The Power Flow Equations

1.0 The Admittance Matrix

Current injections at a bus are analogous to power injections. The
student may have already been introduced to them in the form of
current sources at a node. Current injections may be either positive
(into the bus) or negative (out of the bus). Unlike current flowing
through a branch (and thus is a branch quantity) a current injection
is a nodal quantity. The admittance matri! a fundamental network
analysis tool that we shall use heavily relates current injections at
a bus to the bus voltages. Thus the admittance matri! relates nodal
quantities. "e motivate these ideas by introducing a simple
e!ample. We assume that all electrical variables in this document
are given in the per-unit system.
#ig. $ shows a network represented in a hybrid fashion using one%
line diagram representation for the nodes (buses $%&) and circuit
representation for the branches connecting the nodes and the
branches to ground. The branches connecting the nodes represent
lines. The branches to ground represent any shunt elements at the
buses including the charging capacitance at either end of the line.
'll branches are denoted with their admittance values y
ij
for a
branch connecting bus i to bus j and y
i
for a shunt element at bus i.
The current injections at each bus i are denoted by (
i
.

y
&
y
)
y
$
y
*
(
&
(
)
(
*
(
$
& )
*
$
y
)&
y
*)
y
$)
y
$*
Fig. 1: Networ !or Moti"ating Admittance Matrix
$
+irchoff,s Current -aw (+C-) requires that each of the current
injections be equal to the sum of the currents flowing out of the
bus and into the lines connecting the bus to other buses or to the
ground. Therefore recalling .hm,s -aw (/012/0y the current
injected into bus $ may be written as3
(
$
/(0
$
%0
*
)y
$*
4 (0
$
%0
)
)y
$)
4 0
$
y
$
($)
To be complete we may also consider that bus $ is 5connected6 to
bus & through an infinite impedance which implies that the
corresponding admittance y
$&
is 2ero. The advantage to doing this
is that it allows us to consider that bus $ could be connected to any
bus in the network. Then we have3
(
$
/(0
$
%0
*
)y
$*
4 (0
$
%0
)
)y
$)
4 (0
$
%0
&
)y
$&
4 0
$
y
$
(*)
7ote that the current contribution of the term containing y
$&
is 2ero
since y
$&
is 2ero. 8earranging eq. * we have3
(
$
/ 0
$
(

y
$
4 y
$*
4 y
$)
4 y
$&
) 4 0
*
(%y
$*
)4 0
)
(%y
$)
) 4 0
&
(%y
$&
) ())
9imilarly we may develop the current injections at buses * ) and
& as3
(
*
/ 0
$
(%y
*$
) 4 0
*
(

y
*
4 y
*$
4 y
*)
4 y
*&
) 4 0
)
(%y
*)
) 4 0
&
(%y
*&
) (&)
(
)
/ 0
$
(%y
)$
)4 0
*
(%y
)*
) 4 0
)
(

y
)
4 y
)$
4 y
)*
4 y
)&
) 4 0
&
(%y
)&
)
(
&
/ 0
$
(%y
&$
)4 0
*
(%y
&*
) 4 0
)
(%y
)&
)4 0
&
(

y
&
4 y
&$
4 y
&*
4 y
&)
)
where we recogni2e that the admittance of the circuit from bus k to
bus i is the same as the admittance from bus i to bus k i.e. y
ki
/y
ik
#rom eqs. ()) and (&) we see that the current injections are linear
functions of the nodal voltages. Therefore we may write these
equations in a more compact form using matrices according to3
*
1
1
1
1
]
1

1
1
1
1
]
1

+ + +
+ + +
+ + +
+ + +

1
1
1
1
]
1

&
)
*
$
&) &* &$ & &) &* &$
)& )& )* )$ ) )* )$
*& *) *& *) *$ * *$
$& $) $* $& $) $* $
&
)
*
$
V
V
V
V
y y y y y y y
y y y y y y y
y y y y y y y
y y y y y y y
I
I
I
I
(:)
The matri! containing the network admittances in eq. (:) is the
admittance matri! also known as the ;%bus and denoted as3
1
1
1
1
]
1

+ + +
+ + +
+ + +
+ + +

&) &* &$ & &) &* &$

)& )& )* )$ ) )* )$
*& *) *& *) *$ * *$
$& $) $* $& $) $* $
y y y y y y y
y y y y y y y
y y y y y y y
y y y y y y y
Y
(<)
=enoting the element in row i column j as ;
ij
we rewrite eq. (<)
as3
1
1
1
1
]
1

&& &) &* &$

)& )) )* )$
*& *) ** *$
$& $) $* $$
Y Y Y Y
Y Y Y Y
Y Y Y Y
Y Y Y Y
Y
(>)
where the terms ;
ij
are not admittances but rather elements of the
admittance matri!. Therefore eq. (:) becomes3
1
1
1
1
]
1

1
1
1
1
]
1

1
1
1
1
]
1

&
)
*
$
&& &) &* &$
)& )) )* )$
*& *) ** *$
$& $) $* $$
&
)
*
$
V
V
V
V
Y Y Y Y
Y Y Y Y
Y Y Y Y
Y Y Y Y
I
I
I
I
(?)
@y using eq. (>) and (?) and defining the vectors 0 and ( we may
write eq. (?) in compact form according to3
)
1
1
1
1
]
1

1
1
1
1
]
1

&
)
*
$
&
)
*
$

I
I
I
I
I
V
V
V
V
V V Y I
(A)
"e make several observations about the admittance matri! given
in eqs. (<) and (>). These observations hold true for any linear
network of any si2e.
$. The matri! is symmetric i.e. ;
ij
/;
ji
.
*. ' diagonal element ;
ii
is obtained as the sum of admittances for
all branches connected to bus i including the shunt branch i.e.

+
N
i k k
ik i ii
y y Y
$
where we emphasi2e once again that y
ik
is non%
2ero only when there e!ists a physical connection between
buses i and k.
). The off%diagonal elements are the negative of the admittances
connecting buses i and j i.e. ;
ij
/%y
ji
.
These observations enable us to formulate the admittance matri!
very quickly from the network based on visual inspection. The
following e!ample will clarify.
Exam#le 1
Consider the network given in #ig. * where the numbers indicate
admittances.
&

jB.&

jB.)

jB.$

jB.*

(
&
(
)
(
*
(
$
& )
*
$
*%j)

*%j:

$%j&

*%j&

Fig. $: Circuit !or Exam#le 1
The admittance matri! is given by inspection as3
1
1
1
1
]
1

+
+ + +
+ +
+ +

1
1
1
1
]
1

< . * * ) * B B
) * > . $$ : : * & $
B : * ? . ? & & *
B & $ & * A . > )
&& &) &* &$
)& )) )* )$
*& *) ** *$
$& $) $* $$
j j
j j j j
j j j
j j j
Y Y Y Y
Y Y Y Y
Y Y Y Y
Y Y Y Y
Y
$.0 The #ower !low equations
%efine the net comple! power injection into a bus as 9
k
/9
gk
%9
dk
. (n
this section we desire to derive an e!pression for this quantity in
terms of network voltages and admittances. "e begin by
reminding the reader that all quantities are assumed to be in per
unit so we may utili2e single%phase power relations. =rawing on
the familiar relation for comple! power we may e!press 9
k
as3
9
k
/0
k
(
k
C
($B)
#rom eq. (?) we see that the current injection into any bus k may
be e!pressed as
j
N
j
kj k
V Y I

$
($$)
:

where again we emphasi2e that the ;
kj
terms are admittance
matri! elements and not admittances. 9ubstitution of eq. ($$) into
eq. ($B) yields3
C
$
C
C
$
j
N
j
kj k j
N
j
kj k k
V Y V V Y V S

,
_

($*)
8ecall that 0
k
is a phasor having magnitude and angle so that
0
k
/D0
k
D
k
. 'lso ;
kj
being a function of admittances is therefore
generally comple! and we define E
kj
and @
kj
as the real and
imaginary parts of the admittance matri! element ;
kj
respectively
so that ;
kj
/E
kj
4j@
kj
. Then we may rewrite eq. ($*) as
( )
( )
( )
( )

+
N
j
kj kj j k j k
N
j
kj kj j j k k
j j
N
j
kj kj k k
j j
N
j
kj kj k k j
N
j
kj k k
jB G V V
jB G V V
V jB G V
V jB G V V Y V S
$
$
$
C
$
C C
$
C
) ( ) (
) (
) (
) (




($))
8ecall from the Fuler relation that a phasor may be e!pressed as
comple! function of sinusoids i.e. 0/D0D/D0DGcos4jsinH.
"ith this we may rewrite eq. ($)) as
( )
( )

+

N
j
kj kj j k j k j k
N
j
kj kj j k j k k
jB G j V V
jB G V V S
$
$
) ( ) sin( ) cos(
) ( ) (


($&)
(f we now perform the algebraic multiplication of the two terms
inside the parentheses of eq. ($&) and then collect real and
imaginary parts and recall that 9
k
/I
k
4jJ
k
we can e!press eq. ($&)
as two equations one for the real part I
k
and one for the
imaginary part J
k
according to3
<
( )
( )

+
N
j
j k kj j k kj j k k
N
j
j k kj j k kj j k k
B G V V
B G V V !
$
$
) cos( ) sin(
) sin( ) cos(


($:)
The two equations of ($:) are called the power flow equations and
they form the fundamental building block from which we attack
the power flow problem.
&.0 'ol"ing the #ower !low #ro(lem
The standard power flow problem is as follows3
Eiven that for each bus (node) in the network we know * out of
the following & variables3 I
k
J
k
D0
k
D K
k
so that for each bus there
are two equations available L those of eq. ($:) above and there are
two unknown variables. Thus the power flow problem is to find
solve the power flow equations of ($:) for the remaining two
variables per bus.
This problem is one where we are required to solve simultaneous
nonlinear equations. @ecause most power systems are very large
interconnections with many buses the number of power flow
equations (and thus the number of unknowns) is very large. #or
e!ample a model of the eastern interconnection in the U9 can have
:BBBB buses.
The approach to solving the power flow problem is to use an
iterative algorithm. The 7ewton%8aphson algorithm is the most
commonly used algorithm in commercial power flow programs.
9tarting with a reasonable guess at the solution (where the
5solution6 is a numerical value of all of the unknown variables)
this algorithm checks to see how close the solution is and then if it
is not close enough updates the solution in a direction that is sure
>
to improve it and then repeats the check. This process continues
until the check is satisfied. Usually this process requires :%*B
iterations to converge to a satisfactory solution. #or large
networks it is computationally intensive.
(n this class we are very interested in optimi2ation methods for
finding least cost (or ma!imum surplus) solutions to the problem
of how to dispatch the generation. 9o far we have dealt with
problems where all generation and load was considered to be at the
same bus (node) and were thus able to ignore the network. @ut in
reality generation and load are located at various buses and the
transportation mechanism for moving electrical energy from
supply to consumption is the transmission network. (f there are
losses or constraints in the transmission network (which there are)
these will influence how supply can be allocated and the most
economically desirable solutions may not be feasible.
To account for the network in the economic optimi2ation problems
we have posed we must account for the equations that correspond
to the network. These are the power flow equations. This can be
done and problem that results is referred to as the optimal power
flow (.I#). Mowever because the power flow equations are
nonlinear the .I# requires a nonlinear optimi2ation method for its
solution. 7onlinear optimi2ation (usually called nonlinear
programming instead of linear programming) is a rich interesting
and highly applicable area. ;ou can take entire courses on this
subject (see for e!ample (F <)$).
@ut we do not have time in this class to learn nonlinear
programming methods. 'nd fortunately since we have learned
linear programming we do not need to do so i" #e can convert our
nonlinear problem into a linear one$
"e learned in the last class how to convert our nonlinear objective
function into a linear objective function using piecewise linear
?
appro!imations. This method was very effective in appro!imating
the objective function since the objective function is a separable
function (can be separated into components) and each component
is a function of only one variable. (n other words we were able to
apply piecewise linear appro!imation to each individual cost curve
because each cost curve was a function of only one variable.
The power flow equations are functions of many variables and it is
very comple! to see how to apply piecewise linear appro!imations
since a piecewise linear appro!imation for any one variable will
depend on the value of the other variables.
Thus we seek another method of converting our nonlinear power
flow equations into linear equations.
).0 A##roximations to the #ower !low equations
-et,s reconsider the power flow equations3

( )
( )

+
N
j
j k kj j k kj j k k
N
j
j k kj j k kj j k k
B G V V
B G V V !
$
$
) cos( ) sin(
) sin( ) cos(


($:)
"e will make use of three practical observations regarding high
voltage electric transmission systems.
.bservation $3 The resistance of transmission circuits is
significantly less than the reactance. Usually it is the case that the
!1r ratio is between * and $B. 9o any given transmission circuit
with impedance of 2/r%j! will have an admittance of
jb g
% r
j%
% r
r
% r
j% r
j% r
j% r
j% r j% r &
y
+
+

+

* * * *
* *
$ $ $
($<)
#rom eq. ($<) we see that
A
* *
% r
r
g
+

and
* *
% r
%
b
+

($>)
(f r is very small compared to ! then we observe that g will be
very small compared to b and it is reasonable to appro!imate eqs.
($>) as
B g
and
%
b
$

($?)
7ow if g/B then the real part of all of the ;%bus elements will
also be 2ero that is g/BE/B.
'pplying this conclusion to the power flow equations of eq. ($:)3
( )
( )

N
j
j k kj j k k
N
j
j k kj j k k
B V V
B V V !
$
$
) cos(
) sin(


($A)
.bservation *3 #or most typical operating conditions the
difference in angles of the voltage phasors at two buses k and j
connected by a circuit which is K
k
%K
j
for buses k and j is less than
$B%$: degrees. (t is e!tremely rare to ever see such angular
separation e!ceed )B degrees. Thus we say that the angular
separation across any transmission circuit is 5small.6
Consider that in eqs. ($A) the angular separation across a
transmission circuit K
k
%K
j
appears as the argument of the
trigonometric functions sine and cosine. "hat do these functions
look like for small anglesN "e can answer this question by
recalling that the sine and cosine functions represent the vertical
and hori2ontal components of a unit (length/$) vector making an
angle O/K
k
%K
j
with the positive !%a!is as illustrated in #ig. ).
$B

cosO
O
sinO
#ig. )3 Trig functions of a small angle
(n #ig. ) it is clear that as the angle O/K
k
%K
j
gets smaller and
smaller the cosine function approaches $.B.
.ne might be tempted to accept the appro!imation that the sine
function goes to 2ero. This it does as the angle goes to 2ero. @ut
an even better appro!imation is that the sine o" a small angle is the
angle itsel" (when the angle is given in radians). This can be
observed in #ig. ) from the fact that the vertical line representing
the sine is almost the same length as the indicated radial distance
along the circle which is the angle (when measured in radians).
'pplying these conclusions from observation * to eqs. ($A)3
( )
( )

N
j
kj j k k
N
j
j k kj j k k
B V V
B V V !
$
$
) (
(*B)
7ote that we have made significant progress at this point in
relation to obtaining linear power flow equations since we have
eliminated the trigonometric terms. Mowever we still have product
$$
terms in the voltage variables and so we are not done yet. .ur ne!t
and last observation will take care of these product terms.
@efore we do that however let,s investigate the e!pressions of eq.
(*B) a little.
8ecall that the quantity @
kj
is not actually a susceptance but rather
an element in the ;%bus matri!.
(f kPj then @
kj
/%b
kj
i.e. the ;%bus element in row k column j is
the negative of the susceptance of the circuit connecting bus k to
bus j.
(f k/j then

+
N
k j j
kj k kk
b b B
$
'eactive po#er "lo#3
The reactive power flow equation of eqs. (*B) may be rewritten by
pulling out the k/j term from the summation.
( ) ( )

N
k j
j
kj j k kk k
N
j
kj j k k
B V V B V B V V
$
*
$
Then substitute susceptances for the ;%bus elements3
( )
( )
( )

,
_

+
N
k j
j
kj j k
N
k j j
kj k k k
N
k j
j
kj j k
N
k j j
kj k k k
N
k j
j
kj j k
N
k j j
kj k k k
b V V b V b V
b V V b V b V
b V V b b V
$ $
* *
$ $
* *
$ $
*
7ow bring all the terms in the two summations under a single
summation.
( )

,
_

N
k j j
kj j k kj k k k k
b V V b V b V
$
* *
#actor out the D0
k
D and the Lb
kj
in the summation3
( )

N
k j j
j k kj k k k k
V V b V b V
$
*
(*$)
7ote because we defined the circuit admittance between buses k
and j as y
kj
/g
kj
4jb
kj
and because all circuits have series elements
that are inductive the numerical value of b
kj
is negative. Thus we
can rewrite eq. (*$) as
$*
( )

+
N
k j j
j k kj k k k k
V V b V b V
$
*
(**)
9o there are two main terms in eq. (**).
The first term corresponds to the reactive power supplied (if a
capacitor) or consumed (if an inductor) by the shunt susceptance
modeled at bus k.
The second term corresponds to the reactive power flowing on
the circuits connected to bus k. .nly these circuits will have
non2ero b
kj
. .ne sees then that each circuit #ill have per-unit
reactive "lo# in proportion to (a) the bus k voltage magnitude
and (b) the di""erence in per-unit voltages at the circuit*s
terminating buses. The direction of flow will be from the higher
voltage bus to the lower voltage bus.
'eal po#er "lo#3 7ow consider the real power flow equation from
eqs. (*B) and as with the reactive power flow equation let,s pull
out the j/k term. Thus
( ) ( ) ( )

+
N
k j
j
j k kj j k k k kk k
N
j
j k kj j k k
B V V B V B V V !
$
*
$
) ( ) ( ) (
Mere we see that the first term is 2ero so that3
( )

N
k j
j
j k kj j k k
B V V !
$
) (
(*))
9ome comments about this e!pression3
There is no 5first term6 corresponding to shunt elements as
there was for the reactive power equation. The reason for this is
that because we assumed that 8/B for the entire network there
can be no shunt resistive element in our model. This actually
conforms to reality since we never connect a resistive shunt in
the transmission system (this would be equivalent to a giant
heaterQ). The only place where we do actually see an effect
which should be modeled as a resistor to ground is in
transformers the core loss is so modeled. @ut the value of this
resistance tends to be e!tremely large implying the
corresponding conductance (E) is e!tremely small and it is
very reasonable to assume it is 2ero.
$)
Therefore the term that we see in eq. (*)) represents the real
power flow on the circuits connected to bus k. .ne sees then
that each circuit #ill have per-unit real po#er "lo# in
proportion to (a) the bus k and j voltage magnitudes and (b) the
angular di""erence across the circuit. #urthermore recalling that
@
kj
/%b
kj
and also that all transmission circuits have series
elements that are inductive the numerical value of b
kj
is
negative implying that the numerical value of @
kj
is positive.
Therefore the direction of flow will be from the bus with the
larger angle to the bus with the smaller angle.
.bservation )3 (n the per%unit system the numerical values of
voltage magnitudes D0
k
D and D0
j
D are very close to $.B. Typical
range under most operating conditions is B.A: to $.B:. -et,s
consider the implications of this fact in terms of the e!pressions for
reactive and real power flow eqs. (**) and (*)) repeated here for
convenience3
( )

+
N
k j j
j k kj k k k k
V V b V b V
$
*
( )

N
k j
j
j k kj j k k
B V V !
$
) (
Eiven that B.A:RD0
k
D and D0
j
DR$.B: then we incur little error in the
above e!pressions if we assume D0
k
D/D0
j
D/$.B everywhere that they
occur as a multiplying factor. "e cannot make this appro!imation
however where they occur as a difference in the reactive power
equation because the difference of two numbers close to $.B can
range significantly. #or e!ample $.B:%B.A:/B.$ but $.B$%
$.B/B.B$ an order of magnitude difference.
Saking this appro!imation results in3
$&
( )

+
N
k j j
j k kj k k
V V b b
$
(*&)
( )

N
k j
j
j k kj k
B !
$
) (
(*:)
"ith these equations we can narrow our statements about power
flow.
8eactive power flow across circuits is determined by the
difference in the voltage phasor magnitudes between the
terminating buses.
8eal power flow across circuits is determined by the difference
in voltage phasor angles between the terminating buses.
#inally it is interesting to note that the disparity between the
ma!imum reactive power flow and the ma!imum real power flow
across a circuit.
The reactive power flow equation is proportional to the circuit
susceptance and the difference in voltage phasor magnitudes.
The ma!imum difference in voltage phasor magnitudes will be
on the order of $.B:%B.A:/B.$.
The real power flow equation is proportional to the circuit
susceptance and the difference in voltage phasor angles. The
ma!imum difference in voltage phasor angles will be in
radians about B.:* (which corresponds to )B degrees).
"e see from these last two bullets that real power flow across
circuits tends to be significantly larger than reactive power flow
i.e. usually we will see that
kj kj
! >>
This conclusion is consistent with operational e!perience which is
characteri2ed by an old operator,s saying3 50ars don,t travel.6
*.0 +eal "s. +eacti"e Power Flow
8ecall that our original intent was to represent the network in our
optimi2ation problem because of our concern that network
$:
constraints may limit the ability to most economically dispatch the
generation. There are actually several different causes of network
constraints but here we will limit our interest to the type of
constraint that is most common in most networks and that is
circuit overload.
Circuit overload is caused by high current magnitude. "hen the
current magnitude e!ceeds a given threshold for a particular circuit
(called the rating) we say that overload has occurred.
(n the per%unit system we recall that

+
kj k kj kj kj
I V j ! S
where 0
k
is the bus k nodal voltage phasor and (
kj
is the phasor of
the current flowing from bus k to bus j. Thus we have that3

,
_

k
kj kj
kj
V
j !
I
Taking the magnitude (since that is what determines circuit
overload) we have3
k
kj kj
kj
V
!
I
* *
+

Eiven our conclusion on the previous page that generally I

kj
TTJ
kj

we may appro!imate the above e!pression according to3
k
kj
k
kj
kj
V
!
V
!
I
*
and if D0
k
DU$.B then we have that
kj kj
! I
Thus in assessing circuit overload it is reasonable to look at real
power flows only. 's a result of this conclusion we will build into
our optimi2ation formulation only the real power flow equations
i.e. eq. (*:).
,.0 The %C Power Flow - an exam#le
-et,s study the real power flow e!pression given in eq. (*:).
$<

N
k j
j
j k kj k
B !
$
) (
(t is worthwhile to perform a simple e!ample to illustrate use of
this e!pression.
Consider the &%bus network given in #ig. &. 'll : lines have the
same admittance and this admittance has no real part indicating
we are assuming 8/B for this network. The real power values for
each of the three generators and each of the two loads are given.
'll numerical quantities are given in per%unit.
The #ro(lem is to com#ute the real #ower !lows on all circuits.

y
$)
/%j$B
y
$&
/%j$B
y
)&
/%j$B
y
*)
/%j$B
y
$*
/%j$B
I
g*
/*pu
I
d)
/&pu
I
d*
/$pu
$ *
) &
I
g*
/*pu
I
g&
/$pu
#ig. &3 #our%bus network used in e!ample
"e first write down eq. (*:) for each bus beginning with bus $.
& $& $ $& ) $) $ $) * $* $ $*
& $ $& ) $ $) * $ $* $
) ( ) ( ) (


B B B B B B
B B B !
+ +
+ +
Collecting terms in the same variables results in3
( )
& $& ) $) * $* $ $& $) $* $
B B B B B B ! + +
(*<)
8epeating the process for bus *3
$>
& *& * *& ) *) * *) $ *$ * *$
& * *& ) * *) $ * *$ *
) ( ) ( ) (


B B B B B B
B B B !
+ +
+ +
'gain collecting terms in the same variables results in3
( )
& *& ) *) * *& *) *$ $ *$ *
B B B B B B ! + + +
(*>)
8epeating eqs. (*<) and (*>) together with the relations for buses
) and & yields3
( )
& $& ) $) * $* $ $& $) $* $
B B B B B B ! + +
( )
& *& ) *) * *& *) *$ $ *$ *
B B B B B B ! + + +
( )
& )& ) )& )* )$ * )* $ )$ )
B B B B B B ! + + +
( )
& &) &* &$ ) &) * &* $ &$ &
B B B B B B ! + + +
"e can write these equations in matri! form according to3
1
1
1
1
]
1

1
1
1
1
]
1

+ +
+ +
+ +
+ +

1
1
1
1
]
1

&
)
*
$
&) &* &$ &) &* &$
)& )& )* )$ )* )$
*& *) *& *) *$ *$
$& $) $* $& $) $*
&
)
*
$

B B B B B B
B B B B B B
B B B B B B
B B B B B B
!
!
!
!
(*?)
8emember the left%hand%side vector is the injections which is the
generation less the demand.
To get the matri! it is helpful to first write down the ;%bus3
1
1
1
1
]
1

&& &) &* &$

)& )) )* )$
*& *) ** *$
$& $) $* $$
B B B B
B B B B
B B B B
B B B B
j Y
1
1
1
1
]
1

+ + +
+ + +
+ + +
+ + +

&) &* &$ & &) &* &$

)& )& )* )$ ) )* )$
*& *) *& *) *$ * *$
$& $) $* $& $) $* $
b b b b b b b
b b b b b b b
b b b b b b b
b b b b b b b
j
1
1
1
1
]
1

*B $B B $B
$B )B $B $B
B $B *B $B
$B $B $B )B
j Y
9o we readily observe here that for e!ample @
$$
/%)B @
$*
/$B
@
$)
/$B and @
$&
/$B and it is similar for the other three rows.
$?
9o using the ;%bus values we can e!press eq. (*?) as3
1
1
1
1
]
1

1
1
1
1
]
1

1
1
1
1
]
1

&
)
*
$
*B $B B $B
$B )B $B $B
B $B *B $B
$B $B $B )B
$
&
$
*

(*A)
(.bserve that we omit the j). Then the angles are given by3
1
1
1
1
]
1

1
1
1
1
]
1

1
1
1
1
]
1

$
&
$
*
*B $B B $B
$B )B $B $B
B $B *B $B
$B $B $B )B
$
&
)
*
$

()B)
Mowever when we evaluate the above e!pression by taking the
inverse of the indicated matri! we find it is singular i.e. it does
not have an inverse. The problem here is that we have a
dependency in the & equations implying that one of the equations
may be obtained from the other three. #or e!ample if we add the
bottom three rows and then multiply by %$ we get the top row (in
terms of the injection vector this is just saying that the sum of the
generation must equal the demand).
This dependency occurs because all four angles are not
independent. "hat is important is the angular differences. Thus
we are free to choose any one of them as the reference with a
fi!ed value of B degrees. This angle is then no longer a variable (it
is B degrees) and referring to eq. (*A) the corresponding column
in the matri! may be eliminated since those are the numbers that
get multiplied by this B degree angle.
To fi! this problem we need to eliminate one of the equations and
one of the variables (by setting the variable to 2ero).
"e choose to eliminate the first equation and set the first variable
K
$
to 2ero (which means we are choosing K
$
as the reference).
$A
This results in3
1
1
1
]
1

1
1
1
]
1

1
1
1
]
1

1
1
1
]
1

B*: . B
$: . B
B*: . B
$
&
$
*B $B B
$B )B $B
B $B *B
$
&
)
*

()$)
The solution on the right%hand%side gives the angles on the bus
voltage phasors at buses * ) and &.
Mowever the problem statement requires us to compute the power
flows on the lines (this is usually the information needed by
operational and planning engineers as they study the power
system).
"e can get the power flows easily by employing just one term
from the summation in eq. (*:) which gives the flow across circuit
k%j according to3
) (
j k kj kj
B !
()*)
"e utili2e the ;%bus elements together with the bus angles given
by eq. ()$) to make these calculations as follows3
*: . B ) B*: . B B ( $B ) (
* $ $* $*
B !
: . $ ) $: . B B ( $B ) (
) $ $) $)
B !
*: . B ) B*: . B B ( $B ) (
& $ $& $&
B !
*: . $ ) $: . B B*: . B ( $B ) (
) * *) *)
B !
*: . $ ) B*: . B $: . B ( $B ) (
& ) )& )&
B !
These computed flows are illustrated in #ig. :. The power flowing
into a bus equals the power flowing out of that bus.
*B

I
$)
/$.:
I
$&
/B.*:
I
&)
/$.*:
I
*)
/$.*:
I
$*
/B.*:

I
g*
/*pu
I
d)
/&pu
I
d*
/$pu
$ *
) &
I
g*
/*pu
I
g&
/$pu
#ig. :3 Computed flows for four%bus network used in e!ample
..0 The %C Power Flow - /enerali0ation
"e desire to generali2e the above procedure.
"e assume that we are given the network with the following
information3
Total number of buses is 7 total number of branches is S.
@us number $ identified as the reference
8eal power injections at all buses e!cept bus $
7etwork topology
'dmittances for all branches.
The =C power flow equations based on eq. (*:) are given in
matri! form as
V B !
()))
where
I is the vector of nodal injections for buses * W 7
K is the vector of nodal phase angles for buses *W7
*$
@ , is the 5@%prime6 matri!. Eenerali2ation of its development
requires a few comments.
=evelopment of the @, matri!3
Compare the matri! of eq. (*?) with the ;%bus matri! all repeated
here for convenience3
1
1
1
1
]
1

1
1
1
1
]
1

+ +
+ +
+ +
+ +

1
1
1
1
]
1

&
)
*
$
&) &* &$ &) &* &$
)& )& )* )$ )* )$
*& *) *& *) *$ *$
$& $) $* $& $) $*
&
)
*
$

B B B B B B
B B B B B B
B B B B B B
B B B B B B
!
!
!
!
1
1
1
1
]
1

&& &) &* &$

)& )) )* )$
*& *) ** *$
$& $) $* $$
B B B B
B B B B
B B B B
B B B B
j Y
1
1
1
1
]
1

+ + +
+ + +
+ + +
+ + +

&) &* &$ & &) &* &$

)& )& )* )$ ) )* )$
*& *) *& *) *$ * *$
$& $) $* $& $) $* $
b b b b b b b
b b b b b b b
b b b b b b b
b b b b b b b
j
#rom
the above we can develop a procedure to obtain the @, matri!
from the ;%bus as follows3
$. 8emove the 5j6 from the ;%bus.
*. 8eplace diagonal element @,
kk
with the sum of the non%diagonal
elements in row k. 'lternatively subtract b
k
(the shunt term)
from @
kk
and multiply by %$.
). Sultiply all off%diagonals by %$.
&. 8emove row $ and column $.
Comparison of the numerical values of the ;%bus with the
numerical values of the @, matri! for our e!ample will confirm the
above procedure3
**
1
1
1
1
]
1

*B $B B $B
$B )B $B $B
B $B *B $B
$B $B $B )B
j Y
1
1
1
]
1

*B $B B
$B )B $B
B $B *B
V B
'nother way to remember the @, matri! is to observe that since its
non%diagonal elements are the negative of the ;%bus matri! the @,
non%diagonal elements are susceptances. Mowever one must be
careful to note that the @, matri! element in position row k
column j is the susceptance of the branch connecting buses k4$
and j4$ since the @, matri! does not have a column or row
corresponding to bus $.
Juestion3 "hy are shunt terms e!cluded in the @, matri!N That is
why does e!cluding them not affect real power calculationsN
'lthough eq. ())) provides the ability to compute the angles it
does not provide the line flows. ' systematic method of computing
the line flows is3
) ( + , !
B ()&)
where3
I
@
is the vector of branch flows. (t has dimension of S ! $.
@ranches are ordered arbitrarily but whatever order is chosen
must also be used in = and '.
K is (as before) the vector of nodal phase angles for buses *W7
= is an S ! S matri! having non%diagonal elements of 2erosX
the diagonal element in position row k column k contains the
negative of the susceptance of the k
th
branch.
' is the S ! 7%$ node-arc incidence matri%. (t is also called the
adjacency matri! or the connection matri!. (ts development
requires a few comments.
=evelopment of the node%arc incidence matri!3
*)
This matri! is well known in any discipline that has reason to
structure its problems using a network of nodes and 5arcs6 (or
branches or edges). 'ny type of transportation engineering is
typical of such a discipline.
The node%arc incidence matri! contains a number of rows equal to
the number of arcs and a number of columns equal to the number
of nodes.
Flement (kj) of ' is $ if the k
th
branch begins at node j %$ if the k
th
branch terminates at node j and B otherwise.
' branch is said to 5begin6 at node j if the power flowing across
branch k is defined positive for a direction "rom node j to the other
node.
' branch is said to 5terminate6 at node j if the power flowing
across branch k is defined positive for a direction to node j from
the other node.
7ote that matri! ' is of dimension S ! 7%$ i.e. it has only 7%$
columns. This is because we do not form a column with the
reference bus in order to conform to the vector K which is of
dimension (7%$) ! $. This works because the angle being e!cluded
K
$
is 2ero.
"e can illustrate development of the node%arc incidence matri! for
our e!ample system.
Consider numbering the branches as given in #ig. <. Iositive
direction of flow is as given by the indicated arrows.
*&

:
$
&
)
*
I
g*
/*pu
I
d)
/&pu
I
d*
/$pu
$ *
) &
I
g*
/*pu
I
g&
/$pu
#ig. <3 @ranches numbering for development of incidence matri!
Therefore the node%arc incidence matri! is given as
number branch
-
.
/
0
1

2 1 - 2
1 1 - 2
2 1 - 1
2 2 1 -
1 - 2 2
+
. / 0

number node

1
1
1
1
1
1
]
1

The =%matri! is formed by placing the admittances of each branch
along the diagonal of an S ! S matri! where S/:.
1
1
1
1
1
1
]
1

$B B B B B
B $B B B B
B B $B B B
B B B $B B
B B B B $B
,
Combining ' = and K based on eq. ()&) yields
) ( + , !
B
*:
1
1
1
]
1

1
1
1
1
1
1
]
1

1
1
1
1
1
1
]
1

&
)
*
$B B B B B
B $B B B B
B B $B B B
B B B $B B
B B B B $B

2 1 - 2
1 1 - 2
2 1 - 1
2 2 1 -
1 - 2 2
1
1
1
1
1
1
]
1

1
1
1
1
1
1
]
1

1
1
1
1
1
1
]
1

1
1
1
1
1
1
]
1

)
& )
) *
*
&
)
& )
) *
*
&
:
&
)
*
$
$B
) ( $B
) ( $B
$B
$B
$B B B B B
B $B B B B
B B $B B B
B B B $B B
B B B B $B

B
B
B
B
B
!
!
!
!
!
Ilugging in the solution for K that we obtained which was3
1
1
1
]
1

1
1
1
]
1

B*: . B
$: . B
B*: . B
&
)
*

"e find that the above evaluates to

1
1
1
1
1
1
]
1

1
1
1
1
1
1
]
1

: . $
*: . $
*: . $
*: . B
*: . B
:
&
)
*
$
B
B
B
B
B
!
!
!
!
!
This solution obtained systematically in a way that can be
efficiently programmed agrees with the solution we obtained
manually and is displayed in #ig. :.
*<