IEEE TENCON'93/ Beijng

AN EXTENDED APPROACH FOR NR LOAD FLOW WITH POWER LOSS CORRECTlON MIXHOD
Joong-Rin Shin, Han-Suck Yim
Dept. of Electrical Engineering, College of Engineering,
Kon-Kuk University, Seoul Korea 133-701

ABSTRACT
Newton-Raphson (NR) method is the f a " load flow (LF)
calculation technique, and
"ally
used d w to its
rapidness of numerical convergency. In the conventional NR
method, however, there is a somewhat unrealistic assumption
under which all the systeo power losses are considered to be
supplied by the slack bus generator. In this study, a new
approach for handling the losses and augmenting the
conventional NR method is proposed. The proposed method
estimates the incremental changes of active power on each
generation bus with respect to the total system power loss,
and the estimated value are used to update the slack bus
power. By doing this is improved the performance of power
flow calculation, in (gabal sense.

LF FORMULATION OF CONVENTIONAL NR METHOD

Well-known NR method for LF calculation can be sunmarized as
follows [2-41.
k t tus indices 1 to m denote pure load buses. (m+l) to
(n-1) denote pu'e generation buses. and n h t e slack bus.
then LF problem is to find, mtn-1 unknown variables, bus
voltage magnitudes ( V ) and n-1 bus voltage angles ( 6 ) .
from the following nonlinear simultaneous equations.

INTRODUCTION
The power flow calculations are the aost important and
powerful tools in power systems engineering. Newton-Raphson
(NR) method, amng the various power flow calculation
techniques, is normally used due to its rapidness of
numerical convergency. In the conventional Newton-Raphson
method, however, there is a somewhat unrealistic assumption
under which all the system power losses are considered to be
supplied by the slack bus generator[l-rl]. The unrealistic
feature of the assumption may make little difficulty in
general cases of power systems engineering, but in certain
cases, especially such as voltage stability studies. it can
make the results of load flow solution distorted from those
of the real system. In the situations close to voltage
collapse , power system has relatively large transmission
loss which makes load flow solution often diverges. In those
situations. even in convergence case, the results of load
flow solution would be more distorted than in the nomal. and
these distorted solutions can make the results of voltage
stability analysis inaccurate [5].
Introducing the system power loss formula and augmenting the
conventional Newton-Raphson power flow method, we can
relieve the unrealistic assumption and improve the
performance of power flow calculation. In this paper a new
approach is proposed, in which the conventional load flow
problem is augmented and the system power loss, as well as
bus voltage magnitudes and angles, are corrected at each
iteration step.

*re,

Pi. Pi : active.reactive pure iMection power at bus i.

Gi j. Bi j : conductance. susceptance between bus i and j
SIJ 6i-6j
L = {l. ...m} : index set of load bus
G = {m+l,..,n-l} : index set of generation bus
( slack bus excluded )
In general. P i . V i (iEG), Vn.& (n:slack bus). P j , P j (jEL)
are considered as preliminarily specified in the
cmwntional LF forulations. In some literature, P i . V i ( i ~
G), Vn.Gn are called control variables and the P J , ~(jL)
disturbance variables [l]. NR Method then finds the unknown
variabes V i . iEL and 6 i . iEL*G ( these are called state
variables also in the literature [l] ) basically with
iterative updating scheme f m a given initial point ( V O , 6
0 ) ' .based on the 1st order Tayler approximation of eq. (1)

Unknown variable u-plating scheme

'Ihe iterative scheme "only used can be explained in brief

with the following equations;

............................................................
(*) Hereafter. subscript denotes the component of vector.
superscript denotes the number of iteration count and the
capital letter without subscript means a vector .

- 402 -

I M

1-1
J

I L

r1 M

active slack power CM be exprerwd in a different way

cwpera with eq. (Sa) but it ha6 oam meaning.

active power missmatch ,

reactive
voltage angle difference.
voltage magnitude difference,

Pn

= P Pi
i EL

P~(v.6)- I: PJ

(7)

jEG

sub matrix.
Eq. (7) states also that the system loss are supplied only

Then , newly updated magnitd and angle of bus voltage are

obtained:

by slack bus.

EXTENDED APPROACH FOR NR LOAD FLOW

If V,6 are converged after a few iterations. then the final

values are accepted as the solution of the LF caculation.
Calculation of line power flows and slack bus power
Given the solution the power flow of each transmission line
and slack bus power are obtained with following expressions:
Pij=ViVj(Gijcos6iJ*Bijsin6ij)-Vi*Gij

(4a)

Pij=ViVj (Gijsin6i j -BijC d i j)+Vi2 ( B i J -bij 12)

(4b)

where
bij : shunt susceptwe of the line between bus i and .i
Pij,Qij : Active,reactive power flow from bus i to j
And

n
pn = ZVnVj (GnjCOS& j+Bnjsinbj )

(5a)

j=l

Qn =

ZVnVj(Gnjsin&

j-En jCOSdnj)

(5b)

j=l

where
Pn.Pn : Active,reactive power of slack bus generator
and Vn is normaly set to unity.

Augmenting the conventional NR load flow formulation by

adding the system loss equation. eq. ( 6 ) . to the original
simultanious power equation set. eq.'s (1). and taking the
magnitude of slack bus voltage Vn as a new state variable.
an extended approach for the conventional NR method is
p r o p o d in this pper. In this p r o p o d method, the system
loss PL are treated as a new control variable which is
corrected in every main iteration step of load flow
capitation. At the begiming of main iteration of the
pmposed method the system loss are properly specified ( let
it be denoted by P L S ) and then the state variables are
updated in a similar way with the conventional NR Method,
After me iteration. PI, can be calculated newly with the
updeted state variables. Denoting this calculated value by
PLC.
it probably has
difference from the initially
specified value 9 s . It should be noted that it is self
widient because PL is a dependent variable and practically
not control one. If we can provide a certain correction
scheme of PLS .for each iteration step, which can make PLC
converge to P L ~ .the p m p o s a d method can resolve, in global
sense, the problems concerned with unrealistic performance
of conventional load flow study .

Iterative system loss correction

As seen in eq. (5). slack bus powers(active.reactive) are

dependently determined by the final solution V.6 of other
buses in right hand side. It means that the slack bus powers
cannot be specified beforehand and all the system power
losses are supplied by the slack bus generator. These results
are somewhat unrealistic because in practical sense the system
loss are shared with every generator by its eletrical duty. The
unrealistic feature of conventional Newton-Raphson LF
solution may make little difficulty in general cases of power
systems engineering, but in certain cases it can make the
results of load flow solution distorted from those of the real
world.

m u l a t i o n of system b~
As known generally, the total transmission system loss ( or
sytem loss in brief) PL is calculated with the final V.6 BS
follows:

In this paper the systen loss correction is d.

at each
iteration step, by distributing the system loss difference
A P L = PL' - PLC in proper way to every generator huses. It
means that in next iteration step P L ~is re-specified and
thus corrected. It should be noted that re-specifying the
generation power directly means the change of slack bus
power and thus of system loss by eq. (6). With this
re-specified PLO another Newton-Ram iteration step is
continued to make PL' converge to pratical system loss. bre
detail presentation w.r. t the loss correction scheme is
given in following paragraph.

The augmented power equation set is given in

e q . ' s (8).

it should be noted again that specifyiw system loss

means directly that slack bus power is specified, and it is
a characterized feature compared with convetional method.
&re.

Pis = I: ViVj(Gijco~6ij+Bijsin6ij) i(G,L,)

(8a)

l
'
j

pi' = P ViVj(Gi jcos6ij-Bijsin6ij

J=1

PL = PL (V.6)

- 403 -

iEL

(8b)

n n

which is easy to compte and has good profitability in

mathmatical sense. The profitability is k e d on the fact
that if the magnitude of generation bus voltage is changed
then the calculated APL is changed.

PLO = I: Z ViGij(-Vi+VJcos6ij)
i=Ij=i
j#i

where, superscript s denotes the specified value,

Let the shared part of AF'L to the ith gemration bus be

denoted by U L i . it can be represented using wellknown
chain rule of derivatives as follows:

Then updating equation has the following form,

M j
n
_____
------I

L j
-----

ah

I-------

I-Vn

: aVn
where. the term (aPi/aVi)Vi, (ah/aVi)Vi can be computed
easily in similar ways with the computation of Jacobian
elements:

(1%)
Then at next iteration, i.e., (k+l)th iteration, the
specified active power of generation bus i is readjusted as
follows:

included)
8Pi
Vn = ViVn(GinCoS6in+Bijsin6in)

aVn

PiS(k+l)

8Qi
Vn = - ViVn(Bincos6in-Ginsin6in)
8Vn

t h , updated state variables after kth iteration step are

obtained:
@+I
w.1

= 6k

Vnk*l = Vnk

A6

(AV/V*)
P
+

PiS(k)

ULi

Ct(n)

(14)

I n this procedure. there is another problem to determine the

value of specified system loss.PLS(k*I),before starting at
(k+l)th iteration. Because the system loss difference,A
PL(k+l),obtained at kth step is already distributed to all
generation buses. theoretically P~s(k+*)must be equal to
zero. But. in practice, eq. (12) is an approximate one and
necessarily P~s(k+l) is required to be determined. At this
time eq. (15) is proposed to determine the specified value
of system loss at iteration k+l.

(AV/Vnk) Vnk

As sharn in eq. (9). the active power and voltage angle of

slsdt bus is specified and its voltage magnitude is updated
86 a state variable ( i.e.. unknown variable) . This is the
awific feature of the proposed method, with which the
m l i s t i c concentration of system loss in slack bus, in
conventiaml NR method, can be relaxed.
Using eq. (6) with these updated state variables , P~c(k*i)
is calculated and accordingly the system loss difference A
P~k'l = PLS(~) - PLc(k+i) is obtained. APL is distributed
to every generation buses and next iteration continued. The
convergency of extended approach may be affected greatiy by
the distribution rule of loss difference to generation
buses. Accordingly. the design of loss distribution rule
probably be important.

. .

m i b u t i o n rule of svstem loss difference

In this study, an analytical distribution rule is designed
using the information of derivatives of active power at each
jpnwation bus with repect to the mgnitude of its voltage,

Then the new system loss difference A P L ( ~ * ~can

) be updated
as:
a ( k + I ) = pLs(k+l) - pLc(k*i)
. (16)
Fig. 1. illustrate the logical flow of the extended NR load
flow method which is proposed in this paper.

CASE STUDY AND DISCUSSION

Case Study

Some case studies are carried out to determine the

performance of the proposed method for 5-bus. 30-bus. 57-bus
IEEE test systems[4]. The results of case studies show that
the solutions obtained from the proposed method seem very
close, even if system loss was corrected at each iteration,
to those from conventional method. It can probably be noted
that this shows the availability of the proposed method.
Table 1. shows the comparison between the LF results of

- 404 -

Table 3. hvergency of' the proposed method

compared wtih the conventional.
iter.no.
h

M.AQ

5-bus
conv.

I prop.

3O-bm
conv.

prop.

57-bus
conv.

A
t

P
AQ
conv. = conventional.

prop. = prcvosed

affect greatly the convergency or final solution of proposed

method. Table 3. says that the convergency of proposed
method is a little bad, which could be acceptable.

Discussion
The specific feature of this study can be summarid as

follows:

Fig. 1. Flow chart of the proposed method.

conventional method and proposed one for 5-bus system. Table
2. shows the effect bf initially specified system loss on
the LF solutions for the 5-bus system. From the results it
can be stated that the initial value of system loss does not

Table 1. LF results for 5-bus test system.

busnubei:i

1.0303 1.0440 1.0414 1.0564 1.06oO

convent' a1

-0.0934 -0.0644 -0.0726 -0.0396 O.oo00

-

V
Proposed

1.0255 1.0414 1.0383 1.0554 1.O600

-0,1033 -0.0690 -0.0781 -0.0402 0. oo00

Table 2. Effect of initial system loss change

for 5-bus test system (proposed method).

1,0258 1.0414 1.0384 1.0554 1.06oO

-0,1025-0.0690-0.0780-0.0402 0. Oooo
0.02
1.0255 1.0422 1.0390 1.0554 1.06oO
0.05

-0<034 -0.0669-0.0766-0.0401 0.MX)o

-.

1) In this study, a new approach for the conventionnl NR

load flow method which can make correction of system loss at
every iteration step to relieve the unrealistic assumption
and improve the global performance of power flow
calculation.
2) The proposed method estimates the incremental changes of
active power on each generation bus with respect to the
total systen power loss, and the estimated values are used
to update the slack bus power to relax the concenturation of
power loss in slack bus generator.
3) The results of proposed method seem very close, even if
system loss was corrected at each iteration, to those from
conventional method. It can pr&ly
be noted that this
shows the availability of the prciposed method,
If more theoretical investigations and verifications are
followc+, with respect to the interactions auiung the amOunt
of system power loss, the degree of distortion in load flow
solution and the indices of voltage stability, then the
proposed approach will give more contribution to increase
the performance of power flow techniques in power systems
engineering. especially for the voltage stability analysis.
REFERENCES

[I] R.L.Sullivan. "Power System Planning". McCraw-Hill. 1977

[21 0. L. Elgerd,"Electric Energy Systems Theory-An Introduction", 2nd ed.,McGraw-Hill, 1982
[31 W. F.Tinny. W. L.Powel1, "Notes on Newton-Raphson Method
for Solution of AC Power Problem". BPA Oregon, 1971
[41 M. A. Pai. "Complter Techniques in Power System hlysis",
Tata McGraw-Hi11, 1979
[51 IEEE. "Voltage Stabi 1 i ty of Power System:Concepts.halytical Tools.and Industry Experience",9(yM0358-2-PW.1990

- 405 -