excess of 98%. For high current or power pulsing the efficiency is reduced. Typical efficiency
under high current pulses is still greater than 90%.

10. Thermal Properties

For minimum performance influence over the life of the application it is necessary to
maintain the ultra capacitor core temperature within the rated temperature range of the device.
The lower the temperature is maintained the better for life considerations.
All products are provided with an electrically insulating shrink sleeving around the capacitor
body. For this reason and since all current passes through the capacitor terminals, cooling at the
capacitor ends or terminals is the most efficient means for cooling of the capacitor.
Depending on the duty cycle of the application cooling can be accomplished via heat sinks
(conduction), air flow (convection) or a combination of the two. Consideration should be made
for the duty cycle and resulting capacitor temperature as well as the anticipated ambient
temperature the device will be operating under. The combination of the two should not exceed
the operating temperature for the ultra capacitor.
3.2 Applications of SCESS in Power and Energy Systems
It is the fast response that makes SCESS able to provide benefit to many potential utility
applications. The applications of SCESS are described in the following.
1) Energy Storage An SCESS unit could provide the potential for energy storage of up to
5000 MWh with a high return efficiency (up to 95% for a large unit) and a rapid response time for
dynamic change of energy flow (milliseconds). This aspect makes it ideal for large variations in
energy requirements between daytime peak demand and off-peak back-down as well as large
amounts of energy available for replacement of major unit trips. This may provide for the
potential reduction of spinning reserve requirements.
2) Load Following An SCESS unit has the ability to follow system load changes almost
instantaneously which provides for conventional generating units to operate at constant output.
3) System Stability An SCESS unit has the capability to dampen out low frequency
power oscillations and to stabilize system frequency as a result of system transients.
4) Automatic Generation Control An SCESS unit can be the controlling function

in an AGC system to provide for a minimum of area control error (ACE).

5) Spinning reserve In case a major generating unit or major transmission line is

forced out of service, a certain amount of generation must be kept unloaded as spinning
reserve. An SCESS unit can represent a tremendous amount of spinning reserve capacity
when in the charged mode. This lowers the costs for spinning reserve requirements over
comparable values and methods of maintaining spinning reserve.
6) Reactive volt-ampere (VAR) control and power factor correction

An SMES unit can increase the stability and power carrying capacity of a transmission
system.
7) Black start capability An SCESS unit can provide power to start a generating unit
without power from the grid. This provides for grid restoration when area failures have occurred.
8) Bulk energy management An SCESS unit has the ability to store large quantities
of energy, and thus can act as a storage and transfer point for bulk quantities of energy based on
the economics, potentially lowering the cost of electricity.
9) Transient voltage dip improvement A transient voltage dip lasting for 1020
cycles can result when a major disturbance on the power system occurs. SCESS and associated
converter equipment has been shown to be effective for providing voltage support which can
result in increasing the power transfer limitations on the transmission system.
10) Dynamic voltage stability Dynamic voltage instability can occur when there is a
major loss of generation or heavily loaded transmission line and there is insufficient dynamic
reactive power to support voltages. SCESS has been shown to be effective in mitigating dynamic
voltage instability by supplying real and reactive power simultaneously supplanting loss of
generation or a major transmission line.
11) Tie line control When power is scheduled between utility control areas, it is important
that the actual net power matches closely with the scheduled power. Unfortunately, when
generators are ramped up in one control area and down in the receiving control area to send
power, the system load can change causing an error in the actual power delivered. This ACE can
result in inefficient use of generation. SCESS can be designed with appropriate controls to inject
power to virtually eliminate this error and insure that generation is efficiently used and power
schedules are met.
12) Under-frequency load shedding reduction When the power system suffers
the loss of a major resource such as a generating plant or major importing transmission lines the
system frequency will drop and continue to decline until the generating resourceload balance is
restored. Because SCESS can inject real power rapidly into the system, it is an effective method
to offset, or reduce, under-frequency load shedding because it reduces the mismatch between load
and supply capability of the system disturbance.
13) Circuit breaker reclosing Following clearance of a fault, circuit breakers attempt
to reclose and return the affected transmission line to service. This is accomplished routinely
whenever the power angle difference across the circuit breaker is within acceptable limits.
However, protective relays prevent the circuit breaker from reclosing if the angle difference is too
large. By briefly
supplying some fraction of the power normally transmitted by the transmission line, SCESS can
reduce the power angle difference across a circuit breaker and allow reclosure of the circuit
breaker. This allows restoration of the system power transfers quickly following outages of major
transmission lines.
14) Power quality improvement SCESS can provide ride through capability and
smooth out disturbances on power systems that would otherwise interrupt sensitive customer
loads. When momentary disturbances such as transmission line flashovers or lightning strikes
occur, power can be lost if the transmission line trips, or voltages can dip low. SCESS has very
fast response and can inject real power in less than one power cycle preventing important
customers from losing power.
15) Backup power supply The energy storage capacity of SCESS can be used as a
backup power supply for large industrial customers in case of loss of the utility main power
supply. Studies have shown SCESS can be sized with the appropriate energy storage and capacity
to provide backup through most disturbances and be cost-effective.
16) Sub-synchronous resonance damping Generators which are connected to

transmission lines which have high levels of series compensation (series capacitors) can be
exposed to a phenomenon called sub-synchronous resonance (SSR) which can result in serious
damage to the generator. SCESS as an active device can be designed to provide mitigation of SSR
and allow higher levels of series compensation to be installed.
17) Electromagnetic launcher An electromagnetic launcher requiring high power

pulse sources has been developed as a rail gun for military applications. A rail gun can launch
projectiles at velocities higher than 2000 m/s, surpassing the conventional possibilities. Due to its
high power density, SCESS is a very interesting energy storage device for an electromagnetic
launcher.

18) Minimization of power and voltage fluctuations of wind generator

Due to random variations of wind speed, output power and voltage of wind generator
fluctuate randomly. These fluctuations pose serious problems on the system, for example,
lamp flicker and inaccuracy in the timing devices. Since an SCESS unit is capable of
controlling both the active and reactive powers simultaneously, it can act as a good tool to
decrease voltage and power fluctuations of the wind generator system considerably.
3.3 Configuration of SCESS in Power System
Once the rated voltage in the super-capacitor is reached, the Supercapacitor unit is ready
for automatic generation or load control. The change in ACE is sensed and used to control the
power released or absorbed by the Supercapacitor unit.
The power conversion system (PCS), which is used for interfacing the SCESS with an
electric utility system, are given in Figure 3.4. On one hand, the line-side converter of the PCS
maintains a constant voltage on the dc links, and on the other hand, integration of SCESS into
the dc bus requires a rapid bidirectional interface. In this project, a combined buckboost dcdc
converter technology based on insulated-gate bipolar transistors is used for this purpose.
The SCESS used consists of a string of three Maxwell Boostcap modules BMOD0018-
P390 (18F/390V) yielding a structure of 6F/1170V/750kW and the line side converter is assumed
to maintain a dc voltage of 2.5kV. The nominal voltage for the SCESS is chosen to be 925V.
When there is a sudden rise in power demand in a control area, the stored energy is
almost immediately released by the SCESS through its PCS as a line quantity AC. As the
governor control mechanism starts working to set the power system to the new equilibrium
condition, the SCESS coil stores back its nominal energy. Similar is the action when there is a
sudden decrease in load demand. The SCESS immediately gets charged to its full value, thus
absorbing some portion of the excess energy in the system and as the system returns to its steady-
state, the excess energy is released by SCESS to the system and the stored energy again attains
its nominal value.

Figure 3.2
Single line diagram of two area power system with super-capacitor units.

Figure 3.3
Configuration of super-capacitor bank in control area.
Figure 3.4 Basic configuration of SCESS

As shown in Fig. 3.4, the primary control loop translates the power command received
from the supervisory controller into a current command for the supercapacitor bank (SCB). The
PI controller in the inner current control loop forces the dcdc converter to accomplish the
desired current flow through the SCB.
The bidirectional dcdc converter has basically two modes of operation, namely, the buck
or charges mode and the boost or discharge mode.
For buck/boost mode, the chopper employs modulation of the buck/boost switch. The
following gives the general expression for the dcdc converter average output voltage VSCB in
terms of dc-bus voltage VDC:
VSCB = mVDC.
The modulation index m [0, 1] is given by
m = D in buck mode
= 1-D in boost mode.
Figure 3.5 Charging Mode

Figure 3.6 Discharging Mode
3.4 Converter Operation
When there is surplus power in the system, SW1 and D1 will operate in order to store the
surplus energy in the super-capacitor. Consequently when the system is deficient of power, SW2
and D2 will operate and required energy is released by the supercapacitor. Under normal
conditions, when there is equilibrium between the generation and demand, neither any SCR nor
any Diode will conduct. This will lead the super-capacitor to remain in floating mode. To serve
the purpose, the super-capacitor is employed with the two quadrant chopper arrangement as
shown in Fig 3.7.
Figure 3.7 Two quadrant chopper operation for control of SCESS

The waveforms are shown below
Fig 3.8 (a) Switching Sequence of Chopper (b) Super-Capacitor Current Waveform
(c) Current Waveform Supplied by the system.

Fig 3.9 (a) Switching Sequence of Chopper (b) Super-Capacitor Current Waveform
(c) Current Waveform Supplied to the system.
Model Parameters
Parameter Name Value

Vdc(DC link voltage) 2.5kVolts
Vucb(Nominal Voltage) 925Volts
Cucb(Capacitance) 18 F
Pr(Base Power) 750 kW
Table 3.1

3.5 SIMULINK Model for SCESS


Chapter
4
SUPERVISORY ADAPTIVE PREDICTIVE

CONTROL FOR SCESS
4.1 GENETIC ALGORITHM
A Genetic algorithm GA uses the principles of evolution, natural selection and genetics in
a computer algorithm to simulate evolution. Essentially, the Genetic algorithm is an optimization
technique that performs a parallel, stochastic, but directed search to evolve the fittest population.
Why use evolution as an inspiration for solving computational problems? To evolutionary
computation researchers, the mechanisms of evolution are well suited for some of the most
pressing computational problems in many fields. Many computational problems require
searching through a huge number of possibilities for solutions. An example is the problem of
searching for proteins, in which an algorithm is sought that will search among the vast number of
possible amino acid sequences for a protein with specified properties. Such search problems can
often benefit from an effective use of parallelism, in which many different possibilities are
explored simultaneously in an efficient way. For example, in searching for proteins with
specified properties, rather than evaluate one amino acid sequence at a time, it would be much
faster to evaluate many simultaneously. What is needed is both computational parallelism and an
intelligent strategy for choosing the next set of sequences to evaluate.
The current literature three main types of search methods: calculus based, enumerative
and random. Calculus based methods have been studied extensively. These subdivide into two
main classes: indirect and direct. Indirect methods seek local extreme by solving the usually
nonlinear set of equations resulting from setting the gradient of the objective function equal to
zero. Given a smooth, unconstrained function, finding a possible peak starts by restricting search
to those points with slope of zero in all directions. On the other hand the direct search methods
seek local optima by hopping on the function and moving in a direction related to the local
gradient. This is simply the notion of hill climbing: to find the local best , climb the function in
the steepest permissible direction.
Both the calculus based methods are local in scope: the optima they seek are the best in a
neighborhood of the current point. Clearly, starting the search procedures in the neighborhood of
the lower peak will cause us to miss the main event (the higher peak). Furthermore, once the
lower peak is reached, further improvement must be sought through random restart or other
trickery.

Another problem with calculus based methods is that they depend upon the existence of
derivatives (well defined slope values). Even if we allow numerical approximation of
derivatives, this is serious shortcoming. The real world of search is fraught with discontinuities
and vast multi modal noisy search spaces; methods depending upon restrictive requirement and
derivative existence are unsuitable for all but a very limited problem domain.
Enumerative schemes have been considered in many shapes and sizes. The idea is fairly
straight forward: within a finite search space, the search algorithm starts looking at objective
function values at every point in the space , one at a time. Although the simplicity of the type of
algorithm is attractive , an enumeration is a very human kind of search , such schemes have
y=1 x+ 2
applications parameter set for a line by creating independant bit strings for thetwo
1 2
unknown constants and (parameter set describing the line) and then joining them. A
bit string is a combination of 0s and 1s, which represents the value of a number in binary form.
n
An n bit string can accommodate all integers upto the value 2 1 .
For problems that are solved by the genetic algorithm, it is usually known that the
parameters that are manipulated by the algorithm will lie in a certain fixed range, say {
min , max
}. A bit string may then be mapped to the value of a parameter, say i , by the
mapping
b
i= min ,i + L
(max, imin ,i )
2 1. (4.1)

Where b is the number in decimal form that is being represented in binary form (e.g.,
max
152 may be represented in binary form as 10011000), L is the length of the bit string and
min
and are user specified constants, which depend on the problem in hand.
The length of the bit string based on the handling capacity of the computer being used,
i.e. how long a string the computer can manipulate at an optimum speed.
4.1.1 FITNESS FUNCTION

A fitness function takes a chromosome (binary string) as an input and returns a number
that is a measure of the chromosomes performance on the problem to be solved. Fitness function
plays the same role in GAS as the environment plays in the natural evolution. The interaction of
an individual with its environment provides a measure of fitness to reproduce. Similarly the
interaction of chromosome with a fitness function provides a measure of fitness that the GA uses
on carrying out the reproduction. A GA is the maximization routine: the fitness function must be
a non negative figure of merit.
It is often necessary to map the underlying natural objective function to a fitness function
form through one or more mappings. If the optimization problem is to minimize cost function
J () , where denotes the parameter set, then the following cost to-fitness transformation
may be used:
1
J ( ) =
J ( )+
(4.2)
Where is the small positive number. Maximization of J can be achieved by

minimization of J ; so that desired effectis achieved.
A fitness function can be any non linear, non differentiable, discontinuous positive
function because the algorithm only needs a fitness value assigned to each string.[2]
4.2 Implementation of GA in Present Scheme

Figure 4.1
A non linear average value model for the SCESS and its primary control loop is
developed for its analysis and design. The GA toolbox in MATLAB is employed to tune the
tracking PI controller with ISE (Integral Square Error) as fitness function. Primary control loop
is designed to make system behave similar to a given reference 1st order system as shown in
above figure. It can be observed that the response of the reference model is difficult to
differentiate from the SCESS model.

Figure 4.2 Tracking performance of SCESS model

4.2.1 MATLAB CODE FOR UPDATING KP & KI;

%% GENETIC ALGORITHM
global N simt gain
itr=0;
N=2; % Number of variables
simt=2;
FitnessFunction = @objprocessPI1new33;
numberOfVariables = 2;
LB=[.00001 .00001 ];
UB=[5 5 ];
[x,fval] = ga(FitnessFunction,numberOfVariables,[ ],[ ],[ ],[ ], LB,UB);
gain(1)=x(1);
gain(2)=x(2);
sim('GA_Model_Adil',[0 simt]); % execute the simulation with the current
chromosome
%% OBJECTIVE FUNCTION
function [J,x]=objprocessPI1new33(x)
global simt gain
gain(1)=x(1);
gain(2)=x(2);
sim('GA_Model_Adil',[0 simt]);
J = (ess1(length(ess1)));

4.3 GENERALISED PREDICTIVE CONTROL WITH ADAPTIVE

FEATURES:
Generalized predictive controllers (GPCs) have been successfully applied in process
control during the last decade. The performance of unstable, non-minimum-phase, or linear
processes with dead-time are improved with this type of controller. However, the kind of process
that can be controlled, or the kind of optimization method used to derive the controller, can
present important restrictions: the performance index must be quadratic, and the model of the
process must be linear and without actuator constraints. In other words, GPCs are limited when
used to control real industrial processes. In this paper the genetic algorithms (GA) technique is
used for optimization in GPCs. As this technique is robust under the presence of nonlinear
structures in the cost function and constraints, it will be shown that a GPC optimized using the
GA technique (GAGPC) can perform better in a real industrial environment.
4.3.1 Supervisory Control
The supercapacitor Energy Storage system is inserted in each control area of the
interconnected power system to damp out electromechanical oscillations and keep frequency and
tie power deviations to a minimum. The supervisory control commands SCESS to tackle load
perturbation by immediately releasing/absorbing energy. The SCESS should acquire nominal
state (Energy status) after dealing with a disturbance. The control is constrained because of small
rating of the SCESS system.
The supervisory control should be able to handle this constrained control problem.
Generalized Predictive Control (GPC) with adaptive feature is an attractive option for synthesis
of the supervisory control scheme as it is an optimized control.
4.3.2 GPC WITH ADAPTIVE FEATURE

The GPC comprise of two main parts: the model of the system to be controlled and the
optimizer that determines future control actions. GPC makes use of receding horizon principle.
At each sampling instant, predictions of the controlled and constrained variables are performed
over an interval of Ny samples, and the optimal control problem is solved. Normally, the control
signal is assumed to remain fixed after Nu samples. Ny and Nu are referred to as the prediction
horizon and the control horizon, respectively. On obtaining the solution of the optimal control
problem, the first element of the optimal sequence is applied, and the other elements are
discarded. The entire process is repeated at the next sampling instant.

Figure 4.3
GPC with adaptive feature has attracted researchers of control area. The technique is
known as AGPC. It is an indirect adaptive control scheme. Adaption of controller parameters is
carried out in two steps:
1) online estimation of parameters of plant model

2) online computation of the controller parameters based on the estimated plant model.
The resulting control system thus has a nonlinear and time-varying character. However for such
systems it is difficult to establish stability analytically. Stability proofs require assumptions that
give only partial results. For example, in the case of indirect schemes such as adaptive pole
placement and AGPC, the controller has been shown to stabilize the system if it is assumed that
the estimated plant model corresponds to the exact plant model. The use of an external or internal
excitation signal even for short periods of time is useful for speeding up the convergence of the
indirect adaptive control scheme since it can be shown that the exponential stability is achieved
under richness conditions. However, it is very important to note that, because of the complexity
involved, the effectiveness of the indirect adaptive control schemes in various applications has
been mainly demonstrated through simulation/and experimental results without any analytical
examination of closed loop stability and performance properties.

Figure 4.4
4.3.3 ONLINE MODEL IDENTIFICATION
In order to implement GPC, a suitable prediction model is required. The power system is
a complex nonlinear and a very high order system. However, for control purposes, it is not
necessary to identify online a detailed model but only the part that faithfully represents the
essential features of the system for task at hand. For power-system control problems, a third-
order model is normally sufficient. Errors due to low model order are reduced by online
identification of the model parameters every sampling interval. The parameters of the low-order
model do not have any physical denotation, and the states have no physical meaning in the
normal sense. The causeeffect relationship between control variable u and controlled variable y
for each control area can be thus described by an autoregressive with exogenous input (ARX)
model that has the following form.
n n
y ( k )= i y ( ki ) + i u( ki ) (4.3)
i=0 i=0

From this equation it clear that the current output can be estimated by using n sets of
previous outputs and input measurements. Integer n is called the order of ARX model. In this
project u is the power command for SCESS and controlled variable y is the function of area
control error (ACE) and voltage deviation of SCESS (from the nominal value) because
frequency deviation, tie power deviation and voltage deviation have to be zero in steady state. In

this project, the model parameters i i , which vary with system operating conditions, are
estimated using well known recursive least squares (RLS) identification algorithm with
exponential forgetting. The measurement and the parameter vectors are defined as:
( k )=[ 1 2 N 1 2 N ] T (4.4)
T
( k )=[ y ( k 1 ) y ( k1 ) y ( kn ) u ( k 1 ) u ( k 1 ) u(k n)] (4.5)
Then the least squares algorithm for parameter estimation is given by following recursive
equations:
^ ( k )=^ ( k1 ) + K ( k ) [ y ( k ) T ( k ) ^ ( k1 ) ] (4.6)
P ( k 1 ) ( k )
K (k )=
+ T ( k ) P ( k1 ) ( k )
(4.7)
P( k1)
P ( k )=[I K ( k ) T ( k ) ]

(4.8)
\
Where ^ ( k ) is the estimate of (k ) , P (k ) is the covariance matrix, K (k ) is
the adjustment gain vector , I is an identity matrix and is the forgetting factor. The
forgetting factor is used to progressively reduce the effects of old measurement on
estimation to the benefit of most recent ones, thus maintaining the capacity to follow variations
in the system.
The identification is done online in presence of load disturbances acting on a system.
Standard RLS ( =1 ) is not capable of estimating the time variant process parameters. In
adaptive control where the process is time variant, this can lead to disastrous results because the

controller will use wrong values of parameter estimates. Exponentially weighted RLS (0<
<1 ) will, however, be useful as long as persistency of excitation exists.
4.3.4 GPC FORMULATION;
This project uses state variable based GPC formulation which is regarded as a better
choice than a transfer function model. Therefore, the real time identified model needs to be
converted to a state variable representation. Normally, in GPC formulation an incremental state
space model is used. This is because many control situations require a non zero steady state
control signal which is achieved by penalizing u (k) in the GPC cost function. The incremental
model approach for SCESS control will however defeat the purpose. The non zero offset has to
be avoided to allow the SCESS and governor controls to play their desirable role. For this reason
incremental state variable representation is avoided in this project and a normal state variable
representation is developed. An observable canonical form of state variable representation of the
identified model has the following format.
X(k+1)=X(k)+bu(k) (4.9)
Y(k) = cX(k) (4.10)
where
[ ]
1 1 0 0 0
2 0 1 0 0
A=
n1 0 0 0 1
n 0 0 0 0
[]
1
2
b=
n1
n
c =[1 0 0].

Ny
Based on the given state variable model, the output prediction over prediction horizon and
Nu
control horizon can be written in compact form as [5]
^ =SX ( k ) + LU
Y
(4.11)
It may be noted that A, b, c and hence S and L are functions of time index k.
4.3.5 OPTIMIZATION:
In GPC, the computation of control signal is based on minimization of a cost function.
The optimal control problem which is the core element of GPC is generally formulated using a
quadratic criterion. F or a single input, single output (SISO) system the quadratic criterion can be
expressed as
Ny N u1
J ( k )= {r ( k +i ) y ( k +i)} 2 y ( k +i) + {u( k+ i) } 2 u (k +i) (4.12)
i=1 i=0
Where r defines the pre specified reference trajectory over the prediction
y u
horizon; and are the weighting sequences over the future output and control
respectively. [5]
4.3.6 CONSTRAINTS IMPOSED BY SCESS;
For a small rating SCESS we have following constraints;
a) Power command for the SCESS that cannot exceed its converter size.
The power command for SCESS being the control variable, the following constraint is
put on control vector U:
FU f

Where F= I
I[ ] , and
f=
[ ]
umax
umin
umin umax
Also, and are the minimum and maximum power limits of the converter
respectively.Since the converter rating of the SCESS is 1% of the area capacity, therefore
umax =0.01 p . u u =0.01 p . u
and min
b) Constraint due to the supercapacitor voltage :
An important parameter of the supercapacitor is discharge d, which is defined as the ratio of the
Vm
minimum voltage (below which the SCESS cannot be discharged) to the maximum
VM. WM
admissible voltage In fact, the total energy contained in the supercapacitor
cannot be used in order to control the efficiencies. Only a fraction known as the useful energy
can be utilized, i.e.,
[ ( )]
2
d
W u=W M 1
100
(4.13)
To include the constraints due to voltage in the optimization problem, the prediction of either
supercapacitor voltage or energy should be available in the description given below the
supercapacitor energy is treated as constrained variable. The fact that SCESS has been tuned to
behave like a first order reference model, leads to the idea that the reference model can be
cascaded with an integrator to obtain the energy of SCESS as shown below. The pulse transfer
function of the resulting system can be written as
E (z) 11 z1 + 12 z2
=
PUCB 1+ 11 z1+ 12 z2
(4.14)

Figure 4.5
The transfer function parameters are related to time constant T of the reference model and
T
sample time s . An observable state variable representation of given pulse transfer function
model is:
X c ( k +1 )= A c X c ( k ) +b c ( k ) u(k ) (4.15)
y c ( k )=c c X c (k )
(4.16)
where
A c=
[ 11 1
12 0 ]
bc =
[ ]
11
12
c c =[ 1 0 ]
It may be noted that u(k ) and y c ( k ) correspond to the SCESS power command and
supercapacitor energy respectively. [5]

4.3.7 IMPLEMENTATION OF THE SCHEME AND RESULTS:

The functional block diagram showing the implementation aspects of the propose scheme
for the ith control area is depicted in Fig. 4.6
Figure 4.6
A c ,b c c c ,
Matrices which are used in the calculation of and are fixed for given
values of T and Ts and are therefore passed on to the MPC algorithm in the beginning. On the
other hand, matrices A, b, and c are dependent on online estimates of model parameters

( i , i) and have therefore to be updated at every sampling instant, as in above Fig.


Keeping in view the control objectives, we have made the controlled variable y a function
of ACE (ACE = f +Ptie) and SCESS voltage deviation VUCB.,
The exact expression used for control variable y is
y ( k )= ACE ( k )+ k 1 V UCB ( k ) + k 2 ( 1z1 ) ACE ( k )+ k 3 ( 1z1) V UCB ( k ) (4.17)
k1 k 2k 3
where , are constants and the last two terms in the given equation are derivative
terms which are introduced to improve damping. Since y is a combination of deviation

variables, therefore, reference W is zero.
Figure 4.7
Simulation model with proposed control scheme.
For simulation studies, blocks, such as online identification, conversion to state

space, reconstruction of state vector X, reconstruction of state vector Xc, and generalized
model predictive controller, are represented by S-function code in MATLAB/SIMULINK

blocks. Two case studies were then carried out to demonstrate the effectiveness of the developed
scheme in accomplishing the control objectives. In the first case study, the system is subjected to
a load rejection of 0.01 p.u. in area 1, followed by a load increment of 0.01 p.u.. The simulation
results are summarized as follows.
1) The frequency and tie-power deviations with the SCESS controlled through the proposed
control scheme are reduced by more than 80% and 60%, respectively. The proposed
scheme is thus very effective in arresting the peak frequency and tie-power deviations in
comparison with the previous conventional schemes.
2) The proposed scheme forces the SCESS to operate near its constraints and thus
dampen the power-system oscillations quickly and effectively.
3) The power handled by the SCESS converter never exceeds 0.01 p.u. which is the rated
value of the converter.
4) The SCESS voltage remains within the upper and lower permitted bounds.
5) After dealing with a disturbance, the SCESS acquires its nominal voltage (energy) and
remains ready to tackle a new load perturbation.
Thus with the proposed scheme all the control objectives are effectively accomplished.
Figure 4.8
Frequency deviation in Hz for Area 1 with and without SCESS

Figure 4.9
Frequency deviation in Hz for Area 2 with and without SCESS

Figure 4.9
Tie power deviation in p.u
Figure 4.10
SCESS voltages in volts

Figure 4.11
SCESS Power in p.u
Chapter 5
SUMMARY, CONCLUSION AND REFERENCES.
While an SCESS is integrated with a power system for improving LFC, the SCESS
system imposes certain constraints. During operation, the power and the voltage of SCESS
system has to be restrained within allowable limits. On the other hand the effective and economic
use of SCESS demands that the storage system be operated near its limits to derive maximum
benefit from it. Therefore, the control of SCESS has to be well tailored so that while addressing
the power system requirements, the constraints of SCESS are also respected. This chapter has
presented a new control scheme for improved and profitable operation of an SCESS for LFC
support. This scheme which is based on generalized predictive control, is capable of handling all
the control issues associated with the operation of small rating of SCESS in a systematic manner.
Its long range prediction feature allows the GPC to anticipate and prevent future constraint
violations thus improving the system performance. By allowing SCESS to be safely operated
near its limits, GPC on the one hand enhances control performance and on the other hand
increases profitability of energy storage system.

For implementation of GPC in this project a suitable prediction model of the system was
developed based on the line identification of the system. The resulting scheme is thus adaptive.
AGPC scheme also requires the predictions of the supercapacitor voltage and energy level. For
this purpose the secondary control loop of SCESS was tuned to behave like a reference first
order system, whose model cascaded with an integrator was used for prediction of SCESS
energy. GA was employed for tuning proportional -integral controller of the SCESS secondary
control loop.
Comprehensive simulation studies are carried out in MATLAB/SIMULINK environment.
REFERENCES.
1. S.H Zak, System and Control NY: Oxford, 2003
2. M. Gopal, Digital Control and State Variable Methods, India: Tata McGraw Hill, 2008.
3. S.J.Iqbal, M.D.Mufti, S.A.Lone and I.Mushtaq,Intelligently Controlled Superconducting

Magnetic Energy Storage for Improved Load Frequency Control International Journal
of Power and Energy Systems, vol.29, no 4 pp.241-254, 2009.

4. M.D.Mufti, S.J.Iqbal, S.A.Lone,M.Ahmad and M.Ismail,Supercapacitor Based Energy

Storage System for improved Load Frequency Control, Electric Power Systems
Research vol. 79,no 1, pp.226-233, 2009.
5. M.D.Mufti, S.J.Iqbal, S.A.Lone and Q.U.Ain, Supervisory Adaptive Predictive Control

Scheme for Supercapacitor Energy Storage System, IEEE Systems Journal, vol. pp,
issue 99,2013.
6. P Kundhur,Power System Stability and Control, NY: McGraw Hill Inc.2009.
7. D.P. Kothari and I. J. Nagrath, Power System Engineering, India: McGraw Hill
Education.2008.
8. S.J.Iqbal, M.D.Mufti, S.A.Lone and I.Mushtaq, Improved Load Frequency Control with
Superconducting Magnetic Energy Storage in Interconnected Power Systems, IEEJ
Transactions on Electrical and Electronic Engineering, vol.2, No3, pp.387-397, 2007.
9. C.W.Taylor, K.Y. Lee and D.P.Dave, Automatic Generation Control Analysis with
Governor Deadband Effects, IEEE Transactions on Power Apparatus and Systems,
vol.PAS-98, no 6, pp.2030-2036, 1979.
10. C. Concordia, S.B. Crary and E. E. Parker, Effect of Prime Mover Speed Governor
Characteristics on Power System on Frequency Variations and Tie line Power Swings,
Transaction of the AIEE on Power Apparatus and Systems, vol. 60, no 6, pp.559-567,
1941.
11. C. Concordia, L. K. Kirchmeyer and E. A. Szymanski, Effect of Speed Governor

Deadband on Tie line Power and Frequency Control Performance, Transaction of the
AIEE on Power Apparatus and Systems, vol. 76,no 3, pp.4299-434, 1957.
12. Y. Jia, A Control Method of Prolonging the Service Life of Battery in Stand Alone
Renewable Energy System Using Electric Double Layer Capacitor(EDLC), in 2005 ,
IEEE Conference on Power Electronics and Drives, PEDS 2005, Kula Lumpur, pp.228-
233, 2005.

13. S. C. Tripathy, R. Balasubranium and P. S. C. Nair,Small Rating Capacitive Energy

Storage for Dynamic Performance Improvement of Automatic Generation Control,
Proceedings of the IEE-C, vol. 138, no 1, pp.103-111, 1991.
14. T. Himaya and G. Okabe, Coordinated Load Frequency Control Between lFC Unit and
Small Sized High Power Energy Capacitor System, in 2004 International Conference on
Power Systems Technology- POWERCON-2004,pp.1229-1233, Singapore, 2004.
15. R. J. Abraham, D. Das and A. Patra, Effect of Capacitive Energy Storage on Automatic
Generation Control, in the 7th International Power Engineering Conference 2005-
IPEC2005, vol. 2, pp.1070-1074, Singapore, 2005
16. R. W. Boom, B.C. Haimson, G. E. McIntosh, H. A. Peterson and W. C. Young,

Superconductive energy storage for large systems, IEEE Transactions on Magnetics,
vol. MAG-11.no 2, pp.475-481, 1975.
17. P. A. L. Ham, E. Swidenbank and M. Lown, Frequency control: specification

requirements, practical problems and possible solutions, in IEEE Colloquium on
Frequency Control Capability of Generating Plant, London, 1995.
18. S. C. Tripathy, G.S. Hope and O. P. Malik, Optimization of Load Frequency Control
parameters for power systems with reheat system turbines and governor dead band
nonlinearity, Proceedings of the IEE, vol. 129 , pt. C, no. 1,pp. 10-16, 1982.
19. T. S. Tsay, Load Frequency Control on interconnected power system with governor
backlash nonlinearities, International Journal of Power and Energy Systems, vol. 33,
no. 9, pp. 1542-1549, 2011.
20. Lu, C. -F., Liu, C. -C., & Wu, C. -J. Effect of Battery Energy Storage System on Load
Frequency Control Considering Governor Deadband and Generation Rate Constraint,
IEEE Transactions on Energy Conversion, vol. 1o, no.3, pp.555-561, 1995.

21. A. G. Ter-Gazarian, Energy Storage for Power Systems, UK: IET Power and Energy
Series, 2011.
22. A. J. Wood and B. F. Wollenberg, Power Generation, Operation and Control, NY: John
Wiley & Sons, 1996.
23. P. F. Ribeiro. B. K. Johnson, M. L. Crow, A. Arsoy and Y. Liu, Energy storage for
advanced power applications, Proceedings of the IEEE, vol. 89, no. 12, pp. 1744-1756,
2001.
24. V. Bobal, P. Chalupa, M. Kubalak and P. Dostal, self tuning l predictive controller of
nonlinear servomotot,Mar.1998.
25. P. Chalupa, Predictive Control using Self Tuning Model Predictive Controllers Library,
in Proceedings of the 17th Conference on Process Control 2009. Slovakia, pp. 419-425,
2009.
26. K. J. Astrom and B. Witenmark, Adaptive Control. India: Pearson, 2006.
27. Y. J. Cao and Q. H. Wu, Teaching Genetic Algorithm using MATLAB, International
Journal of Electrical Engineering Education, vol.36, pp.139-153, 1999.


Aadilllllllllllllllllllllllllaaaa

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Aadilllllllllllllllllllllllllaaaa

Diunggah oleh

Hak Cipta:

Format Tersedia

Application of Supervisory Adaptive Predictive Control Scheme in

Electrical Power Systems

Department Of Electrical Engineering Page 1

Load Frequency Control Problem

1.1 Active Power and Frequency Control

Department Of Electrical Engineering Page 2

Figure 1.1. AGC and AVR

Department Of Electrical Engineering Page 3

1.2 Generator Response to Load Change

Figure 1.2. Generator Supplying Isolated Load

1.3 Speed Governing System

Department Of Electrical Engineering Page 4

Figure 1.3. Turbine Speed Governing System

Department Of Electrical Engineering Page 5

1.3.1 Model of Speed Governing System

Where Pc is the commanded increase in power.

Two factors contribute to the motion of C:

Department Of Electrical Engineering Page 6

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

The net movement of C is therefore

Department Of Electrical Engineering Page 7

Equation 1.4 is represented in the form of a block diagram in figure 1.4

Department Of Electrical Engineering Page 8

Figure 1.4. Speed Governor

1.3.2 Speed Governor Dead-Band

Department Of Electrical Engineering Page 9

Figure 1.5. Governor with dead-band

1.4 Turbine Model

Department Of Electrical Engineering Page 10

1.4.1 Generation Rate Constraints (GRC)

Department Of Electrical Engineering Page 11

Figure 1.7. Governor Model with GRC

1.5 Generator Load Model

Department Of Electrical Engineering Page 12

Rate of change of K.E. is therefore,

changes in frequency f and can be expressed as

Department Of Electrical Engineering Page 13

where the constant B can be determined empirically.

Writing the power balance equation, we have:

Taking the Laplace transform and rearranging (in p.u), we get:

Figure 1.8. Generator Load Model

Department Of Electrical Engineering Page 14

Figure 1.9. LFC of an Isolated Thermal Power System

1.6 MATLAB Modelling and Simulations

Department Of Electrical Engineering Page 15

Department Of Electrical Engineering Page 16

Department Of Electrical Engineering Page 17

1.7 Effect of PI Controller on the LFC Loop

Department Of Electrical Engineering Page 18

Figure 1.12. Proportional Plus Integral Load Frequency Control

1.8 MATLAB Modelling and Simulations (with PI Controller)

The block diagram in Fig. 1.12 is developed in SIMULINK.

Department Of Electrical Engineering Page 19

Figure 1.13. Simulink Model for Block Diagram in Fig.1.12

Department Of Electrical Engineering Page 20

Figure 1.14 Dynamic Response of LFC with and without PI Controller

Department Of Electrical Engineering Page 21

Figure 1.15 Dynamic Response of LFC for different values of KI

Department Of Electrical Engineering Page 22

Two-Area Load Frequency Control

2.1 Development of Model for Two-Area Control

Power transported out of area 1 is given by

Department Of Electrical Engineering Page 23