Вы находитесь на странице: 1из 9

IEEE Transactions on Power Systems, Vol. PWRS-2, No.

2, May 1987
A NEW APPROACH FOR MINIMIZING POWER LOSSES AND IMPROVING VOLTAGE PROFILE

287

J. Qiu

S.M. Shahidehpour

Electrical & Computer Engineering Department Illinois Institute of Technology Chicago, IL 60616
Abstract - This paper presents a new method for minimizing transmission line losses and improving voltage profile in a given system by adjusting control variables, i.e., tap position of transformers and reacTransmission tive power injection of VAR sources. losses are considered as a function of voltage increments. The control variables and voltage increments are related by a modified Jacobian matrix. Linear Programming (LP) is used to calculate the voltage increments which minimize the transmission losses, and the adjustments of control variables are obtained by amodi-

distribution. He used incremental transmission line equaand transformer methods and linearized networks tions. The problem was solved by a special LP technisystem. que by giving priorities to generators in the is to Because the main purpose of these two methods find the values of control variables which would cause the dependent variables to vary within a certain limit, these two methods can also be used to study contingency analysis. They, however, are not suitable for finding minimizes an optimal dispatch of reactive power which the transmission losses of the system. fied Jacobian matrix. Since this method does not need any matrix inversion, it will save computational time Mamandur and Chenoweth [5] developed a method to and memory space. profile minimize system losses and to improve voltage by controlling generator voltages, transformer tap setINTRODUCTION tings and adjustable VAR sources. The method employs linearized sensitivity relationships to minimize system In the last twenty years, more attention was paid losses and the system performance sensitivity relating to optimal reactive power dispatch, because the reallolinear dependent and control variables. It is a dual adjusting programming technique to determine the optimal adjustcation of reactive power generations, by transformer taps, changing generator voltages and by ment to the control variables, simultaneously satisfyswitching VAR sources, can approach two goals: ing the constraint. Unrestrict step sizes are used for the dependent variables during the first iterations and 1) To improve the voltage profile. restricted step size for the subsequent iteration. The power flow calculation was performed at the end of each 2) To minimize the system losses. Ramalyer et al [7] also presented an algoiteration. rithm to minimize power losses and to improve voltage Power system operator should ensure the quality and reliability of supply to the customers by maintaining profile. Power flow calculation is not included in each iteration and the algorithm incorporates a method to the load bus voltages in their permissible limits. Genavoid zigzagging of the solution around the optimal erally, power losses in the transmission of electrical point. The dependent variables are eliminated from the energy cause a loss of revenue. So, even a small percentage of savings inloss will be very much appreciated problem by means of the sensitivity matrix, based on since the total generated power is on the order of power flow equations. Both of these methods need to megawatts. This paper is mainly concerned with optimal inverse or partially inverse the Jacobian matrix to dispatch of reactive power for minimizing transmission obtain sensitivity matrix, which is a time consuming process for the large scale systems. losses. Elangovan [8] developed a new algorithm for the In the past, several methods using sensitivity relationship have emerged to solve the complex problems. evaluation of the sensitivities of line losses with of respect to the control variables, i.e., voltages reacThere are mainly two different methods to study generators, tap settings of the transformers and reactive power planning: nonlinear programming and linear programming. Shoults and Sun [1] developed a nonlinear tive power of VAR sources. These coefficients are oboptimizing strategy based upon the gradient method tained by the eigenvector of the transposed Jacobian employing the sequential unconstrained minimization matrix. Although this method does not require the intechnique. To make the method suitable for nonfeasible version of the Jacobian matrix, it may not be easy to starting point, an out-side-in penalty was chosen to find the eigenvector for a larger matrix. force the dependent functions to be feasible. Since This paper presents a new method to find optimal linear programming is a fast, reliable, versatile and state of reactive power dispatch which minimizes the accurate method to solve the optimal sufficiently transmission losses and improves voltage profile of dispatch problem, more work has been done in this area power systems. The constraints are on transformer-tap and Chen [2] developed an Shoults in recent years. positions, VAR sources and bus voltage increments. The find algorithm using least-square minimization, to transrelationships between control variables, i.e., suitable changes of dependent variables such as voltage former-tap positions, VAR sources (generator reactive magnitudes and currents flowing in branches. Hobson [4] power outputs and switchable capacitors), and bus voltage increments were derived. The correlations between these variables are given by a modified Jacobian matrix. LP techniques are used to find optimal voltage profile all which minimizes transmission losses and satisfies A paper recommended and approved 86 SM 346-1 the constraints simultaneously. The required adjustof by the IEEE Power System Engineering Committee ments of control variables are obtained using modified the IEEE Power Engineering Society for presentation Jacobian matrix. The searching process will stop when at the IEEEIPES 1986 Summer Meeting, Mexico City, the difference in real power generated by slack bus for 20 - 25, 1986. Manuscript submitted Mexico, July a prescribed two consecutive iterations is less than available for printing September 4, 1984; made value. Since restricted step size is used on each variMay 1, 1986. able, zigzagging will only be in a small range and would
er

presented

method of finding the network reactive

pow-

Printed in the U.S.A.

not cause large errors.

optimal real

power

The state of the system with initial generation schedule is the

0885-8950/87/0500-0287$01.00O1987

IEEE

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

288 condition for this technique.


LIST OF SYMBOLS

Note: The symbols with a bar on top, e.g., V, represent a matrix or a vector and symbols with a dot, e.g., V represents a phasor.
N : number of buses in the system Vi : voltage of bus i
6.: voltage angle of bus i
1
.

except at the slack bus. To formulate the objective function and constraints, the load and tap-changing transformer models have to be introduced at this state.

Load Model
The losses of the system are reduced by the variations of voltage profile. The variations of voltage profile is controlled by generated reactive power distribution in the system which will also change the reactive power received by the system load. Hence, it is important to consider the variations of load with the changes of voltage profile. Reactive load can be represented by the following equation:

complex power injection at bus i

P. : real power injection at bus i


.~~~~

A. : reactive power injection at bus i


power flow from bus i to bus j where

QL QLS( QL= QLs


"

Vq

)q

(1)

S.. =

Pij +Qij:

T.. : the setting for a tap-changing 3 transformer between buses i and j : transmission loss of the system pL

0 q = { 1 2

for constant power load for constant current load for constant impedance load

y.. : admittance of the branch between '3 bus i and bus j

apd QLS, VS are the value at base condition. The load changes due to the variations of voltages may be calculated by the following equation,

QiG

: reactive power generation at bus i


:

QiL
Q.

reactive load at bus i

AQL=

aQL

V sAV=

q * QL

q- 1

v~ ~ ~

* AV.

(2)

: initial value or base value of reactive power injected at bus i


:

Tap-Changing Transformer
of a tap-changing By adjusting the tap-setting transformer we can improve the voltage profile and retransactive power distribution. The tap-changing former can be modeled using one of the following two

QiT

tap position adjustment : number of buses with reactive LI power sources L2 : number of tap-changing transformer in the system L : coefficient vector for the

injected power equivalent

to the

techniques:

objective function A : coefficient matrix for the constraint equations Jacobian matrix A1 :

V.

1:

ZT.

V.

A2 A3

the coefficient matrix considering the load effects : the coefficient matrix considering the effect of tap-changing transformer. METHODOLOGY

I
AQiT
=

iT

ij 11

A.

AQjT

X,_
13

AT

ij

The optimization problem is described by the following equations: = -TL -AV minimize

APL

Figure 1
ure 1.

subject to AQin

AQ

A.AV

<

A%1a

1)

The voltage magnitudes are mainly controlled by Also, the reactive power distribution in the system. the reactive power distribution is heavily influenced by the real power flow. Therefore, the proposed method is applied after the real power has been optimally distributed. At this point, it is reasonable to make the following assumptions: 1) bus voltage angles are assumed to remain constant in each iteration of optimization. 2) real power injection at each bus is constant

When

process. Assume that the power flow from bus i to bus j denoted by

Tij has an increment equal to ATij the power flow through zT.. will change. We model this situation by 1iJ and j which two reactive power injections at nodes i flow causes the same amount of reactive power to through zTij as Tij changes to Tij + A Tij. Here, the assumption about constant voltage will cause some errors because Vi, Vj may change during the optimization

Consider the tap-changing transformer model in Figconstant. Assume that voltages Vi and Vj are

ij

is

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

289

S.. ij
Note

**

T.

.-V. -I. .-T. .-V. [(V. -T.

-V.)-y.

AQiT
AT
=

T. 1j
1 ZT

y. .(V.
13

T..-V.*V.). 13 1
b
6

bij -Vi
/
6

Tij

.sin6 ij-b.ij V.i v.(gij

cos6ij)
(6)

ij
so
S.. 13

1= 13131J
*

jT
i
-V.
1

V. = V.

.V.

(gij.

sin

6..
13

b.
13

cos

6..)
13

V. = V. /
*

6..

- 6

-6

~22 (g.13 .-j b. .)(V.1 *T. .-V. V. *T.13 .cos 6.. 1j 1j 1 1j jV. *V. *.T . *sin 6..)
-

If we know the value of AQiT3 or AQjT which are used to determine the optimal voltage profile, tap position could easily be calculated by equation (6).

13

13

[g.

.*(V. *..

-V.

-V. *T.
-

.cos
1

2) The alternative way to deal with tap-changing transformer is to represent it by a "II" circuit, as shown in Figure 2.
V.

* .V. *V. *T. . *sin 6 ..)]


13 13

V
1:
..

V.

j[-g. .*V *V..-T .-sin 6.. 1


(V3
T~
1

b..
1
*cos

~yijj
6..).

ij

V. -V. -T.
1

ij

ij

The reactive power flowing through


Q..
=

the

transformer is,

-[gi.
-

V.V. *T. . *sin 6.. + b.. *(V *

V.

V. -V. T. .*cos 6..)] 1 13 13

T3.
ij~~1

y.

The derivative of

Qij

with respect to

Tij

is

aT13
is,

3Q. .2
=

-2b

ij.ViTij
-

Vi Vj. *(gij *sin

6ij
(3)
to

bij
-

cos 6ij ).

(T3

T13

(1-T ij.)Y.

In the same way, the power flowing from bus j


S.j
=

bus i

V. 'I..
(gij

V.

*[(Vj

Ti
Tij .V..V.

bij) (Vj

cos

6..

Figure 2.
It is obvious from Figure 2 that,
11

+ j

T.j .Vi Vj-sin 6ij)


Ti -V -Vj
- T. 13
cos

[ij * (V.

6 ij)
6..) 13

S.

P.+jQ.

.(V
2

S(Tij
_jV

~~2*
-

Tij) Y)

jb
+ j

(Vb

-V. -V.*cos 1 3j

(Tij- ij)
2
2

2.

gij ji

(ij

(T~.-

T..).b..

i)

and,

gij*(j -Ti j V*Vi-Vj *si co 6 .]i + .**T. . '.sin


13 13

Qi

-b

.Vi (Tij -Tij) ij

13

Here, we will consider the effect of voltage changes on the calculation of AQiT and AQjT. The derivatives of Qi with respect to Vi and Tij are
aT

cos

aQi
-b

2
.V..(2T..

1)

aQ. '
a
V.

V(gij

6ij

bij *cos

6..).

(41

AQ
1.

-b ij =

Vi (Tij
V

If ATi- is given, AQi. and AQji can be calculate(d using equatlons (3) and (43. Note that AQiJ is from bu! i to bus j, and AQ-i is from bus j to bus i and both ar( originated from ATij. According to Figure 1, we cai write the following equations:

Since

AQiT
AQ

-AQi
A
1

aQ.
13ij
V

AiT

...

(7)

AQ iT

_AQ.
+

i3 3Ti V.V * (g'i sin6..-b.


1

13 =

=[2b
.cos6
[-V.

TQ

VoT..
13
1

2bi

(T.

13

.)]AT..
V.,

And, S.

=
=

Pi+jQ.
2
V

= V.

*[ (l-T

13.

.-T.

AVi

bi

.V2(2T.

-1) -AT...

ij).Yij
=

Vj ]

Qji QjTAQ= -AQ


.(g.j .sin6.

aQ.
aT
.

.(l-T

AT..

ij).(gij-j bij)
)*b..

ij
+

ij

i j

V.(1-T ij)'g.ij

b.

.ocos6ij) ]-AT.
Q.
=

jV. *(l-T.
'V2.

and

-b..
1

(1-T.. ) 1

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

290
3)

Adding Tap-Changing Transformer Factor

aT ij
SQ.
and

23

aQ

21~~~~~~~ 1jT = -AQ =


=

-Q 2

ij*v

av'

-T

AV. SQ. . .) ,

1T.i

AT. a ij
2

2-b.13.-V..(1-T 121.) -AV. - b.12 VV.2 *AT.11j. .

(8)

Having derived the models for load and tap-changing transformer, we are ready to formulate coefficient matrix , the limits of constraints and the objective function.

Coefficient Matrix A Matrix A is a NXN matrix which gives the relation between voltage increments and reactive power injection increments. The formulation of matrix X follows three steps: forming Jacobian matrix, adding load effect and adding tap-changing transformer effect.
1)

changing the tap As it was mentioned earlier, position is equivalent to the injection of reactive power into one of the buses which is connected to the We will consider two distinct transformer terminals. cases for the buses connected to the terminals of the tap-changing transformer: 1) At least one of the tapconnected to a changing transformer terminals is not generator or switchable capacitor bus. So, for example, if a tap-changing transformer is connected between bus i and bus j, the assumption is that the bus j has no generator or switchable capacitor. In this case, if AQ is injected into bus j, a proper value AQiT will also be injected into bus i and there is a defand inite relationship between the values Of AQiT AQi determined by AQj. The value ofby AQiG, therefore, reactive power generator is denoted

AQj
then

jT

AQi = AQiG

AQiT

a2j I,
=

2jN
..

[a 2il

a2iN

AV

Forming the Jacobian Matrix

The portion of matrix A, which is formed by Jaco, bian matrix is denoted by 1I From equation AQ = X AV, we notice that the elements of T1 are partial derivative of Qi with respect to V. in a linearized model of the power system. Using the well known equation of
= I yij Qi 1 Vi j=111 Vj .sin(6ij.12

QQiT ... a2iN] * we will have AQiG = 0 if bus i is not connected to any The effect of tap-changing reactive power source. transformer in formulation of matrix A is discussed as
follows:
a)
for the model represented by Figure 1,

AQiG= [a2il

E0.)
11

Since bus j is not connected to any generation unit, from equation (6) we will have

we will have

ATij
will have

V. *V. (g.
1

21

12

.sin6.. 121

b. cos6 jT) 121 121


(5), we

aQ.
SV. - =
i '2

Substituting the above equation into equation


2

ii
a

21

V *y
1

sinn
11 1

i,j -1, -

A'

-(2b. .-T .V1 +gij V.1 V 2 Sin6. .-b. .V.-V. cos6.. 13 1j 1j 12 1 1j1 V.1 V. (g.121.sin.13 + b. .cos6..) 121 121

where ali 2)

is the

SV.

ij -th element of matrix A.

AQjT

H*AQjT.

(9)

Adding Load Effect

Since the voltage at each bus is a function of load and generated power at the bus, if we want to study the variation of the reactive power with voltage, it is necessary to consider the effect of load separate from
the effect of VAR sources and tap-changing transformer in formulation of X matrix. We call the matrix considering load effect as T2. If generation and load are only sources of reactive power, then

Let A3 be the matrix which includes the tap-changing transformer effect, so A2 can be modified as a3k = a 2ik + H-a 2jk k = 1, ..., N 3k where i,j represent all the rows which correspond to the buses i and j in Figure 1. The elements in other rows are unchanged.

b)

for the model represented by Figure 2.


From equation (7) and (8), we can write
121

AQ - AQL A1AV AQcG A1 AV + AQL from equation (7) AQL = [ - ]

AQ

b.
13

.(2-b.
j

V.

(l-T ) -AV.
2

means the matrix is a diagonal one and

AV where

subscript D

and

AQiT = 2'b...V.(T.1 -Ti ) AV.+b.121.V.i(2Tij 1) 1 13 121 1 1 1


b.

VLi iLis SV.1 =qi QLi


Therefore
and
G-

iS

7)
+

1 V2 (2b. *V .(1-T.i ).AV. - AQ.) ij j jv i V2


Vj -1 . 1Vi 21

AQ =A_-AV= (A

S~~V

D V.

AV
q

=2*b.21.'V.-(T.2 .-T.13 )AV.+2b.b 1


(2 -T.11 -1) 13(l-T. -)AV 21 -

V2

a2ii = a.ii + qi Qiis

V1 is
..
N

-.1
.

2 V2

(2T.-1).AQ.. 1 3jT'

.2ij

alij

To consider the effect of tap-changing transformer in formulation Ta, T2 can be modified as follows:

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

a3. 3i
a

2ii -2

bi 3 jij'3 IV1 (T.j -Tj) -2 (2T -1)*a.2ji


-V.
+

.~ ~
3

291

V2
=a

2 b *.
1~~~~~

V.

(2 T

1) * (1-T )

optimal values of voltages should be in the equation. Since these values are unknown at this state, the optimal value is considered to be 1 p.u. There are two different conditions for buses numbered from 1 to L1 + L2. First kind of buses are the ones from 1 to L1. These are VAR source buses. Generally, Qmax' Qmin are given for these buses, therefore,
where s indicates the initial value. For the buses represented by Figure 3, the injected power is determined by reactive power sources and the tap-changing transformer. The limits on AQi and AQj are formulated as follows:

AQmax Qmax Qs

AQmin Qmin Qs

3ik= a2ik

2
i
k
=

2 (-l).(-T. ilk

1,

N 1..,

i, j

The values of i and j represent all the rows which correspond to the buses i and j in Figure 2. The other elements of the matrix '2 remain unchanged. So, if the values of AVi and AQj are determined by solving the linear programming, ATi- and AQiT will be calculated according to the equations which were derived in parts a) and b). After determination of A3. the constraints equation is as follows, AQ = A3'AV. For the generation or switchable capacitor buses, AQ is the increment of reactive power that should be injected to the bus and for the buses which are non-generation terminal of tap-changing transformer, AV is used to find the tap increment. 2) If both terminals of the tap-changing transformer are connected to the reactive power sources (Figure 3), a sub-linear programming formulation will determine AQjc, AQjT and AQjG. This technique is discussed in the following sections.

AQi
so$0

AQiG + AQiT

AQ iT=

H.AQjT

AQ. = Qj

AQi

AQiG

+ H +

AQ.j AQ.j T Qjc Qj

AQ jT

so the limits are,

AQ. jmax =AQ. jTmax + AQ. jcmax

AQ. . =AEQ. Tmin + AQ. cmin j jmin j

'Qimax
AQ.mm
=

QiGmax +QiTmax AQiGmin +QiTmin-

The other kind of buses are the ones which are numbered + from L1 1 to L1 + L2. Since Tmax and Tmin are given for every tap-changing transformer, we have
AT . = Tm. - T ATmax = Tmax - T s mmn mmn s AQmax, AQmin will be calculated from ATmax, ATmin and this calculation is discussed as follows: a) for the transformer model of Figure 1: From equation (4) we have AQJT = -V.1 .V3 * (gij *sin&.. + b.1 3.cos6) AT... 13 13 13 3 << Usually for a power transformer, r xi so Yi-= jbij also bij < 0, and 6ij < 900 which leaa to the iollowing equation

j
1 : T..

0-HA~

Qj c
Figure 3.

AQjT =Vi jVjb ij.-cos6

ATij
(10)

where (-V. V. *b.13 cos6.13. ) > 0, therefore 1 3

Determination of the Limits for the Inequality Constraints


Let L1 represent the number of buses with reactive power sources, either generators or switchable capacitors, and L2 represent the number of buses which are the terminals of tap-changing transformers and are not connected to a reactive power source. (L2 equals the number of tap-changing transformers.) Therefore, all the buses with reactive power sources are numbered from 1 to L1, where slack bus is bus #1. All the buses without reactive power source which are also the tapchanging transformer terminals (one for each tapchanging transformers) are numbered from Ll + 1 to Ll + The rest of the buses could be named arbitrarily L2. from L1 + L2 + 1 to N. These buses can be load buses, junction buses or one of the terminals of the tapchanging transformer. We have modified the matrix X to incorporate the effect of load and second terminal of tap-changing transformers (without VAR sources) in the equations. Since reactive power may not be injected into these buses, the constraints for these buses become equality constraints. To find the limits of AQ, the

AQ.jTmax = -V.i1 -V.3 -b.13.*cos6..j AT..j 13 13max V.b.-b . AQ.Tmin = -V.1 ..cos6.. * AT.. 13 13 13max
b)
For the transformer model in Figure 2: From equation (8), we have 2 AQjT = 2.b1. *V.j (l-T. )-AV. - b. -V. AT..

3j3 13

i 3

13

13

since the value of Tij is close to 1 and AV- is very small, the first term on the right hand side of the above equation is negligible compared to the second term. Therefore, we can determine the limits of AQ.T by only using the second term, so = -bij (ATij)max AQjT
.

AQj

jT

= min

2 -b..'V. 1-33

(AT..) min 13

(11)

Objective Function
The transmission loss increments of the system,

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

292
which are the functions of voltage increments are used to formulate the objective function: (12) APL= L . AV
where L = [Q

*.
N

Pi

av

MAPL

9 QN

= iI V. G.i =l'13

cos6.1

(13)

ed. The changes in reactive power sources are given by the first L1 elements of AQ and required value of AT can be calculated by equation (6) or (7) using the last L2 elements of AQ. If both terminals of the tap-changing transformer are connected to reactive power sources, the solution of the following sub-linear programming will determine AQjc, AQjT, and AQiG.
min a*
subject to

The derivation of objective function is;described in Appendix A. After modifying the A matrix and computing constraints and objective function, the optimization problem appears as follows, minimize APL = L* AV

Q.jc

+ 6 AQ T

AQiG + H*AQ. TAQi

AQ iGmin- QiG <QiGmax <AQ


AQ
. < jcmin < AQjjc -

AQi

AQjT

AQi
AQ.jcmax

subject to AQin ' AQ


A
-it
_

A3

<

AVmin- av< AVmax . < where A3 is a (L1 + L2)xN matrix. The inequality contraints in this case, are for reactive power sources and also for tap-changing transformer terminals which are not connected to a reactive source. V is a (NLl-L2)xN matrix and the equality constraints are for the buses numbered from L1 + L2 + 1 to N. In linear programming formulation the variables have to be non-negative. This difficulty is overcome by transforming the unconstrained variable to non-negative values. Define au = AV - AVmin so the constraints for voltage would change to, < <A o - AU - umax (15) where - v =AV -AV. v au
max

*AV = 0

(14)

AQjTin-S AQjTT-

AQj

ax

where a and a are the parameters which represent switching preferences. Before running sub-linear programming, the following conditions must be examined. Let AQi and AQj be the solutions to the linear programming. If AQj has reached the limit, we do not need to run the sublinear programming and the reactive power distribution for the other bus connected to the same tap-changing transformer is given by the following equations

AQ. = j

AQm 3max

AQiG = AQi - HQAQ.m


If the value Of AQiG obtained from the above equation exceeds the limit of AQiG, the value of the limit will be used for AQiG6
DIGITAL SOLUTION ALGORIHTM

AQjT AQj Tmax

max

min

max

min

The reactive power increments are, = A3 AU AQ = A3* AV = A3

(AU+AVmin)
_*

A3 -AVmin

(16)

Also, the limits for AQ would change to, AQmin = AQmin - A3 A Vmin -, -' 1.(17) Qma = AQmx -A 3 * mmn max max The equality constraints are modified as follows, (18) A 3 *AU=-A 3 *AV min . The objective function changes to, (19) APL =L(AU + AVmin) =LAU + LAVmin Since L * AVmin is just a constant, we can delete it from the objective function without changing the nature of the problem. The new objective function is, APL =L * AU
* _ _

The linearized model of power system is used, so, the method will give satisfactory results only in a small range in the neighborhood of the base condition. The range of variation of AQGstep, AQCstep and ATstep is monitored in each iteration to ensure the accuracy of the results. The following describes the steps in the process of the optimal dispatch of reactive power. Figure 4 is the corresponding flow chart.
Starti

AVmin

The optimization problem can be formulated as: I minimize APL = L * AU


_

subject to
-A3
-1

AUmax
AU <

X } min -1~~~1

(20)
20

A3

ax axj_

AU> . -A3 .AVmin 3 By solving the linear programming problem, AU is obtained. By using equation (15) and AQ = T3.AV, the increments of reactive power injection can be calculatA

Figure 4.

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

293

1)

2)

3)

4)
5)

6)

Perform the load flow calculation to determine the state of the system with optimal real power generation schedule. If the corresponding data is available, this step can be skipped. . Select the values of AQstep' AT step and AV step The basic step size variation limits are chosen as 5% for transformer taps, 0.25p.u. for voltage changes, and 0.lp.u. for power injected by the VAR sources. Q, T and V are permitted to vary within these limits around the values which are determined by power flow calculation. If any of these step sizes is beyond the limits, the value of limit will be used in the calculations. Calculate A3 A matrices and A AQm tps AUmax-step usingAQ step' AT step and AV _step rather AT, than A:m max' ax' AVmax and using_-AQstep' -AT step' provided -AVstep instead of AQmin' ATmin' AVmin' that none of these values falls beyond limits. Solve the LP problem. The modified simplex method is used in the program. Other method could also be used to solve the LP problem. After solving LP, AQG, AQC and AT are obtained. Use QG + AQG, QC + AQC and T + AT to perform power flow calculation. The results are used as base data for the next ite'ration. Check whether the real power injection at slack bus is larger than that of last iteration. If the answer is negative, the process will be repeated from the second step. If the answer is positive, the result will be printed out and process will be terminated.
EXAMPLES

optimal point. If the base case is an ill-condition, i.e., some variables are beyond the limits, there may be no feasible range in the first iteration, if the step size is too small. For the 6-bus system V6 at base case is lower than the limit so, there is no feasible range when the step size is equal to 0.5. To avoid this situation, a larger step size must be used for the first iteration and the process will continue with the original step size in the following iterations.
Table I. Line Number 1 2 3 4 5 6 7

Line Data of 6-Bus System on 100 MVA Base


Bus Number From To 1 3 1 4 4 3 3 5 4 2 2 6 4 6

Impedance R 0.123 0.080 0.097 0.000 0.282 0.723 0.000

(p.u.) X 0.518 0.370 0.407 0.300 0.640 1.05 0.133

Tap Ratio

0.9756

0.9091

Table II. Tap


T

Limits on the Variables

Voltage (p.u.) VAR Srce.(MVAR) Gen. Load Gen. Caps. I tr IrX1Itr xr r -E v1 IV2 JV6 Ql 12 Q3IQ4 Low 0.91 0.9ll.0 l.1010.9 0.9 0.9 0.9 -20 -2010. 0.01 h 1.11 1.11 1.1 1.15 1.1 1.1 1.1 100 100 5. 5.5

13

11 ,f I-

It

'I 1

1,

V3

V4

V5

Table III.

Result of LP and Power Flow at 2nd Iteration (Transformer Model Figure (1)) Power Flow _G QC V 6(deg) _G_ _G LC 37.54 - 1.06 0.0 93.9 38.3 19.80 1.10 --3.79 50.1 20.07 5.5 0.943 -11.95 5.5 50 5. ____ 5.0 0.962 -9.47 5.0 - 0.944 -12.32 _30 18. _ 10.935 -13.181 155113. 1.035 T 46 = 0.993
LP

A computer program implementing the proposed algorithm was prepared and used to test the method on 6 and 39 bus systems. The one line diagram and the base case for 6-bus system is given in Figure S. Table 1 gives line data and tap positions. Table II gives limits on taps, voltages and VAR sources. Table III and Table IV give the results of linear programming and power flow calculation of second iteration, where basic step size was used. Table III corresponds to tap-changing transformer model in Figure 1 and Table IV toi the one in Figure 2. The program was tried for different step sizes and the results show that the smaller the step size is, the more iteration are required to achieve optimal condition. With smaller step size, the losses at optimal state are smaller than the ones calculated with larger step size. It is interesting to know that different step sizes lead to different reactive power dispatch at

Bus V 1 1.06 2 1.1 3 0.945 4 10.964 5 0.949 6 0.939 T35 =

LIQL

Table IV.

Result of LP and Power Flow at 2nd Iteration (Transformer Model Figure (2))

Bus V QG QC V 6 (deg) 1G QG 'QC IL QL 0.0 93.8 39.0 1 1.07 38.7- 1.07 2 1.1 19.0 1.10 -3.41 50.1 18.6 - 5.5150. 5.0 3 0.952 5.510.957 -11.70 4 0.971 5.0 5.010.977 -9.28 _ 5 0.954 0.950 -12.05 130.18.0 ____ 6 0.941 55.13.0 0.941 -12.95
T -103 T35 = 1.033
Table V.
Iteration
Trans-

LP

Power Flow

= 461

0.991

System Losses at Each Iteration


Start 1 2 3 5 4 10.789 9.208 9.013 9.084 9.112 9.032
7
9.029

former

p(i)
L
)

Transformer Model

FModel

i-)L

-1.481 -0.295 0.071 0.028 -0.08

-0.003
8.933 8.933

IpL)

10.789 9.203

8.900

8.704 9.281

8.798

Fig. (2) jP1i)-Pi1(i

j-1.586 -0.303[-0.136 0.517 -0.483( 0.135 0.000

Figure 5.

There are several methods to examine the convergence of the solution, one of them is to consider the difference between the losses in two consecutive iterations. This approach is examined using the 6-bus sys-

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

294
tem and, when the difference is smaller than 0.01 MW, the procedure stops. The losses at each iteration and difference between two consecutive iterations are given in Table V. This table shows that the procedure converges faster when using transformer model of Figure 1, but it gives better result when using transformer model of Figure 2. It is worth to indicate that even though there is a small difference in real losses after second iteration, the reactive power distribution changes from one iteration to the other. The results in Table III, IV show that the bus voltages calculated by LP are very close to the voltages from power flow calculation and in most cases the differences are less than 0.05p.u. This means that the required tap and reactive power increments obtained from LP can be used to achieve optimal voltage profile. The method has also been tested on 39-bus system and the losses were reduced by 44% after 17 iterations. It is important to point out that for this system a larger step size was used and the results of each iteration tends to reach the optimal point by oscillation.

ACKNOWLEDGEMENT
The Authors wish to acknowledge the useful information provided by Mr. H.L. Forgey of Consumer Power Co. Special thanks also go to Ms. Deborah Waddy for her dedicated typing of the entire manuscript. APPENDIX A

The system power losses including real and reactive power losses is equal to the sum of total injected powT** er. . NT-* SL = V *I = V Y *V _ (A-1) *YV. .V
-

Since where Y is system admittance matrix. voltages can change in equation (A-1) * N ** ** . * ASL (Yij AV .V + Y VV jAV)
-T T -* *V =AV -Y *-* + V *-Y AV* . Note AVi, AV. have the same angle as Since ASL is a real number,

only

the

(A-2)

CONCLUSIONS
A new method is presented to find optimal reactivepower dispatch to minimize transmission losses and to improve voltage profile by adjusting VAR sources and transformer tap positions. The inverse of Jacobian matrix is not required by this method so, the method is It can be time efficient and needs less memory space. used as a tool to assist the power system operators to improve the system voltage profile and reduce losses. The method has been tested on the 6 and 39-bus systems and can easily be implemented to a larger system. Since sensitivity matrix is not used, the relationship between system losses and control variables is not discussed.

Vi,

V. respectively.

..*T *T..T .-V) (AV y .V) = (AV .y *~T =v y *AV

=V Y AV. Therefore AS L = -*T-* .AV + -T-* -AV V *Y -* V *Y


=

V {[cos6 -j

sinS.] D Y[CS 1
1

sin6.] D 1
(A-3)
The

[cos6 i+j sin6i] -Y 1 1D

[cos.i-j sin6i] }-AV 1D


diagonal

REFERENCES

[1]

Feb. 1982, pp. 397-405. [2] Raymond R. Shoults and M.S. Chen, "Reactive Power Control by Least Squares Minimization", IEEE Trans. P.A.S., Vol. 95, No. 1, Jan./Feb. 1976, pp. 315334. [3] H.H. Happ, "Optimal Power Dispatch - A Comprehensive Servay", IEEE Trans. P.A.S., Vol. 96, No. 3, May/June 1977, pp. 841-850. [4] Eric Hobson, "Network Constrained Reactive Power Control Using Linear Programming", IEEE Trans. P.A.S., Vol. 99, No. 3, 1980, pp. 868. [5] K.R.C. Mamandur and R.D. Chenoweth, "Optimal Control of Reactive Power Flow for Improvements in Voltage Profiles and for Real Power Losses Minimization", IEEE Trans. P.A.S., Vol. 100, No. 7, July 1981, pp. 3185-3193. [6] W.O. Stadlin and P.L. Fletcher, "Voltage vrersus Reactive Current Model for Dispatch and Control", IEEE Trans. P.A.S., Vol. 101, Oct. 1982, pp. 37513758. [7] S. Ramalyer, R. Ramachandran, and S. Haribaron, "New Technique for Optimal Reactive-Power Allocation for Loss Minimization in Power System", IEE Proceedings, Vol. 130, Pt.c, No. 4, July 1983, pp. 178-182. [8] S. Elanogovan, "New Approach for Real Power Loss Minimization", IEE Proceedings, Vol. 130, Pt.c, No. 6, Nov. 1983, pp 295-299.

Raymond R. Shoults, D.T. Sun, "Optimal Power Flow Based Upon P-Q Decomposition", IEEE Trans. P.A.S.,

where D means the matrix is a first term

matrix.

[cos6i-j sinmi] -Y .[cos6i+j sinSi]l 1 1 1D 1D = ([cos. i] .Y*j[sin 1D *Y ).*[cos6i+j sinms] i] 1 D ~ = {[cos] Y [cos] + [sinS] -Y *[sin6i] } D D D D
+

j{[cos6.]

[sinmI]

-Y [sin6.] D D } Y .[cos6]
1

(A-4)

The second term

[cos6i +jsin6i] D .Y [cos6i - j sin i]D 1D = {[cos i]D 'Y+j[sin6i] 1Y 1-[cosSi - j sinms]
=

o 1 1 {[cos6iD.] .Y.[cos6]

~D

Y sin + [ sinA. ] *y-**[s di] }

[in6iD

1ID

1D

+
-

j{[sini]
[coS]
Y

.Y
-

.[cos6i]l D 1
[sin6]
1

(A-5)

Substitute (A-4), (A-5) into (A-3), ASL 2'V *{ [cos8- ] .Y *[cos6A]


L

also

[sin6i]

.Y

[sinr6] }DAV
D =

(A-6)

[cos i] .DY [cos] D D

[cos6.]

DD

*G.[cosSi]

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

295
- j [cos6.]

and

1D

*B.[cos6.]
1D
=

1D

(A-7)

[sinAi]
-

*Y.[sin6.]

[sinei] D *G.[sin6.] 1D
(A-8)
Sub-

j [sin6.]

D*B'[sin6.]D

where G, B are the real and imaginary parts of Y. stituting (A-7), (A-8) into (A-6), we have
AS L = 2V *vr{[cos6.] *G.[cos6] 1 D D + [sin6i] *G.[sin6.] } AV D D - j2'V *{[cos6.] *. B*[cos6.] D D
+

[sinSi] D *B.[sinM] D } 1
D

AV.

(A-9)

The real power loss is .,T = 2'V

APL

{I[cos6.] *G.[cos6.] 1

+
=

[sin6i] D *G [sin.i] D }1*V 1


2*V1 *{G ij *[cos6i .cos6.
+

sin6. sin6] }.AV

2.[

DAP

where

a1

3APL ]-AV L
2

2*L*AV

(A-10)

9AP
Dv

L =

V. G. .cos.. i=l 1 1J 1J

(A-11)
(-l

So, the objective function is given by following equation where the factor 2 has been dropped from the equation (A-10).

APL

L*AV.

(A- 12)

Authorized licensed use limited to: PONTIFICIA UNIVERSIDADE CATOLICA DO RIO DE JANEIRO. Downloaded on May 04,2010 at 16:32:40 UTC from IEEE Xplore. Restrictions apply.

Вам также может понравиться