212 Operation and Control in Power Systems

Without integral control

Fig. 6.25 Response to integral control with different gain settings

1. The control loop must be stable.

2. The frequency error should return to zero following a step load change. The deviation
in the transient state must also be minimized.
3. The integral of the frequency error should not exceed a certain value (say 150Hz or
3 seconds)

6.21 PID Controllers

From the above analysis, it is clear that proportional integral and derivative control strategy can
be applied for load frequency control. While proportional control is inherent in the feedback
through the governor mechanism itself, derivative control when introduced improves transient
performance and ensures better margin of stability for the system.

where K3 is the gain of derivative controller.

The response to such a control is

By proper choice of K J' K~ and K3 all the specifications for the system performance can
be satisfied.
Load Frequency Control 213

E 6.1 A \00 MW generator has a regulation parameter R of 5%. By how much will the
turbine power increase if the frequency drops by 0.1 Hz with the reference unchanged.
Solution:
Actual change in frequency = 5% of 50Hz = 0.05 x 50 = 2.5Hz
R = 2.5Hzl100Mw = 0.025 HzlMw
If M = -0.1 Hz, the increase in turbine power
I I
~P= -- M= ---x(-O.I) =4 MW
2 0.025
The turbine power increase = 4MW.

E 6.2 A 100 MW generator with R = 0.02 HzlMW has its frequency fallen by 0.1 Hz. If the
turbine power remains unchanged by how much the reference power setting be changed.
Solution:
The signal to increase the generation is blocked. Thus at the input summing point the reference
power setting must be changed. Such that

~ Pref - ~R M= 0

I 1
(i.e .. ) ~ Pref - R M= 0.025 xO.l = 4 Mw

E 6.3 Two generators with ratings 100 MW and 300 MW operate at 50Hz frequency. The
-s.ystem load increases by 100 MW when both the generators are operating at about half
of their capacity. The frequency then falls to 49.5Hz.
If the generators are to share the increased load in proportion to their ratings
what should be the individual regulations? What should be regulations if expressed in
, per unit Hertz/per unit megawatt?

1
Solution: AP= --M
R

I
~P = --M
2 R2
J
M=0.5 Hz
Power is shared in proportional to their ratings
214 Operation and Control in Power Systems

100
~P = 100 x - = 25 MW
I 400

300
~P = 100 x - = 75 MW
2 400

Hence R = - 0.5 = -0.00667

I 75

If regulation is expressed in p.u. then with f = 50Hz

0.00667 300
R2 = - x- = 0.04 pu HzlPu MW
50 I

Both have the same value, even though based on their individual ratings, they have
different regulation.

E 6.4 Determine the primary load frequency control loop parameters for a control area having
the following data:
Total rated area capacity Pr = 1000 Mw
Normal operating load = 500 MW
Inertia constant H = 4.0 sec
Regulation R = 2.5 Hzlpu MW
Solution:
OP 500MW
Load damping D = Of = 50Hz = 10 MW/Hz

[Here, the load damping is assumed linear and percentage change is assumed to be
the same]
10
In per unit D =- = 0.01 pu MW/Hz
1000

2H 2x4.0
Tp = fsD = 50x 0.01 = 16 sec

1 1
K
p
= -
D
= -O.oI~
- = 100 Hzlpu MW
Load Frequency Control 215

E 6.5 Determine the area frequency response characteristic and the static frequency error for
a system with the following data, when I % load change occurs?
D = 0.0 I pu Mw/Hz
R = 2.5Hzlpu MW
Tp = 16 sec
Kp = 100 Hzlpu MW
Solution:
1
Area frequency response characteristic ~ =D+ R

1
= 0.01 + 2.5 = 0.41 MW/Hz

M=- M =- =0.02439Hz
~ 100 x 0.41
E 6.6 In the example E6.5, the governor is blocked so that it does not change the generation.
In that case what would be the steady state frequency error?
Solution:
When the governor is not acting, the feedback loop is not existing. In such a case R is infinite.

I
~=D+ R =D=O.OI puMW/Hz

M
Hence, ~f = - - =- -0.01 = -I Hz
~ 0.01
Frequency falls by 1 Hz (i.e.) f = 50 - I = 49 Hz.
It may be noted that with the generator acting the frequency from E6.5 is
50 - 0.02439 = 49.9756 Hz
The importance of feedback through governor mechanism can be understood from the
above.

E 6.7 A 100 MVA synchronous generator operates initially at 3000 rpm, 50Hz. A 25MW load
is suddenly applied to the machine and the steam valve to the turbine opens only after
0.5 sec due to the time lag in the generator action. Calculate the frequency to which the
generated voltage drops before the steam flow commences to increase to meet the new
load. The value of the stored energy for the machine is 5kW-sec per KVA of generator
energy. Also calculate the value of H constant for the generator.
216 Operation and Control in Power Systems

Solution:
Stored energy = 5KW per/kVA
(i.e.,) = 500 MW sec pr 100 MVA
Load increase =25 MW
Energy required to supply this load for 0.5 sec = 25 MW sec
Frequency at 500 MW sec stored energy = 50Hz
Frequency fall = M

M i1f 25xO.5 MW
-=-= sec
f 50 500

i1f = 50 x 25 x 0.5 = 1.25Hz

500
Frequency falls to 50 - 1.25 = 48.75 Hz

H constant = stored kinetic energy at rated frequency = 5MW - sec x I OOMV A = 5 sec
machine rating MV A I OOMV A

E 6.S Given the following parameters, obtain the frequency error. Plot it when a step load
disturbance of (i) I % and (ii) 2% occur in the system.
Tp = 22 sec
R = 2.5
Kp= 100
Solution:
Gp
The expression for i1F(s) = --I-~--i1PD(S)
l+-GsGToG p
R

Neglecting the turbine dynamics and governor action (G s GTO :::: 1.0)

k
i1F(S) =
I +STp I +~_~ s
R 1+ STp

k] = 0.01 and k2 = 0.02

simulating the transfer functions in .MATLAB the response is obtained and shown
in Fig. E(6.8).
Load Frequency Control 217

Fig. E6.8 Step load frequency error characteristic without supplementary control 1. k = 0.01
=
2. k 0.02 (governor action and turbine dynamics are neglected

E 6.9 Show the effect of governor action and turbine dynamics, if they are not to be neglected
in E6.8 given Ts = 100 msec and TTG = 0.5 sec.
Solution:
For this the exact frequency error is used

LlF(S) = I Gp . LlPD (S)

1+ RGSGTGG p

_ Ks
LlPD(S) = O.Olpu; G s --'"""-- G TG -- K TG
1 +STs I+STTG

Ks KTG ::::: 1
The response is shown in Fig. E(6.9).
It can be seen that greater the step load change, larger the error and the governor action
and turbine dynamics does not cause any change in the response in the steady state, except for
a transient deviation at the beginning of the distribution.
218 Operation and Control in Power Systems

Fig. E6.9 Step load frequency error characteristic without supplementary control 1. k =0.01
= = =
2. k 0.02 Ts 100 msec and TTG 0.5 sec

E 6.10 An isolated power station has the following parameters

Turbine time constant T t = O.5sec
Governor time constant Ts = 0.2sec
Governor inertia constant H = 4 sec
Governor speed regulation R = R per unit
The load varies by 0.8 percent for a I percent change in frequency, i.e. D = 0.8 . The
governor ~peed regulation is set to R = 0.05 pu. The turbine rated output is 250 MW at nominal
frequency \>f 50Hz. A sudden load change of 50MW(dP L = 0.2 pu) occurs.
(a) Construct the SIMULINK block diagram and obtain the frequency deviation
response.
(b) Set integral gain to 7 and obtain the frequency deviation response and compare
both the responses.
- - - - - - - - - - - -

Load Frequency Control 219

t.

Solution:
To create a SIMULINK block diagram presentation select new (model) file from FILE menu.
This provides an untitled blank window for designing and simulating a dynamic system. Copy
different blocks from the simqIink libraries or other previously opened windows into the new
window by depressing the mouse button and dragging.
I. Open the continuous library and drag the transfer function block to the window.
Double click on transfer function to open the dialog box. Enter the numerator and
denominator values (the coefficients in the descending powers of s, if any power of
s is missing, enter zero) of the transfer function .
2. Open the math library and drag the sum block in to the window. Open the sum dialog
box and enter + - under list of signs.
3. From the math library drag the Gain block into the model file right click on the gain
block and click on the Flip option to rotate the gain block by 180degrees.
4. Open the source library and drag the step input block to window. Double click on it
to open its dialog box and set up the step time(step duration), initial and final
values(which will be same) to represent the step input.
5. Open the sink library and drag scope to window to observe the response.
By using the left mouse button, connect all the blocks.
Before starting simulation, set the simulation parameters. Pull down the simulation
dialog box and select parameters. Set the start time, stop time and for a more accurate integration,
set the maximum step size.
In this example the parameters for all the blocks for the system in the fig. are initialized.
Open m-fiIe and enter the parameter values. The following m-fiIe has to be run prior to the
simulation.( of model file)
Open new m-fiIe and enter the parameter values as shown below.
Tg = 0.2;
T t = 0.5;
H= 5;
D = 0.6;
R = 0.05;
%for integral control
Kj = 7;
Save the m-fiIe under parameters and run the file. SIMULINK block diagram and results
for LFC
By using above procedure construct the simulink block diagram for the load frequency
control of isolated power system as shown in Fig. E 6.1O(a).
220 Operation and Control in Power Systems

Constant Governer Turbine Inertial.load

Step

l/R
'-----------<-K-I4-----------'

Fig. E 6.1 O(a) Block diagram Model of Load Frequency Control (Isolated Power System)

Pull down the file menu and use save as to save the model under AFC. Start the
simulation. Double click on the scope, click on the auto scale, the result is displayed as shown
in Fig. E 6.1 O(b).
Add integral controIIer block for above system and save the file under LFC 1.
Start the simulation. Double click on the scope, click on the auto scale, the result is
displayed as shown in figure below.

Fig . E6.10(b) Response of LFC

1. Uncontrolled response
2. Integral control response
Load Frequency Control 221

E 6.11 A single control area system with the following data experiences a sudden load change
of3%.
Kp= 100
Tp= 25
R = 2 Hzlp.u. MW
with integral control using a gain of 10 obtain the frequency plot with time and show that the
frequency deviation is reduced to zero.
Solution:
Using MATLAB, the solution is obtained and shown in Fig. (E 6.11).
Kj=IO Kp = IOO Tp = 25 R =2 num = Kp/Tp
d l = (I1Tp) + (K/ (R * Tp)) d2 = Kj * K/Tp den = [I d l d 2]

Transfer Fan

Fig. (E 6.11)
 222 Operation and Control i!!.!Qwer Systems

6.22 The optimal Control Problem

A dynamic system is described mathematically by
x = f(x,u, t) ..... (6.82)
Where x denotes the state vector and u the control vector
Let it be desired to minimize the performance index or cost functional
I

J = fL(x,u,t)dt ..... (6.83)

°
Pontriagyn's method for optimal control involves in creating an augmented cost function
for minimization. .
L ' (x,x,u, t) = L(x,u, t) + At[f(x,u, t) - x ..... (6.84)
Applying Euler-Lagrange equations, we obtain
.au
A.=-- ..... (6.85)
ax
. au
x=--
aA
..... (6.86)

0= au ..... (6.87)
au
Where u = L (x, u, t) + At [f(x, U, t)] ..... (6.88)
and A. denotes Lagrange multipliers. Simultaneous solution to eqn. (6.85, 6.86, 6.87)
subject to boundary conditions gives optimal control Uo. But, the equations are non linear and
time varying and hence even for simple systems the solution will be difficult to obtain.

6.23 The Linear Regulator Problem

Consider the linear dynamic system

~(t) = [A](t)X(t) + [B][t]!:! (t) ..... (6.89)

Let the cost functional be

I I 12
J = _XT (t 2 )HX(t 2 ) +- fX(t)Q(t)X(t) + u T(t)P(t)u(t)]dt ..... (6.90)
2 21J

Where Hand Q are real symmetric positive semi definite matries and P is a real symmetric
positive definite matrix. Again t2 is fixed, but X(t 2) is free. For simplicity, omitting the arguments .

..... (6.91 )
Load Frequency Control 223

For optimality ..... (6.92)

..... (6.93)

..... (6.94)

From the above VO =-P-I(t)B T(t)90(t) ..... (6.95)

substituting this VO in equation (6.92)

XO (t) = A(t)X(t) - B(t)p-I (t)B T(t)9° (t)}

..... (6.96)
XO (t) = -Q(t)X(t) - AT (t)9° (t)

..... (6.97)

or ..... (6.98)

where + is called state transition matrix.

..... (6.99)

From the boundary conditions we obtain )..0 (t 2) = HXO (t 2) and hence,

H XO(t 2) = +21 XO(t)+ +12 AO(t) ..... (6.100)

It can be derived from the above that

A° (t) = [+22 - H+ 12 ]-1 [H+ 12 - +22]x° (t)

= R(t) XO (t) ..... (6.101 )
From eqn. (6.95) VO(t) = _P- 1B T(t)Ao(t)

= _p-l(t) BT(t) R(t) XO(t)

= _[Kt (t) XO(t)] ..... (6.102)
The optimal control is a time varying linear combination of system states.
224 Operation and Control in Power Systems

6.24 Matrix Riccati Equation

Consider eqn (6.10 1)
°
A (t) == R(t)Xo (t)

Differentiating i Get) == R(t)Xo(t) + R(t)Xo (t) ..... (6.103)

Substituting XO(t) and AO(t) from eqn. (6.92) and (6.93)

== Q(t) X(t) - AT(t) AGet)

== R(t) X(t) + R(t) A(t) X(t) - R(t) B(t) p-l(t) BT(t) AO(t) ..... (6.104)
Eliminating A(t) and rearranging the terms
R(t) B(t) p-l(t) BT(t) R(t) XO(t) - R(t) A(t) X(t) ..... (6.105)

- R (t) XO(t) - Q(t) XO(t) - AT(t) R(t) X (t) == 0 ..... (6.106)

since the above eqn. must be satisfied for at XO(t)
R(t) + R(t)A(t) + AT (t)R(t) - R(t)B(t)p-l(t)B T(t)R(t) + Q(t) ..... (6.107)

• The above equation is called matrix Riccati equation. R is an n x n symmetric matrix.

n(n + 1)
2 first order differential equations only need be solved to get the elements ofR matrix,
so that the elements of feed back gain K(t) can be computed further from equation (6.102)
For controllable dynamic systems H = 0, A, B, P and Q are time invariant matrices and as
S~ ex) R(t) ~ RO, a constant matrix. The matrix Riccatti equation becomes

ROA + ATRo+ Q - ROBp-l BT RO = 0 ..... (6.108)

In the steady state R (t) ~ o.

The gain elements from K matrix can be determined as constants after the elements
of RO matrix are determined. This involves solution to simultaneous non-linear algebraic
equations. There are several methods available for the solution of which Klineman's method is
simple and elegant.

6.25 Application of Modern Control Theory

The load frequency control (LFC) problem can be viewed as a dynamic optimization problem.
Integral square error criterion can be applied and controllers can be designed using optimal
control theory. Controller design using this theory was first proposed by Fosha and Elgerd.
The results obtained by using this theory contradict the existing practice in the selection of the
bias parameters and the integral controller gains.
The system equations are put in the state variable form denoted compactly by
X==AX+Bu+Fd ..... (6.109)
Load Frequency Control 225

where A.B and F are the system, input distribution and disturbance distribution matrices
respectively and X, U and d are the state, control and disturbance vectors respectively.
By a suitable transformation, eqn. (6.24) is transformed into the form.
X' =A'X' +B'U' ..... (6.110)
The LFC problem is then treated as a linear regulator problem and the solution is obtained
by solving algebraic Riccatic equation.

6.26 Optimal Load Frequency Control- Single Area System

The states to be minimized are the frequency deviation Mand the time integral of the frequency
deviation f~f.dt. The contrl signal is to be weighted and included in the cost function so that
too large a control is not required. The cost function selected is

J= f 2"1 ('T~
00
Qx + UT
)
Pu dt ..... (6.111)
o
Where Q and P are the weighting matrices for the state and control vectors and X and u.
Consider the block diagram for the single area system shown in Fig. 6.26.

r---II/R~--------------.....,

~...-+t.F(s)

Fig. 6.26 Block diagram for single area system.

As indicated in section (6.18) Ks' KrG is set equal to one.

The following states are defined

XI = filf(t)dt = integral of frequency error

~ = ~ f(t) = frequency deviation

x3 = ilPTdt) = change in turbo generator output

x4 = ~X v(t) = valve displacement at speed governing system.

Thus, xT = [J~f ilf LlPTG ~Xv ]

Let u be the control vector and d = LlPd be the disturbance vector.
Here, the control is a single input scalar u, and the disturbance d is the change in the load
demand ilPd
226 Operation and Control in Power Systems

The perfonnance index is

qll
XT Qx = {x j x_}
--2 { q2l ..... (6.112)

Since only ~f and fMdt are to be minimized.

Identifying the states and the state variable system model, we obtain
Xl (t) = X2 ( t) ..... (6.113)

Since Xj(t) = fM{t) dtand ~(t) = ~f(t)

From the block diagram

M(s) = [ M>TG (s) -M>p {s)J (I +KiT )

p
..... (6.114)

Cross multiplying and rearranging

S ~ F(s) 1
= __ ~F(s) + Kp ~PTG (s) - Kp M>D (s) ..... (6.115)
Tp Tp Tp
In the time domain, using the state variables

. () 1 Kp Kp
x2 t =--x2+-x3-~' d ..... (6.116)
Tp Tp. Tp
Also, from the block diagram

~PTG(S)=( 1+ST1 l~Xv (S) ..... (6.117)

TG
As before,
STTG ~PTG(S) = ~xv(s) - ~PrG(s) ..... (6.118)
and in time domain

X2 (t) =_I-x4(t) __I_ X 3 (t) ..... (6.119)

TTG J-ro
Finally, from the block diagram

~X v (s) = (_1) [U(S)-~M(S)]

I+ST5 R
..... (6.120)

1
i.e., STs ~(s) = -~ ~ (s) + U(s) - - M(s) ..... (6.121)
R
In the time domain

..... (6.122)
Load Frequency Control 227

Putting eqn. (6.113, 6.116, 6.119 and 6.122) in matrix from, the state variable model is
obtained as

0 0 0
XI (t) 0 0
-I Kp
0 0 0 kp
x2 (t) Tp 1p [ x2
Xl
-T
= __1_ _1_ + 0 + p d
x3 (t) 0 0 x3 0 ..... (6.123)
TTG TTG I
x 4 (t) __1_ I
x4 Ts
0 0 0
RTs -Ts

redefining the state and control variables.

XI = Xl

"2 = x 2
,
x3 = X3 + LlPD
x4 = X4 + LlPD
U + u' + LlPD ..... (6.124)
Eqn. (6.112) can be reduced to

0 1 0 0
.' 1 Kp 0

[~}
Xl 0 0
Tp 1Tp 0
X2
= 0 I u' ..... (6.125)
0 0
X3 Tr Tr I
x4
.' 1
0 0 Ts
RTs Ts
The optimal control to minimize the performance index given in equation (6.111) can be
determined using the solution technique described for the linear regulator problem. It is required
to solve the algebric matrix Ricoati equation.
Q + ATR + RA- RBp-l BT R =0 ..... (6.126)
for the elements of the R matrix which is positive defmite and symmetric. Kleinman's method
may be used to solve the matrix Riccati equation. An initial feedback gain vector KI is selected
such that the matrix ~ = [A - B K I] has eigenvalues with negative real parts. Then the matrix
equations.
Ai RT + ~ Al + Q + ~T P KI = 0
are to be solved for the elements of RI" The new gains are computed using
K(l) = p-l BT R
I I
The procedure is repeated till convergence is obtained for the elements of R I .
228 Operation and Control in Power Systems

The optimal control

UO =- p-l BT RX(t)
= - KT X(t) ..... (6.127)
can be calculated.
E 6.12 For the single area system with the following data, determine the optimal control.
T p = 0.04s R = 2Hzlp.u. MW
TT = O.Ss Kp = 100 Hzlp.u. MW
Ts=O.IS Pn = 0.01 p.u.

Assume Q[~ ~1
Solution:
Substituting the parameters in the algebraic matrix Riccati equation and solving the equation
using, Kleinman's method with an initial vector of K j = [1 1 1 1]
The values of Ki converge to
K = [1.0000 1.1368 1.7092 0.2976]
The optimal control is shown in Fig. E 6.12.

8
7

r
';'
4

::: 2
~
0
::s
0

-1

1.0 2.0 3.0 4.0

o
Time (s) ---.

Fig. E 6.12 Optimal control

Load Frequency Control 229

6.27 Optimal Control for Tandem Compound Single Reheat Turbine- Generator
System
The model for tandem compound single reheat turbine is discussed in
Chapter 3 : The system is shown in Fig. 6.27.

Fig. 6.27 Tandem compound single reheat turbine system

The state - variable model of this system is given by

x = AX+Bu+ Fd
with the initial conditions X(O) = 0
where XT = [Xl X2 X3 X4 Xs X6]
= (f~f M ~PG ~PRH ~PCH ~XV ] .... (6.128)

0 1 0 0 0 0
1 Kp
0 0 0 0
Tp Tp
1 FHP
0 0
Tco TJ T2 TCH
A= 0 1 1
0 0 0
TRH TRH
1
0 0 0 0
TCH TCH
1
0 0 0 0
Ts

The constant T 1 and T2 are given in chapter 3.

1 F FIP FIP
- = -LP
-+-----
Tl Tco Tco TRH
230 Operation and Control in Power Systems

BT = [0 0 0 0 0 ;s 1
FT = [0 - ~: 0 0 0 0 1
u=APc and d=AP D
The above equations are transformed into the form
Xl =AX I +Bu l
using the transformation

0
0
d
X=X'+
d
d
d

and u~1 + d
A quadratic cost function

= r ±(X'T.QX· +u'TRu'~t
is selected and the algebraic matrix Riccati equation is solved as before to obtain the
optimal gains using the equations.
Q-RB p-l BTR + RA +ATR= 0
and u opt = _p-l BT RX
=_LTX ..... (6.129)
The results are plotted in Fig. E.6.
E 6.13 Compute the optimal load frequency control for a single thermal power system with
tandem compound single-reheat steam turbine with the following data
R = 20s TRH = 7.5s Kp = 2 p.u. Hz / p.u. MW
Ts = O.ls Tco = 0.45s Tp = 20s
TCH = 0.25s FHP = 0.3 D = 0.01 p.u.
FJP = 0.4 FLP = 0.3
Load Frequency Control 231

Plot the variations of the controlled states with time

Solution:
Q and P matrices are selected as follows

1 0 0 0 0 0
0 1 0 0 0 0
0 0 0 0 0 0
Q= ; P=1
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
The solution is obtained using a computer program for the algorithm explained
in 6.26. The feedback gains, starting from initial values of unity for each, converge to the
following solution.
LT = [1.0 0.6617 0.0269 0.3037 0.0047 0.0177]
Plots of fM and ,M, the two optimally controlled states are shown in Fig. E. 6.13 (a)
and (b)
o 5 10 15

Time (5)
I x 10-3
....
..::: -2 x I0- 3

1 -3xI0-3

Fig. E6,13 Optimal response

(a) Optimal response of time error to step load change

-4xI0-4 ,.-------------------

....~t 5
Or------+-~---_r-------+_--~-
10

+
-4xI0-4
15

-8x 10-4

-12x 10-4

Fig. E6.13(b) optimal response of frequency error to step load change

232 Operation and Control in Power Systems

6.28 Optimal Control of Hydro Speed Governing System

Consider the system shown in Pig. 6.20. The state variable mode for the system is

X =AX +Bu+ Cu + P.d ..... (6.130)

with the initial conditions XeO) = 0
where XT = [XI X 2 X3 X 4 Xs X6 ]
= U~f ~f ~PG ~PGV ~Ppv ~ppv]
0 1 0 0 0 0
1 Kps
0 0 0 0
Tps Tps
1 1
A= 0 0 --- --- 0
O.5Tw 0. 5Tw 0. 5Tw
0 0 0 0 1 0
0 0 0 0 0
0 -KI -K2 -K3 -K4 -Ks

BT =[0 0 0 0 0
TpT~TG 1
Tp~Gl
T
C =[0 0 0 0 0

pT = [0
_ Kps
Tps
0 0 0 Kps
Tp TR Tps
1
u = ~ Pc = control input
and d = ~ PD = step-load disturbance.
The constants Kl' K:z, K3 K4 Ks are given by
- Tps -TR
K1-
TpTGTRTpS
_ Kps
K2 -
Tp TG Tps
cr
K 3 =---
TpTGTR
K4 = TR(cr+o)+TG
. TpTGTR
1 1
Ks =-+-
Tp TR
Load Frequency Control 233

The state and control variable are transfonned using the relation,

Xl
0
Xl
X 0
X2 2
X d
X3 3
= + d ..... (6.131)
X4 X
4
0
X5 X5 u'
X6 X6 TpTG

U = u' + d (j

The transfonned equations are

j( = A'X' + B' .u'

where A'=A

and B'T = [0 0 0 0 TP~J

The cost functional to be minimized is selected as

J= ~ r(X'TQX' +u'Tpu')dt
Selecting Q and R matrices are :

1 0 0 0 0 0
0 1 0 0 0 0
0 0 0 0 0 0
Q=
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0

andP=l
the algebraic matrix Ricccati equation
Q = R B' p-l B'T R + R A' +A'T R = 0
is solved for the elements of R.
The optimal control is obtained using
u = _p-l B'T R X'
opt
=_LTX ..... (6.120)
234 Operation and Control in Power Systems

The states with optimal control are determined from

..... (6.133)

E 6.14 For the hydro system shown in Fig. 6.14.1; using the following data, compute the
optimal control

Distributor volve and

Pilot volve and
servomotor

Vet min Turbine and Power system

penstock

Permanent speed droop

Speed governing system

Fig. ES.14.1 Hydro speed governing system

Time (s) --+

o 4 8 12 16 20

~ -0.004
~
5 (i) Optimal response
..,l: -0.008
.., (ii) Uncontrolled response
E
E=
-0.012

1 -0.016

(a)

Fig. ES.14.2 Uncontrolled and optimal response of hydro speed governing system
(a) uncontrolled and optimal response of time. error
Load Frequency Control 235

TG = 0.2 s KPS = 2 p.u. Hz! p.u. MW

Tp = 0.04s Ll PD = 0.01 p.u.
Tw = 0.6 s
Tps = 20s
TR = 3s
(J = 0.05

8 = 0.2
The 100 % load conditions on the plant.
.,

Time (s) ----+

0.0004

0.0004

t 0.0008

E 0.0012
:::- (i) Optimal response
'- (ii) Uncontrolled response
<l 0.0016

+ 0.0020

0.0024

Fig. E6.14.2 (b) uncontrolled and optimal response of frequency error

Solution:
Following the procedure outlined in Sec. 6.20, the optimal feed-back gain components are
computed as
LT =[1.0 3.4061 0.078 0.5393 0.1656 0.0063]
The uncontrolled and optimal response are shown in figure E6.14.2 (a) and (b).

6.29 A Review of Optimal Control

The optimal control determined by eqn. (6.116) is quite often impractical due to the following
reasons:
1. The optimal control is a function of all the states of the system. In practice, all the
states may not be available. The inaccessible states or missing states are required to be
estimated.
2. It may not be economical to transfer all the information over long distances.