IEEE Transactions on Power Systems, Vol. 10, No.

2, May 1995 73 1

ECONOMIC LOAD DISPATCH MULTIOBJECTIVE OPTIMIZATION

PROCEDURES USING LINEAR PROGRAMMING TECHNIQUES
AHMED FARAG" SAMIR AL-BAIYAT T.C.CHENG
KING FAHD UNIVERSITY OF PETROLEUM & MINERALS UNIV. OF SOUTHERN CALIFORNIA
DHAHRAN , SAUDI ARABIA LOS ANGELES ,U.S.A.
* CURRENTLY AT UNIVERSITY OF SOUTHERN CALIFORNIA, LOS ANGELES, U.S.A.
ABSTRACT

This paper outlines the Optimization problem of real and l h i s paper presenis a new method to determine the
reactive power, and presents the new algorithm for studying optimal shift in power dispatch related to contingency states
the load shedding and generation reallocation problem i n or overload situations in the system. The approach
emergencies where a portion of the transmission system is incorporates the condition that the additional cost incurred i n
disabled and an a x . power solution cannot be found for the shifting generation should be minimized (4-7).
overloaded system.
The method used for solving the load flow problem is the
l h e paper describes a novel and efficient method and Newton-Raphson method. The Sensitivity KRi between line
algorithm to obtain the optimal shift i n power dispatch
related to contingency states or overload situations in power flow of line II and power injection of bus i has also been
system operation and planning phases under various calculated.
objectives such as economy, reliability and environmental
conditions. In order to demonstrate the effectiveness of the
proposed algorithm, two examples: IO bus system and the
The optimization procedures basically utilize linear IEEE 30 bus, six generators system are considered. Objectives
programming with bounded variables and i t incorporates the selected vary from economy, security to minimal
techniques of the Section Reduction Method and the l h i r d environmental impact. Numerical results have clearly shown
Simplex Method. that the optimal solution by means of the proposed algorithm
i s successfully and favourably compared to the existing
l h e validity and effectiveness of the algorithm i s techniques and algorithms (8-1 1).
verified by means of two examples: a 10-bus system and the
ICE€ ~O-BUS,six generators System. PROBLEM FORMULATION

INTRODUCTION The multi-objective power flow optimization problem

can be formulated as i n the following steps:
The main purpose of the economic dispatch of electric
energy systems have so far been confined to determine 1. OPTIMUM SCHEDULING OF REAL GENERATION
generation schedule that minimizes the total generation and
operation cost and does not violate any of the system T he objective function of the optimum scheduling of the
operating constraints such as line overloading, bus voltage real generation i s to:
profiles and deviations. I n general, power system possesses
multiple objectives to be achieved such as economic
operations, reliability, security and minimal impact on
environment. which inherently have different characteristics,
and hence conflicting relations hold among these objectives.
A new technique which can be almost generally used in
extremizing the quadratic objective function with a linear
programming method has been developed. It is considered to where
be more effective than such a technique as the
"Multi-Segment Curve Method" or the Linear Objective n
Function Method". In addition to the algorithm for a = ? a
calculating real power generation which minimizes the total i=l i '
fuel cost, the algorithm for optimizing bus voltage and
reactive power i s presented. ai is the Basic Cost Coefficient for generator i.
As a measure to counter the line overload, the algorithm n i s the number of generators inclusive of the slack bus
i s developed for settling the Economic Load Dispatch (ELD) 9s
condition and the overload elimination condition generator.

1:; 1,
simultaneously. As a I-inear Programming technique, the
Third simplex Method has been contrived for reducing
execution time and memory size (1-3)- b= bi is the Linear Cost Coefficient for

94 SM 573-6 PWRS A paper recommended and approved

by the IEEE Power System Engineering Committee of the
IEEE Power Engineering Society for presentation at
the IEEE/PES 1994 Summer Meeting, San Francisco, CA,
July 24 - 28, 1994. Manuscript submitted December 13,
1993; made available for printing June 10, 1994.

0885-8950/95/$04.00 0 1994 IEEE

732

cl
Xgi 4
AP . t iP
91
: -P z (6)
so that the conditions of Eqn. 5 reduce to:
c2 0 ci is the Quadratic Cost
(-1)
c - Then we can express the original variables AP with
gi
C. Coefficient for generator i respect to the new variables X as follows:
9i
U C AP .
gi
X
gi
- Po.gi t Pm
gi
(8)
Substituting of Eqn. 8 into Eqn. 4. we have
AJ =b'(Xgs - P o9s + Pm)
9s
= b' X t b' ( -Po t Pm ) (9)
9s 9s 9s
Since the constant terms of the objective function to be
extremized are insignificant, we can use the following
objective function

" ,
AJ=dX =CgSb X
gs k=l k gk

which i s a linear function with respect to the new variables

'gi.

CONSTRAINTS
1 aking the perturbation of the both sides o f Eqn. 1, we get
The Newton-Raphson method formulates a set of linear
3tA3=atb1-(P tAP )t(P tAP )l[c](P tAP ) (2) equations expressing the real and reactive power injection
9s 9s 9s 9s 9s 9s
then errors a t buses as follows:

AJ=blAP tP1 [ c N P ,APT [c]P +APT [c]AP

9s 9s 9s gs 9s gs 9s

= (bTt2PT ap
9s
[c])AP +APT [c]AP
9s 9s 9s
(3)
where, H. N. 3. L are Jacobian matrices defined as s ,ap
3~
As the value of P
9s
approach to the optimal point, AP
9s , 2 and %respectively (4-7).
becomes so small that we can neglect the term of (AP 9s)2. In the real power optimization procedure, setting AV =
0 firstly. l h e relationships of AP, AQ with A e can be
1hen we have the new objective function expressed by the following equations (3-5):
A J = boTAP (4)
gs AP = [HlAQ
AQ = [ J ]Ae
where
bIT = bT t 2Ps: [c] (i)
Real Power Balance Condition
, 1 * I

= (bl b2 ... bk ... bn )

The real power balance condition is defined as the
Equality Constraints. l h e slack bus real power balance
9s condition can be written similar to Eqn. (12(a)) as follows:
APs = [ H ] A 0 (13)
with bk = bk t 2c P
k 9k From Eqn. (12(a)), we have
1he control variables AP . must obey:
91

Pm -Po. < A P < P M - P o .

(5) P n 1
g1 gi - gi - gi gi
where H i s the inverse of the Jacobian matrix H (3-7).
where )xq, )-$,l-$ are matrices consisting of the column vectors o f
i s the initial value of real power generated at the matrix H pertaining to generation buses, control buses,
iP
: and load buses, respectively.
generating bus i
Since the bus power of the non-generator buses cannot
is the upper l i m i t of real power generated at be changed, i.e. APc = 0 and Apt= 0. Equation (14) can be
:P
generating bus i written as follows:
i s the lower l i m i t of real power generated a t A@= [H,l AP9 (1 5 )
:P
generating bus i Substituting Eqn. (15) into Lqn. (13) leads to:
aPs = CHsl [H 1AP (16)
Since some of the APgi may be negative and linear 9 9
Using equation (8)
programming can only operate on nonnegative variables, new
variables should be introduced through i
APgi= .
X 91 - Po.
91 t Pm
91
 733

and substituting in Eqn. (16) in view of Eqn. (8). then [N ] is the (n x n ) matrix consisting o f the r o w
9s 9s b
xgs - P O t P; = [lis] [Hgl(Xg - Pgo t P$ of the Jacobian matrix [NI pertaining to the
9s generator buses.
or exoressed in t h e form
Substituting Eqn. (21) for A P9s into Eqn. (4) leads t o

A3 = blT [Ngs]AV
(ii) Line Overload Prevention Condition

l h e line overload prevetion condition is defined a s the

Inequality Constraint (4.5). The overloading in a transmission
line can lead t o system collapse in an extreme case. The
following condition must be satisfied for preventing the line
overload; a s a security constraint or security index:
"

Z
n
AP ( UQ
where: bnT
0

b ' l (NqJ = (b , b2, . bid.

Q ..?
1=1
kei gi 1 he control variable AVi must satisfy the following
conditions:
M
where V y ( V p tAVi<Vi (23)
is the line flow through line Q
ZQ Thus, by introducing the new nonnegative variables Xvi. Eqn.
is the upper limit of transmission power of line Q
UQ (23) is reduced to:
and KQi is the sensitivity between line flow of line Q and
0<Xvi( v y -vm (24)
power injection of bus i (i.e. AZQ/APi)
where:
Substituting from Eqn. (8) for AP in Eqn. (18), we get:
9i Xvi AVi t Vp - Vm. (25)

n
m
and AVi = Xvi - Vy + V m
2
Q
t ?
i=l
KQi(X
gi
- P"gi t P . ) (UQ
gi
Substituting Eqn. (25) into Eqn. (22) and eliminating
the constant terms, the linear objective function , with
respect to, the new variable is obtained as:
or
n n AJ = b X v
CONSTRAINTS

l h e problem size can be reduced by taking into (i) Conditions on Reactive Power
consideration only the constraints of the lines whose loading
is approaching their maximum rating IJQ. I t can be taken as Separating the Eqn. (21) into two parts, one pertaining t o
the voltage-controlled buses and the other pertaining t o the
90% of normal loading (6.7).

1 I: [
load buses, we get:
( 5 ) Upper/Lower Bounds on Each Generator Output Power: AQgc
Rewriting Eqn. (17) in a compact form leads to: (27)
where [AQQ = -AV
O I X t P
-P" (20)
p- 9 9
1 he Simplex Method with Bounded variables handles [Lgc]and[LQ] a r e the matrices consisting of the
these constraints on the control variables implicitly without
increasing the problem size. rows of the matrix [L] pertaining t o
the voltage-controlled buses and the
(iv) Environmental Impact (91: load buses respectively.
For the voltage-controlled buses, the reactive power
Generation emission can be taken a s an index for must be inside the permissible range given by:
environment conservation. The following condition must be
sa tisf ied:
Qg"c
5 Q i c + AQgc -
< Qgc
M (28)
-.
C-P For the load buses, the reactive power cannot be changed: i.e.
2 G AQQ = 0 (29)
a tf3P t y P t Eexp V
G G Substituting Eqn. (27) into Eqns. (28). (29) leads to
where a, 0. y. 5 . E are coefficient of generator
emission characteristic and V is the allowable upper limit. Qic + [Lgcl AV 5 Q E

QZc t EL 9c1 AV5Q;

2. _OPTIMUM SCHEDULING OF REACTIVE GENERATION
['-a] AV = O
The gbiective function of the optimum scheduling of the or in the form of Eqn. (25) a s follows:
reactive generation is obtained a s follows:

Setting A 0 = 0 in Eqn. ( I I), the changes in the active

power and reactive power caused by the bus voltage change
can be expressed a s follows:
AP = [Ngs] AV
AQ = [t.] AV (21)
where:
734

(ii)Upper and Lower Bounds on Each Bus Voltaqe: (iii) Bounds on Xci and XLi

Rewriting Eqn. ( 2 4 ) in a compact form such as:

The bounds on XGi and Xei can be presented as:
O<Xv5VM-Vm (32)
which can be implicitly treated without increasing the 0 5 xci 5 Pzi- Pgi
problem size.
0 5 xu 5 ;x
3. CONTINGENCY ANALYSIS where
;X i s the upper l i m i t of load shedding quantity
When it is impossible to dispatch the load without
overloading the lines, the solution of the linear programming AP,-i a t bus i.
problem in the real power optimization procedure appears to
be infeasible. In such cases, the solution can be feasible by SECTION REDUCTION METHOD
assuming the possibility of load shedding. In this case, the
objective function i s written as the sum of the following two A t the first iteration, the temporary section of each
terms: the first is the objective function in the real power control variable is set to be the part of the original range
optimization procedure, i.e. Eqn. (IO) and the second is a within the distance of AD from the initial point of the
weighted sum of load shedding quantities given as: control variable, where D denotes the width of the original
range between upper and lower limits of the control variable.
n n The value A i s determined according to the estimated
1
3= ?
k=l
b' X
k gk
t
j=l
E (WF).
1
X
Qj
(33) maximum distance between the initial points and the optimal
points (0.1 </h 5 1.0).
where: As the iteration i s repeated, the temporary section of
XQj = APQj is the variable representing the load the next iteration is reduced to k times as wide as that of the
shedding quantity APQj of the load bus j. previous iteration.
which is nonnegative and needs not be In general, it is reasonable to choose k as 0.5. but may be
replaced, and somewhat adjusted in accordance with the circumstances.
The constant k is called the Section Reduction Factor.
(wF)j i s the weighting factor assigned to reflect
the load priority of load bus j. which must The procedure for extremizing the problem of n
be made greater than the largest values o f generators i s depicted i n Figure (1).
b' for k = 1. 2. n ......
9s'
THIRD SIMPLEX METHOD AND SOLUTION PROCEDURE
CONSTRAINTS
It i s possible to develop the modified simplex method by
(i)Real Power Balance Condition:
reconstructing an identity matrix from the tables a t every
iteration by extracting the unit vectors and rearranging
Since the power change quantity of a bus equals to them. By using this developed method, the execution time
the sum of the power generation change and the load change.
and memory size are substantially reduced as a unit vector i s
we can write
generated a t every iteration instead of the identity matrix.

AP =APc t A P L (34) I n general, the linear programming problem is written as

APS = APcs t APLs (35) follows.
In analogy with Eqns. (16). (17). it i s possible to Minimize J CZ
rewrite: Subject to [A] X 2 b with X >_ 0
A P c s + APLc = [HSI [HI (APc t APL) (36)
where:
or i n view of Eqn. (8) and the definition of XQ AP L, f A = [B N] is the constraint matrix.
B = (C. C) i s the Cost Coefficient Vector,
XCs - PEs t P& iXQs = [HS] [HI (Xc - P E + P; t X $; or x = (X. X) is the variable vector,
b i s the basic matrix,
Etis] EH 1 (xC N is the non-basic matrix, and
XQ) - xGS - xQS B. N are the underscripts pertaining to basic and
non-basic variables respectively.

(37) A t the initialization step, the slack and surplus variables,

and the artificial variable are introduced as shown i n Table
(ii) Line Overload Prevention Condition
(1).
In the same manner as i n the case of real power A t the main step, the leaving and entering variables are
optimization, we get: determined. and, after the pivoting operation is carried out,
Table ( 2 ) is obtained.
n n
b o m b
Z + E K (X E K X < U ,or Comparing Tables (1) and ( 2 ) . it is observed that the unit
Q i=l Q i Gi -'Ci) j = l Qj Q j- Q vector of the r - t h column has been transferred to the k-th
column. 1P we exchange the unit vector of the k-th column
n n n for the new r-th column vector, we obtain the same result by
b b b o m
E K .X . t . E K .X . < U - Z t E K (P - P ) operating with only the right side part of the matrix [ 1 B- 1
i=l Ci
Qi j=1 QJ QJ Q Q i = l QiCi Ci
N I.
 735

Iteration
- - (11 Iteration 111

I
1 (pG1)

Iteration1

Generator 111

Figure (1) Graphic depiction of the Section reduction Method.

T a b l e (1): Variables a t t h e F i r s t I t e r a t i o n
The number of operations per iteration and the required
memory size of the Simplex Method, the Revised Simplex
Method, and the Third Simplex Method are shown i n Table (3).
From the table it can be seen that if m is significantly larger
than n, the Third Simplex Method results i n a substantial
saving i n the executing time and memory size.
0 0 0 ... 0 0 ...zj-c j . . . zk-ck . . .
The iterative technique for the economic load dispatch
...Y ... ... optimization procedure i s shown i n figure (2).
x1
1 0 0 ...... 0 Y lk

x2
0 I 0 ...... 0 ...Y Z j ... YZk ...
x3
0 0 I ...... 0 ...Y J j ... YJk ...
. .
Method Operation Memory size
. .
...Yrj ... Yrk ... Calculation Pivoting Auxiliary Total
. . I I I I I
. .
xm
0 0 0 ...... 1 ...Ymj ... Ymk ... Simplex Multiplication (m+l) (m+n+l) (m+i)'+mn+n ( m i l ) (rntnti)
L- Method Addition m(m+n+l) m(m+nt I )

Revised Multiplication (m+I)' mn (m+i)'+mn

(zj-cj) Simplex ( m t 1)'+n+ I

0 0 .. ck-zk/Y* .. 0 /I-Y ri.(Z k-Ck)/Yrk 0 Cgb-(Zk-Ck)br/Yrk Method Addition m(m+i) mn m(mtnt1)

XI
1 0 .. -Yik/Yrk .. 0 !..Yrj-(vr{Yrk)Yik 0 1
I;
bl-(Ylk/Yrk)br
Third Multiplication (m+l) (n+2) ( m t l ) (n+2)
......

Y ./Yrk I
I
I
Simplex

/Method /Addition

I
1
I
m(nt1) 1
I
tn(n.1)

I
1 I
(m+l) (ntZ)+m+n

where m is the number of constraints (number of rows o f the constraint matrix A)

.... -Ymk/Vrk .. and n is the number of original variables.
736

Set Initial Conditions of Table (5): Bus voltage and Connected Load
P,, and VI
-
Bus
No. voltage at
(Per unit) Iteration

-
1 2 Not specified 0.2 + j0.097
2 2 Not specified 0.3 + j0.145 0.9737
3 2 Not specified 0.2 + j0.097 0.9650
4 2 Not specified 0.3 + j0.145 0.9726
Perform the Real Power Optimization Algorithm 5 2 Not specified 0.2 + j0.097 0.9996
and Update Poi to PZ+AP$ 6 3 I .o 0.3 + j0.145 1.0123
1---- 7 3 1.o 0.15 + j0.0726 0.9875
1 8 3 I .o 0.2 + j0.097 1.0125
S o h the Power Flow of the System on the 9 3 1.o 0.2 + j0.097 0.9882
Update Values of Pi 10 1 1.05 0.2 + j0.097 1.0175
.-

Calculate b”, [NI. [ L ]

I Solve the Power Flow of the System on the

Update Valuer of Vi I
I

Figure (2): Iterative Procedure

RESULTS ON EXAMPLE (1) SYSTEM STUDIES

A IO-bus model system shown in Figure (3) was used to

test the method of scheduling real and reactive power and t o
study the convergence characteristics of the optimization Figure (3): 10 Bus Model System
process prsented in the paper. RESULTS ON EXAMPLE (2)SYSTEM STUDIES

The impedance and line charging data is given in Table The proposed optimization algorithm is applied to the
(4) and the bus voltage and load data a r e given in Table (5). IEEE 30 Bus test system with 6 generators and 41 lines given
The operating limits and cost data for each generator and in Fig. (4). The system data a r e given in Tables (7)and (8). as
real power generations of each iteration a r e shown in Table obtained for Reference (9). The results using new algorithm
(6). In consequence of continuing the iteration procedure until are shown in Tables (9)and (IO) for the optimal solution of
the cost flucturation AC becomes smaller than 0.05% of the subproblem as compared to Table (1 I) of Reference (9).
cost, the number of load flow calculation is 3 times. The numerical results on this power system have verified the
validity of the algorithm with respect to the existing oms.
The memory size and execution time using the Third Simplex
Table (4):Line Impedance and Charging Susceptance Method has been reduced tremendously.

-ine Bus Line Line charging

lumber Number Impedance susceptance
1 1-2 0.02 + j0.08 0.03
2 1-6 0.06 + j0.24 0.02
3 1-9 0.04 + j0.16 0.015
4 2-3 0.06 + j0.24 0.02
5 2-6 0.06 + j0.24 0.02
6 3-7 0.06 + j0.24 0.02
7 4-7 0.04 + jO.16 0.015
8 4-8 0.06 + j0.24 0.02
9 5-6 0.04 + j0.16 0.015
10 5-10 0.06 + j0.24 0.02
11 6-9 0.01 + j0.04 0.01
12 8-10 0.04 + j0.16 0.015
13 9-10 0.08 + j0.32 0.025
Fig. (4): 30 Bus Test System 191
 737

Bus Bus Rating of each senerator Convergence of Power generation 1

No. NO. Fuel Iter me1 Iter

I
2 1 I I
1 6 27+6P + P 1.5 0.05 0.6 30.960 0.05 27.303 0.4153
61 g1

2
2 ? 3 5 + 1 0 P + P 1.5 0.05 0.45 39.702 0.05 35.503 0.05
g2 82

2
3 8 29 + 5 P + 0.5 P 1.5 0.05 0.35 30.811 i.075 34.953 1.0421
e
3 s3

2
4 9 31c8P + p 1.5 0.05 0.5 35.250 0.05 31.403 0.05
sb 84

2
5 10 28 + 6 P + 0.5 P 1.5 0.05 0.37 30.289 1.1445 35.522 0.762
s5 g5

Total Fuel Cost I 167.012 164.682

CONCLUSIONS Table (7): Specified bus data

~-
Bus Type Active Power Reactive Power JS Voltage
- ~~~ ~~ ~

An optimization technique has been presented for the

1 P-Q -0.106 -0.0 19 -
economic allocation of real and reactive power applying the
linear programming method and an algorithm has been 2 P-Q -0.024 -0.009
3 P-Q 0.000 0.000 -
developed for finding a post-emergency schedule with the -
minimum of load shedding. 4 P-Q 0.000 0.000
5 P-Q -0.035 -0.023
Satisfactory results are obtained by adapting the 6 P-Q 0.000 0.000
program to the IO-bus model system and the 30-bus model 7 P-Q -0.087 -0.067 -
system and found that i t is very similar and even better to 8 P-Q -0.032 -0.016 -
9 P-Q 0.000 0.000 -
that of other optimum dispatch methods.
10 P-Q -0.175 -0.112
The technique of linear programming is shown to be a 11 P-Q -0.022 -0.007
12 P-Q -0.095 -0.034 -
powerful and practical tool for obtaining an approximate
solution of a linearized optimization problem. Above all, the 13 P-Q -0.032 -0.009
Section Reduction Method and the Third Simplex Method may 14 P-Q -0.090 -0.058
have applications elsewhere. 15 P-Q -0.035 -0.018 -
16 P-Q -0.082 -0.025 -
But, the confirmation of its usefulness must await 17 P-Q -0.062 -0.0 16
further testing on actual power systems. Extensions and 18 P-Q -0.1 12 -0.075
refinements on the present algorithm are expected t o include 19 P-Q -0.058 -0.020 -
reliability indices. As an example, the algorithm relies on the 20 P-Q 0.000 0.000
convergence of the power flow at each iteration and 21 P-Q -0.228 -0. I09
therefore may be interrupted by a single solution divergence. 22 P-Q 0.000 0.000
The efficiency can also b e improved by exploiting sparsity of 23 P-Q -0.076 -0.0 16
the problem. 24 P-Q -0.024 -0.0 12
25 P-v 0.000 0.000 1.071
Table (8): Line flow capacity 26 P-v 0.000 0.000 1.082
27 P-v -0.300 - 1.010
- 28 P-v -0.942 - 1.010
29 P-v -0.217 1.045
Line No Line NO. P: Line NO. psi 30 5 0.000 0.000 1.Ob0
P:
I I I I -
1
I I I 1
15 29 0.20
2 16 30 0.20
3 17 31 0.20
4 18 32 0.50
5 19 33 0.50 M.S. Bazaraa. 3.3. Jarv. "Linear Programming and
6 20 34 0.50 Network Flows", John Wiley & Sones, pp. 1-219. 1977.
7 21 35 0.15
8 22 36 0.50 D. Leven, Y . Wallach. V. Conrad, "Mathematical
9 0.15 23 37 0.50 Programming Solution of Optimum Dispatch". Power
10 0.50 24 0.50 38 0.50 Technology Computer Applications Conference (PICA),
11 0.50 25 39 0.50 CH 1318-3, pp. 137-141, 1979.
12 0.50 26 40 0.50
13 0.30 27 0.20 41 0.50 B. Scott, J.L. Marinho, "Linear Programming for Power
14 0.50 28 lP.U.1 System Network Security Applications", IEEE Trans. on
Power Apparatus and Systems, Vol. PAS-98, No. 3, pp.
Line capacity is 110% of standard value (Ps) 837-848. May/June 1979.
738

kl Bul
IO. w.

1 30

2 29

3 26

4 27

5 26

6 25

Total CO!

Emission Characteristic of generators

2 eR: Total Iter
cm Bus F = a + b P C c P + d e w Initial tm/h 1
no. BO. C G

-2 2 -4 2.851 PG
1 30 10 (b.091 - 5 . 5 5 4 P + 6 . 4 9 0 P ) + 2910 exp 0.45 0.029'8 0.38
c c

-2 2 -4 3 . 3 3 3 PC
2 29 10 ( 2 . 5 4 3 - 6.047 P + 5.638 P ) + 5*10 em 0.50 0.01194 0.51
G G

-2 2 -6 8 . 0 PC
3 28 10 ( 4 . 2 5 8 - 5 . 0 9 4 7 c 4.586 P )+ l*:O exp 0.60 0.02685 0.51
c c
-2 2 -3 2 . 0 PG
4 21 10 (5.326 - 3.550 P t 3.38 P ) + 2'10 em 0.PO 0,04892 0.42
G c
_____
-2 2 -6 8 . 0 ZG
5 26 10 (4.258 - 5.094 P + 1.586 P I + 1*10 exp 0.60 0.02865 0.53
c G

6
-2
10 ( 6 . 1 3 1 - 5.555 P t 5 . 1 5 1 P ) + 1'10
2 -' 6.63'
''r.30 0.04936 0.44
"p
c G
2 5 1

Minimized objectives Generation Emission Line

cost minimum overload
minimum index
minimun

Total generation cost ($/h) 606.04 645.88 640.62

Total emission (ton/h) 0.2350
Total flow deviation (PA.) 0.0381 0.7778 -
0.0

1 0.124 0.381 0.589

2 0.310 0.515 0.300
enerator output 3 0.543 0.562 0.525
(P.U.1 4 1.016 0.399 1.055
5 0.514 0.522 0.355

(IO) J. Barras. S. Alec. C. Pasche, P.A. Chamorel. "Network

Simplex Method Applied to AC Load-Flow Calculation".
IEEE Trans. on Power Systems, Vol. 21. No. 1. Feb.
1987. pP. 197-203.

(1 1) E. Habsen, D.L. Fletcher, W.O. Stadlin. "Network Flow

Linear Programming Techniques and their Application
t o Fuel Scheduling and Contingency Analysis." IEEE
Trans. Power App. System, Vol. PAS- 103. No. 7, 3uly
1984.