Load Flow Studies 193

The elements of this matrix, of course, agree with those previously calculated in Example 6.1.
If there is mutual coupling between elements i and j of the network, the primitive Y matrix will not be
diagonal. Thus, the element corresponding to the ith row and jth column (also, jth row and ith column) will
be equal to the mutual admittance between the elements i and j. Thus, YBUS is also modified as given by Eq.
(6.26).

6.4 LOAD FLOW PROBLEM

The complex power injected by the source into the ith bus of a power system is
Si = Pi + jQi = Vi Ii*, i = 1, 2, ..., n
Since it is convenient to work with Ii instead of I *i, we take the complex conjugate of the above equation,
Pi – jQi = V *i Ii, i = 1, 2, ..., n
Ê n ˆ
Substituting Ii = Á Â (YikVk )˜ from Eq. (6.2b) in the above equation, we have
Ë k =1 ¯
Ê n ˆ
Pi – jQi = V *i Á Â (YikVk )˜ , i = 1, 2, ..., n
Ë k =1 ¯
Equating real and imaginary parts, we get
Ê Ê n ˆˆ
Pi (Real power) = Real Á Vi* Á Â (Yik Vk )˜ ˜ (6.27a)
Ë Ë k =1 ¯¯
Ê Ê n ˆˆ
Qi (Reactive power) = – Imaginary Á Vi* Á Â (Yik Vk )˜ ˜ (6.27b)
Ë Ë k =1 ¯¯
jd i jd k
Let, Vi = | Vi | e , Vk = | Vk | e
jqik
Yik = | Yik | e
then,

n
Pi (Real power) = | Vi | Â | Vk | | Yik | cos (qik + dk – di) (6.28a)
k =1
n
Qi (Reactive power) = – | Vi | Â | Vk | |Yik | sin (qik + dk – di) (6.28b)
k =1
(i = 1, 2, ..., n)
Equations 6.28(a) and (b) are called power flow equations. There are n real and n reactive power flow
equations giving a total of 2n power flow equations.
At each bus there are four variables, viz., | Vi |, d i, Pi and Qi, giving a total of 4n variables (for n buses).
If at every bus two variables are specified (thus specifying a total of 2n variables), the remaining two
variables at every bus (a total of 2n remaining variables) can be found by solving the 2n power flow Eqs
6.28(a) and (b). When a physical system is considered, specifying variables at every bus depends on what
devices are connected to that bus. In general there are four possibilities giving rise to four types of buses as
follows.
194 Modern Power System Analysis

6.4.1 Slack Bus/Swing Bus/Reference Bus

This bus is distinguished from the remaining types by the fact that real and reactive powers at this bus are
not specified. Instead, voltage magnitude (normally set equal to 1 pu) and voltage phase angle (normally set
equal to zero) are specified. Usually, there is only one bus of this type in a given power system. The need of
such a bus for a load flow study is explained in the example that follows. The slack bus is numbered 1, for
convenience.

6.4.2 PQ Bus/Load Bus

At this type of bus, the net powers Pi and Qi are known (PDi and QDi are known from load forecasting and
PGi and QGi are specified). The unknowns are |Vi | and di. A pure load bus (no generating facility at the bus,
i.e., PGi = QG i = 0) is a PQ bus. PQ buses are the most common, comprising almost 80% of all the buses in
a given power system.

6.4.3 PV Bus/Generator Bus

This bus has always a generator connected to it. Thus PGi and | Vi | are specified. Hence the net power Pi is
known (as PDi is known from load forecasting). Hence, the knowns are Pi and | Vi | and unknowns are Qi
and di. PV buses comprise about 10% of all the buses in a power system.

6.4.4 Voltage Controlled Bus

Frequently the PV bus and the voltage controlled bus are grouped together. But they have physical
differences and slightly different calculation strategies. The voltage controlled bus has also voltage control
capabilities, and uses a tap-adjustable transformer and/or a static var compensator instead of a generator.
Hence PGi = QGi = 0 at these buses. Thus, Pi = – PDi, Qi = – QDi and |Vi | are known at these buses and the
unknown is d i.

Example 6.5 This example will also demonstrate the need of a slack bus. Consider a simple four bus
system [Fig. 6.5(a)] out of which two buses are PQ buses, one is PV bus, and the remaining one is a slack bus.
The buses are numbered as shown in Table 6.3 [Refer Fig. 6.5(c)].
Table 6.3

No. Bus Type Known Unknown

1. Slack Bus | V1 |, d 1 P1, Q1
2. PQ Bus P2, Q2 | V2 |, d 2
3. PQ Bus P3, Q3 | V3 |, d3
4. PV Bus P4, | V4 | Q 4, d 4

Total Real Power Loss, PL = SPi = P1 + P2 + P3 + P4

= P1 + K1
where, K1 = P2 + P3 + P4 (known).
Total Reactive Power Loss, QL = SQi = Q1 + Q2 + Q3 + Q4
= Q1 + Q4 + K2
where, K2 = Q2 + Q3 (known).
 Load Flow Studies 195

Since PL and QL are not known prior to load flow solution, the real and reactive powers (Pi and Qi) cannot
be fixed at all the buses. In the above example, P1, Q1 and Q4 cannot be fixed prior to load flow solution. After
the load flow solution is complete, the total real and reactive powers, PL = Â Pi and QL = Â Qi become
i i
known and the slack bus has to supply excess real power P1 = PL – K1, and excess reactive power Q1 = QL
– Q4 – K2. This is why a slack bus is needed in load flow solution.
Thus, after a load flow solution is complete, real power at the slack bus (P1) is known, and hence the
real power generation PG1 (as PD1 is known from load forecasting) is known. As V1 is already specified
for a slack bus, the slack bus must be a generator bus. It is required that the variations in real and reactive
powers at the slack bus be a small percentage of its generating capacity. So, the bus connected to the largest
generating station is normally selected as the slack bus.
Equations 6.28(a) and (b) which are written below for convenience are referred to as Static Load Flow
Equations (SLFE).
n
Pi (Real power) = |Vi| Â (|Vk| |Yik| cos (qik + dk – d i)) (6.28a)
k =1
n
Qi (Reactive power) = – |Vi| Â (|Vk| |Yik| sin (qik + dk – di))
k =1
where, i = 1, 2, ..., n (6.28b)
By transposing all the variables on one side, these equations can be written in the vector form
f (x, y) = 0 (6.29)
where, f = vector function of dimension 2n ¥ 1
x = vector of dependent or state variables of dimension 2n ¥ 1 (2n independent
variables specified a priori)
y = vector of independent variables of dimension 2n ¥ 1 (2n unspecified variables)
Some of the independent variables in x can be used to manipulate some of the state variables. These
adjustable independent variables are called control variables. The remaining independent variables which
are fixed are called fixed parameters. Vector x can then be partitioned into a vector u of control variables
and a vector p of fixed parameters,
Èu˘
x= Í ˙ (6.30)
Î p˚
Control variables may be voltage magnitude on PV bus, PGi at buses with controllable power, etc. Fixed
parameters are those which are uncontrollable.
For SLFE solution to have practical significance, all the state and control variables must be within
specified practical limits. These limits are dictated by specifications of power system hardware and
operating constraints, and are described below:
1. Voltage magnitude | Vi | must satisfy the inequality
| Vi | min £ | Vi | £ | Vi | max (6.31a)
This limit arises due to the fact that the power system equipment is designed to operate at fixed
voltages with allowable variations of ±(5–10)% of rated values.
2. Certain of the di s (state variables) must satisfy
| di – dk | £ | di – dk | max (6.31b)
196 Modern Power System Analysis

This constraint limits the maximum permissible power angle of transmission line connecting buses i
and k and is imposed by considerations of stability.
3. Owing to physical limitations of P and/or Q generation sources, PGi and QGi are constrained as
follows:
PGi, min £ PGi £ PGi, max
QGi, min £ QGi £ QGi, max (6.31c)
Also, we have
 PGi =  ( PDi ) + PL (6.32a)
i i

 QGi =  (QDi ) + QL (6.32b)

i i
where PL and QL are system real and reactive power losses.
The load flow problem can now be fully defined as follows:
Assume a certain nominal bus load configuration. Specify PGi + jQGi at all PQ buses (this specifies
Pi + jQi at these buses, specify PGi (this specifies Pi) and | Vi | at all PV buses; and specify | V1 | and
d 1 (= 0) at the slack bus. Thus all the 2n variables of vector x are specified. Thus 2n SLFE, which are
nonlinear algebraic equations,* can be solved (iteratively) to determine the values of the 2n variables
of the vector y, comprising voltages and angles at the PQ buses; reactive powers and angles at the PV
buses; and active and reactive powers at the slack bus.
The next logical step is to compute the line flows.
From the above definition we can state two versions of the load flow problem. In both the cases, bus 1 is
assumed as the slack bus.
Case I We assume all the remaining buses are PQ buses. We have:
Given V1, S2, S3, ..., Sn
Find S1, V2, V3, ..., Vn
Case II We assume both PV and PQ buses. We number the buses so that buses 2, 3, ... m are PQ buses and
m + 1, ..., n are PV buses.
Thus,
Given V1, S2, ..., Sm, (Pm + 1, | Vm + 1 |), .., (Pn, | Vn |)
Find S1, V2, ..., Vm, (Qm + 1, d m + 1), ..., (Q n, dn)
Since the load flow equations are essentially nonlinear, they have to be solved through iterative numerical
techniques. At the cost of solution accuracy, it is possible to linearise load flow equations by making suitable
assumptions and approximations so that fast and explicit solutions become possible. Such techniques have
value, particularly for planning studies where load flow solutions have to be carried out repeatedly but a
high degree of accuracy is not needed. An approximate load flow solution is dealt in Example 6.6.
Once a load flow problem is formulated it can be seen that there exists a range of every independent
variable in the vector x for which there is no solution, or there are multiple solutions. For visualising this
problem, we consider a simple two bus system in Example 6.6.

Example 6.6 In Fig. 6.11, bus 1 is the reference bus with V1 = 1 –0º and bus 2 is PQ bus. We are to
find S1 and V2.

* It is because cosine and sine functions are involved in real and reactive power Eqs 6.28(a) and (b).
 Load Flow Studies 197

Fig. 6.11 A simple two-bus system

Solution The complex power injected at the bus 2, S2 is given by

S2 = SG2 – SD2 = – PD2 + j0.2 (i)
Also from Sec. 5.9 of Ch. 5, we have
S2 = S21
| V2 |2 | V2 | | V1 |
= | Z | –q - Z –(q + d2 – d1) (ii)
where, Z = | Z | –q = 0.5 –90º (iii)
d1 = 0º as V1 = | V1 | –d 1 = 1 –0º
From (i), (ii) and (iii), we get
– PD2 + j0.2 = j2 | V2 |2 – 2 | V2 | cos (q + d 2) – j2 | V2 | sin (q + d 2)
Equating the real and the imaginary parts on both sides, we get
PD2 = 2 | V2 | cos (q + d 2)
0.2 = 2 | V2 |2 –2 | V2 | sin (q + d 2)
or, we get
PD2
cos (q + d2) = (iv)
2 | V2 |
2 |V2 |2 - 0.2 |V2 |2 - 0.1
sin (q + d2) = = (v)
2 |V2 | |V2 |
Squaring and adding, we get
2
PD2 (|V2 |2 - 0.1) 2
1= +
4 |V2 |2 |V2 |2
or,
4 | V2 |4 – 4.8 | V2 |2 + (P D2
2 + 0.04) = 0

or,
4.8 ± 4.82 - 16( PD22 + 0.04)
| V 2 |2 =
8
Thus, if 16 (P2D2 + 0.04) > (4.8)2 or PD2 > 1.183, there are no real solutions. If PD2 = 1.183, | V2 | = 0.775,
the solution is unique. If PD2 < 1.183, there are two real solutions, and hence the solution is non-unique. Let
PD2 = 0.5; then | V2 | = 0.253 and 1.066 pu.
Thus, there may be no solution, or at least no unique solution, depending on the given data. In most
cases having non-unique solutions, the practical solution is the one with voltage closest to 1 pu. Thus, with
PD2 = 0.5, the desired solution is | V2 | = 1.066, and the angle d2 is given by Eq. (iv) as
198 Modern Power System Analysis

PD2
cos (q + d2) = cos (90 + d2) = – sin d2 =
2 | V2 |
or,
PD2 0.5
sin d2 = – = = – 0.2345
2 | V2 | 2 ¥ 1.066
or,
d2 = – 13.56º
Also,
S1 = – S2 = PD2 – j0.2 = 0.5 – j0.2
The solution of a load flow problem becomes simple when only two buses are present. But, in a normal
power system there are large number of buses and the solution becomes quite complex, tedious and time
consuming. Hence, there is no choice but to use iterative methods. The most commonly used iterative
methods are the Gauss–Siedel, the Newton–Raphson and fast-decoupled load flow methods.

6.4.5 An Approximate Load Flow Solution

Let us make the following assumptions and approximations in the load flow Eqs 6.28(a) and (b):
1. Line resistances, being small, are neglected (shunt conductance of overhead lines is always
negligible), i.e., PL, the active power loss of the system is zero. Thus, in Eqs 6.28(a) and (b), qik =
90º and qii = – 90º.
2. (d i – d k ) is small [< (p/6)], so that sin (di – dk) (di – dk). This is justified from considerations of
stability (see Ch. 12).
3. All buses other than the slack bus (numbered as 1) are PV buses, i.e., voltage magnitudes at all the
buses, including the slack bus, are specified.
Equations 6.28(a) and (b) then reduce to
n
Pi = | Vi | Â | Vk | | Yik | (di – d k) i = 2, 3, ..., n (6.33)
k =1
n
Qi = – | Vi | Â (| Vk | | Yik | cos (di – d k)) + | Vi |2 | Yii |, i = 1, 2, ..., n (6.34)
k =1
k πi
Since | Vi |s are specified, Eq. (6.33) represents a set of linear algebraic equations in dis, which are
(n – 1) in number as d1 is specified at the slack bus (d1 = 0). The nth equation corresponding to the slack
bus (n = 1) is redundant as the real power injected at this bus is now fully specified as P1 = S PDi –S PGi;
(PL = 0). Equation (6.33) can be solved explicitly (non-iteratively) for d2, d3, ..., dn, which, when substituted
in Eq. (6.34), yields Qis, the reactive power bus injections. It may be noted that the assumptions made
have decoupled Eqs (6.33) and (6.34), so that these need not be solved simultaneously but can be solved
sequentially. Solution of Eq. (6.34) follows immediately upon simultaneous solution of Eq. (6.33). Since
the solution is non-iterative and the dimension is reduced to (n – 1) from 2n, it is computationally highly
economical.

Example 6.7 Consider the four-bus sample system of Fig. 6.12, wherein line reactances are indicated
in pu. Line resistances are considered negligible. The magnitudes of all four bus voltages are specified to be
1.0 pu. The bus powers are specified in Table 6.4.
 Load Flow Studies 199

Solution Figure 6.12 indicates bus

injections for the data specified in Table
6.4. As bus voltages are specified, all the
buses must have controllable Q sources. It
is also obvious from the data that buses 3
and 4 only have Q sources. Further, since
the system is assumed lossless, the real
power generation at bus 1 is known a priori
to be PG1 = PD1 + PD2 + PD3 + PD4 – PG2
= 2.0 pu. Therefore, we have 7 unknowns
instead of 2 ¥ 4 = 8.
In the present problem, the unknown
state and control variables are d 2, d3, d4, Fig. 6.12 Four bus lossless sample system
QG1, QG2, QG3 and QG4.
Table 6.4

Bus Real Demand Reactive Demand Real Generation Reactive Generation

1. PD1 = 1.0 QD1 = 0.5 PG1 = ? QG1 (unspecified)
2. PD2 = 1.0 QD2 = 0.4 PG2 = 4.0 QG2 (unspecified)
3. PD3 = 2.0 QD3 = 1.0 PG3 = 0.0 QG3 (unspecified)
4. PD4 = 2.0 QD4 = 1.0 PG4 = 0.0 QG4 (unspecified)

Though the real losses are zero, the presence of the reactive losses requires that the total reactive
generation must be more than the total reactive demand (2.9 pu)
From the data given, YBUS can be written as follows:
1 2 3 4
1 È - j 21.667 j 5.000 j 6.667 j10.000 ˘
Í j 5.000 - j 21.667 j10.000 j 6.667 ˙˙
YBUS = 2 Í (i)
3Í j 6.667 j10.000 - j16.667 j 0.000 ˙
Í ˙
4 ÍÎ j10.000 j 6.667 j 0.000 - j16.667 ˙˚
Using the above YBUS and bus powers as shown in Fig. 6.12, approximate load flow Eq. (6.33) is expressed
as (all voltage magnitudes are equal to 1.0 pu)
P2 = 3 = 5(d2 – d1) + 10(d2 – d3) + 6.667(d2 – d4) (ii)
P3 = – 2 = 6.667 (d3 – d1) + 10(d3 – d2) (iii)
P4 = – 2 = 10(d4 – d1) + 6.667(d4 – d2) (iv)
Taking bus 1 as a reference bus, i.e., d1 = 0, and solving (ii), (iii) and (iv), we get
d2 = – 0.077 rad = 4.41º
d3 = – 0.074 rad = – 4.23º (v)
d4 = – 0.089 rad = – 5.11º
Substituting d s in Eq. (6.34), we have
Q1 = –5 cos 4.41º – 6.667 cos 4.23º – 10 cos 5.11º + 21.667
Q2 = – 5 cos 4.41º – 10 cos 8.64º – 6.667 cos 9.52º + 21.667
Q3 = – 6.667 cos 4.23º – 10 cos 8.64º + 16.667
200 Modern Power System Analysis

Q4 = – 10 cos 5.11º – 6.667 cos 9.52º + 16.667

or,
Q1 = 0.07 pu
Q2 = 0.22 pu (vi)
Q3 = 0.132 pu
Q4 = 0.132 pu
Reactive power generations at the four buses are
QG1 = Q1 + 0.5 = 0.57 pu
QG2 = Q2 + 0.4 = 0.62 pu (vii)
QG3 = Q3 + 1.0 = 1.132 pu
QG4 = Q4 + 1.0 = 1.132 pu
Reactive line losses are
4 4
QL = Â QG i -Â QDi
i =1 i =1

= 3.454 – 2.9 = 0.554 pu (viii)

Now, let us find the line flows. Equation (5.68) can be written in the form (| Z | = X, q = 90º)
| V | | Vk |
Pik = – Pki = i sin (d1 – dk)
X ik
where Pik is the real power flow from bus i to bus k.
1
P13 = – P31 = sin (d1 – d3) = sin 4.23º/0.15 = 0.492 pu
0.15
1
P12 = – P21 = sin (d1 – d2) = – sin 4.41º/0.02 = – 0.385 pu (ix)
0.2
1
P14 = – P41 = sin (d 1 – d4) = sin 5.11º/0.1 = 0.891 pu
0.1
Real power flows on other lines can be similarly calculated. For reactive power flow, Eq. (5.69) can be
written in the general form (Z = X, q = 90º)
| Vi |2 | Vi | | Vk |
Qik = - cos (d i – d k )
X ik X ik
where Qik is the reactive power flow from bus i to bus k.
1 1
Q12 = Q21 = – cos (d1 – d2) = 0.015 pu
0.2 0.2
1 1
Q13 = Q31 = - cos (d1 – d3) = 0.018 pu (x)
0.15 0.15
1 1
Q14 = Q41 = - cos (d1 – d4) = 0.04 pu
0.15 0.15
Reactive power flows on other lines can similarly be calculated. Generations and load demands at all the
buses and all the line flows are indicated in Fig. 6.13.
 Load Flow Studies 201

Fig. 6.13 Load flow solution for the four bus system

6.5 GAUSS–SIEDEL METHOD

The Gauss–Siedel (GS) method is an iterative algorithm for solving a set of nonlinear algebraic equations.
We consider a system of n equations in n unknowns x1, ..., xn. We rewrite these n equations in the form:
xi = f i (x1, ..., xn), i = 1, ..., n
To find the solution, we assume initial values for x1, ..., xn, based on guidance from practical experience
in a physical situation. Let the initial values be x 01, .., x0n. Then, we get the first approximate solution by
substituting these initial values in the above n equations as follows:
(
x11 = f1 x10 , x20 … , xn0 )

(
x1i = fi x11 , x12 , … , x1i -1 , xi0 , … , xn0 )

(
x1n = f n x11 , x12 ,… , x1n -1 , xn0 )
This completes one iteration. In general, we get the kth approximate solution in the kth iteration as follows:
k -1
xik = fi (x1k, ..., xki – 1, x ik – 1, xi + 1 , ..., x nk – 1), i = 1, 2, ..., n
We note that the value of xi for the kth iteration is obtained by substituting x1k, x2k, ..., xki – 1 (obtained in
k th iteration) and xik – 1, ..., xnk – 1 (obtained in (k – 1)th iteration) in the equation xi = fi. If | xik – xik – 1| < e (a
very small number), for all i values, then the solution is said to converge. The iterative process is repeated
till the solution converges within prescribed accuracy. The convergence is quite sensitive to the starting
values assumed. Fortunately, in a load flow study, a starting vector close to the final solution can be easily
identified with previous experience.
To solve a load flow problem by GS method, we consider two cases depending on the type of buses
present. In the first case we assume all the buses other than the slack bus are PQ buses. In the second case,
we assume the presence of both PQ and PV buses, other than the slack bus. Later on we also include the
presence of voltage controlled bus, whose voltage is controlled by a regulating transformer.
Case I The slack bus is numbered one, and the remaining (n – 1) buses are PQ buses (i = 2, ..., n). With
slack bus voltage assumed, the remaining (n – 1) bus voltages are found through iterative process as
follows:
202 Modern Power System Analysis

The complex bus power injected into the ith bus is

Si = Pi + jQi = Vi I *i
or,
Pi - jQi
Ii = (6.35)
Vi*
From Eq. (6.2b)
Ê ˆ
n
1 Á ˜
Vi = I i - Â (Yik Vk )˜ , i = 2, ..., n (6.36)
Yii Á
ÁË k =1
k πi
˜¯
Substituting Ii from Eq. (6.35) into Eq. (6.36)
È ˘
1 Í Pi - jQi n
˙
Vi = Í *
- Â (Yik Vk ) ˙ , i = 2, ..., n (6.37)
Yii Í Vi k =1 ˙
Î k πi ˚
We assume a starting value for Vi (i = 2, ..., n). These values are then updated using Eq. (6.37). The voltages
substituted at the right hand side of Eq. (6.37) are the most recently (updated) values for the corresponding
buses. Iterations are repeated till no bus voltage magnitude changes by more than a prescribed value.
If instead of updating voltages at every step of an iteration, updating is carried out at the end of a
complete iteration, the process is known as Gauss Iterative Method. It converges much slower and may
sometimes fail to do so.

6.5.1 Algorithm for Load Flow Solution (Case I)

1. The slack bus voltage magnitude and angle are assumed, usually V1 = 1 –0º pu. With the load profile
known at each bus (i.e., PDi and QDi known), we allocate PGi and QGi to all generating stations. With
this step, bus injections (Pi + jQi) are known at all buses other than the slack bus.
2. Assembly of bus admittance matrix (YBUS): With the line and shunt admittance data stored in the
computer, YBUS is assembled by using the algorithm developed earlier. Alternatively, YBUS is
assembled using YBUS = AT YA, where the input data is in the form of primitive admittance matrix Y
and singular connection bus incidence matrix A.
3. Iterative computation of bus voltages (Vi ; i = 2, ..., n): To start the iteration, a set of initial voltage
values is assumed. Since, in a power system the voltage spread is not too wide, it is normal practice
to use a flat voltage start, i.e., initially all voltages are set equal to (1 + j0), except the slack bus
voltage, which is fixed. This reduces the n equations to (n – 1) equations in complex numbers
(Eq. (6.37)) which are to be solved iteratively for finding complex voltages V2, V3, ..., Vn. If complex
number operations are not available in a computer, Eq. (6.37) can be converted into 2(n – 1)
equations in real unknowns (ei, fi or | Vi |, di) by writing
Vi = ei + jfi = | Vi | e jdi (6.38)
We also define
Pi - jQi
Ai = , i = 2, ..., n (6.39)
Yii
Bik = Yik /Yii i = 2, ..., n
k = 2, ..., n, k π i (6.40)
 Load Flow Studies 203

Now, for the (r + 1)th iteration, the voltage Eq. (6.37) becomes
È A i -1 n ˘
Vi(r + 1) = Í r * - Â ( Bik Vk ) - Â ( BikVk( r ) ) ˙ , i = 2, ..., n
i ( r +1)
(6.41)
ÍÎ (Vi ) k =1 k = i +1 ˙˚
The iterative process is continued till the change in magnitude of bus voltage, | DVi(r + 1) |, between two
consecutive iterations is less than a certain tolerance for all bus voltages, i.e.,
| DVi(r + 1) | = |Vi(r + 1) – Vi(r) | e, i = 2, ..., n (6.42)
Also, we see if | Vi | min £ | Vi | £ | Vi |max, i = 2, ..., n.
If not, we fix | Vi | at one of the extreme values, i.e.,
| Vi |min if | Vi | £| Vi |min or | Vi | max if | Vi | ≥ | Vi | max.
Depending on the nature of the problem, we can also check
if | di – d k | £ | di – dk |max (i = 1, ..., n; k = 1, ..., n; i π k)
4. Computation of slack bus power: Substitution of all bus voltages computed in step 3 with V1 in Eqs
6.28(a) and (b) with i = 1 yields real and reactive power at the slack bus i.e., S1 = P1 + jQ1.
5. Computation of line flows: This is
the last step in the load flow analysis
wherein the power flows on the various
lines of the network are computed.
This also enables us to check whether
any line is overloaded. Consider the
line connecting buses i and k. The line
and transformers at each end can be
represented by a circuit with series
admittance yik and two shunt admittances
yik0 and yki0 as shown in Fig. 6.14. Fig. 6.14 p-representation of a line and transformers
The current fed by bus i into the line can be connected between two buses
expressed as
Iik = Iik1 + Iik0 = (Vi – Vk) yik + Viyik0 (6.43)
The power fed into the line from bus i is
Sik = Pik + jQik= Vi I*ik = Vi(Vi* – Vk*) yik* + ViVi*yik0* (6.44)
Similarly, the power fed into the line from bus k is
Ski = Vk (V*k – Vi*) yik* + VkViko*y*ki0 (6.45)
The power loss in the (i – k)th line is the sum of the power flows determined from Eqs (6.44) and (6.45).
Total transmission loss can be computed by summing all the line flows (i.e., Sik + Ski for all i, k).
If there are any mutually coupled transmission lines, then we use the Eqs 6.10(a) and (b) to find the
currents between lines and then compute for the line flows as above.
It may be noted that the slack bus power can also be found by summing up the flows on the lines
terminating at the slack bus.

6.5.2 Acceleration of Convergence

Convergence in the GS method can sometimes be speeded up by the use of the acceleration factor. For the
ith bus, the accelerated value of voltage at the (r + 1)st iteration is given by
Vi(r + 1) (accelerated) = Vi(r) + a(Vi(r+1) – Vi(r))
204 Modern Power System Analysis

where a is a real number called the acceleration factor. A suitable value of a for any system can be obtained
by trial load flow studies. A generally recommended value is a = 1.6. A wrong choice of a may indeed slow
down convergence or even cause the method to diverge.
This concludes the load flow analysis for PQ buses only. Next, we consider the presence of PV buses
(see Example 6.9).
Case II Here we consider (m – 1) PQ buses, (n – m) PV buses and the slack bus.
Given V1, (P2, Q2), ..., (Pm, Qm), (Pm + 1, | Vm + 1 |), ..., (Pn, | Vn |)
To find S1, V2, ..., Vm, (Qm + 1, dm + 1), ..., (Qn, d n).
Algorithm (Case II) We first repeat the iteration for PQ buses as in Case I, then continue the iteration for
PV buses. At the PV buses, P and | V | are specified and Q and d are unknowns to be determined. Therefore,
the values of Q and d are to be updated in every GS iteration through appropriate bus equations. This is
accomplished in the following steps:
1. From Eq. (6.27b),
ÏÔ n ¸Ô
Qi = – Im ÌVi* Â YikVk ˝ , i = m + 1, ..., n
ÔÓ k =1 Ô˛
The revised value of Qi is obtained from the above equation by substituting most updated values of voltages
on the right hand side. For the (r + 1)th iteration we can write,
ÏÔ ( r ) i - 1 ¸
( )
n
( )
Qi(r + 1) = –Im Ì Vi * Â YikVk
( r +1) r r Ô
+ Vi ( ) * Â YikVk ( ) ˝ , i = m + 1, ..., n (6.46)
ÓÔ k =1 k =i ˛Ô
2. The revised value of di is obtained from Eq. (6.41) immediately after step 1.
È (r +1) i -1 n ˘
Í Ai ( r) ˙
di(r + 1) = –Vi(r + 1) = Angle - Â BikVk ( r +1)
- Â BikVk (6.47)

where,
ÍÎ( )
Í (r )
Vi * k =1 k = i +1
˙
˚˙
Pi - jQi ( )
r +1
Ai(r + 1) = , i = m + 1, ..., n (6.48)
Yii
Physical limitations of Q generation require that Q demand at any bus must be in the range Qmin to Qmax.
If at any stage during iteration, Q at any bus goes outside these limits, it is fixed at Qmin or Qmax as the case
may be, and the bus voltage specification is dropped, i.e., the bus is now treated like a PQ bus. Thus, Step 1
above branches out to Step 3 as follows:
3. If Qi(r + 1) £ Qi, min, we set Qi(r + 1) = Qi, min or if Qi(r +1) ≥ Qi, max, we set Qi(r + 1) = Qi, max , and treat the
bus i as a PQ bus. We compute Ai(r + 1) and Vi(r + 1) from Eqs (6.48) and (6.41) respectively.
Now, all the computational steps are summarised in the detailed flow chart of Fig. 6.15. It
is assumed that out of n buses, the first is slack bus, then 2, 3, ..., m are PQ buses, and the remaining,
m + 1, ..., n are PV buses.
Case III We consider here the presence of voltage controlled buses in addition to PQ and PV buses other
than the slack bus. This case will be dealt in Example 6.9.

Example 6.8 For the sample system of Fig. 6.7 the generators are connected at all the four buses,
while loads are at buses 2 and 3. Values of real and reactive powers are listed in Table 6.5. All buses other
than the slack are PQ type.
Assuming a flat voltage start, find the voltages and bus angles at the three buses at the end of the first GS
iteration.
Load Flow Studies 205
206 Modern Power System Analysis

Fig. 6.15 Flow chart for load flow solution by the Gauss–Siedel iterative method using YBUS
 Load Flow Studies 207

Solution

Table 6.5 Input data

Bus Pi, pu Qi, pu Vi, pu Remarks

1 – – 1.04 –0º Slack bus
2 0.5 – 0.2 – PQ bus
3 – 1.0 0.5 – PQ bus
4 0.3 – 0.1 – PQ bus

2 . 991 - j 9 . 253
= = 1.025 – j0.0093 pu
3 - j9
The YBUS for the sample system has been calculated earlier in Example 6.3(b) (i.e., the dotted line is
assumed to be connected). In order to approach the accuracy of a digital computer, the computations given
below have been performed on an electronic calculator.
Bus voltages at the end of the first iteration are calculated using Eq. (6.41).
È ˘
1 Í P2 - jQ2
V 12 = - Y21 V1 - Y23 V 30 - Y24 V 40 ˙
Y22 Í V 20 *
Î ( ) ˙
˚
1 È 0 .5 + j 0 . 2 ˘
= Í - 1. 04 (-2 + j 6) - ( - 0 .666 + j 2) - ( - 1 + j 3)˙
Y22 Î 1 - j 0 ˚
4 . 246 - j11. 04
= = 1.019 + j0.046 pu
3 .666 - j11
È ˘
1 Í P3 - jQ3
V 13 = - Y31 V1 - Y32 V21 - Y34 V 40 ˙
Y33 Í V30 *
Î ( ) ˙
˚
1 È - 1 - j0 .5 ˘
= Í - 1.04(-1 + j 3) - (-0.666 + j 2)(1.019 + j 0.046) - (-2 + j 6) ˙
Y33 Î 1 - j 0 ˚
2 .81 - j11. 627
= = 1.028 – j0.087 pu
3 . 666 - j11
È ˘
1 Í P4 - jQ4
V41 = - Y41 V1 - Y42 V21 - Y43 V31 ˙
Y44 Í V40 *
Î ( ) ˙
˚
È
1 0 . 3 + j 0 .1 ˘
= Í - (-1 + j 3)(1.019 + j 0.046) - (-2 + j 6)(1.028 - j 0.087) ˙
Y44 Î 1 - j 0 ˚

Example 6.9 In Example 6.8, let bus 2 be a PV bus now with |V2| = 1.04 pu. Once again assuming
a flat voltage start, find Q2, d2, V3, V4 at the end of the first GS iteration.
Given 0.2 £ Q2 £ 1 pu
208 Modern Power System Analysis

Solution From Eq. (6.46), we get (Note: d 20 = 0, i.e., V 20 = 1.04 + j0)

Q12 = – Im ((V 02)* Y21 V1 + (V 02)* (Y22 V 02 + Y23 V 03 + Y24 V 04))
= – Im (1.04 (– 2 + j6) 1.04 + 1.04 ((3.666 – j11)1.04 + (– 0.666 + j2) + (– 1 + j3)))
= – Im (–0.0693 – j0.2097) = 0.2079 pu
From Eq. (6.47),
È ˘
1 Í P2 - jQ21
d 21 = – - Y21 V1 - Y23 V30 - Y24 V40 ˙
Î ( )
Y22 Í V20 * ˙
˚
1 È 0 . 5 - j 0 . 2079 ˘
= – Í - (-2 + j 6)(1.04 + j 0) - (-0.666 + j 2)(1 + j 0) - (-1 + j 3)(1 + j 0) ˙
Y22 Î 1. 04 - j 0 ˚
È 4 . 2267 - j11. 439 ˘
= –Í ˙ = –(1.0512 + j0.0339)
Î 3 . 666 - j11 ˚
= 1.84658° = 0.032 red
V 21 = 1.04 (cos d 12 + j sin d 21)
= 1.04 (0.99948 + j0.0322)
= 1.03946 + j0.03351
È ˘
1 Í P3 - jQ3
V31 = - Y31 V1 - Y32 V21 - Y34 V 40 ˙
Î ( )
Y33 Í V30 * ˙
˚
1 È - 1 - j0 .5 ˘
= Í - (-1 + j 3)1.04 - (-0.66 + j 2)(1.03946 + j 0.03351) - (-
-2 + j 6) ˙
Y33 Î 1 - j 0 ˚
2.7992 - j11.6766
= = 1.0317 – j0.08937
3.666 - j11
È ˘
1 Í P4 - jQ4
V41 = - Y41 V1 - Y42 V21 - Y43 V 31 ˙
Î ( )
Y44 Í V40 * ˙
˚
1 È 0 . 3 + j 0 .1 ˘
= Í - (-1 + j 3)(1.0394 + j 0.0335) - (-2 + j 6)(1.0317 - j 0.08937) ˙
Y44 Î 1 - j 0 ˚
2 .9671 - j8 . 9962
= = 0.9985 – j0.0031
3 - j9
Now, suppose the permissible limits on Q2 (reactive power injection) are revised as follows:
0.25 £ Q2 £ 1.0 pu
It is clear that other data remaining the same, the calculated Q2 (= 0.2079) is now less than the Q2 min.
Hence Q2 is set equal to Q2 min i.e.,
Q2 = 0.25 pu
Bus 2, therefore, becomes a PQ bus from a PV bus. Therefore, | V2 | can no longer remain fixed at 1.04 pu.
The value of V2 at the end of the first iteration is calculated as follows. (Note: V 20 = 1 + j0 by virtue of a flat
start).
 Load Flow Studies 209

1 È P2 - jQ2 ˘
V21 = Í 0 *
- Y21 V1 - Y23 V30 - Y24 V 40 ˙
Y22 ÍÎ (V 2 ) ˙˚
1 È 0 . 5 - j 0 . 25 ˘
= Í - (-2 + j 6)1.04 - (-0.666 + j 2) - (-1 + j 3) ˙
Y22 Î 1 - j 0 ˚
4 . 246 - j11. 49
= = 1.0559 + j0.0341
3 . 666 - j11
1 È P3 - jQ3 ˘
V31 = Í 0 *
- Y31 V1 - Y32 V21 - Y34 V 40 ˙
Y33 ÍÎ (V3 ) ˙˚
1 È - 1 - j0 .5 ˘
= Í - (-1 + j 3)1.04 - (-0.666 + j 2)(1.0559 + j 0.0341) - (-2 + j 6) ˙
Y33 Î 1 - j 0 ˚
2 . 8112 - j11. 709
= = 1.0347 – j0.0893 pu
3 . 666 - j11
1 È P4 - jQ4 ˘
V41 = Í 0 *
- Y41 V1 - Y42 - V21 - Y43 V 31 ˙
Y44 ÍÎ (V 4 ) ˙˚
1 È 0 . 3 + j 0 .1 ˘
= Í - (-1 + j 3)(1.0509 + j 0.0341) - (-2 + j 6)(1.0347 - j 0.0893) ˙
Y44 Î 1 - j 0 ˚
4 . 0630 - j 9 . 4204
= = 1.0775 + j0.0923 pu
3 - j9

6.6 NEWTON–RAPHSON METHOD

The Newton–Raphson (NR) method is a powerful method of solving nonlinear algebraic equations. It
works faster, and is sure to converge in most cases as compared to the Gauss–Siedel (GS) method. (For
its convergence properties see Appendix I). It is indeed the practical method of load flow solution of large
power networks. Its only drawback is the large requirement of computer memory, which can be overcome
through a compact storage scheme. Convergence can be considerably speeded up by performing the first
iteration through the GS method, and using the values so obtained for solving the NR iterations.
We consider a set of n nonlinear algebraic equations
fi(x1, x2, ..., x n) = 0 where i = 1, 2, ..., n (6.49)
We assume initial values of unknowns as x01, x02, º, x0n. Let Dx01, Dx02, º, Dx0n be the corrections to be
found out, which on being added to the initial values, give the actual solution. Therefore,
fi(x01 + Dx01, º, x0n + D x0n) = 0; i = 1, 2, ..., n (6.50)
Expanding these equations around the initial values by Taylor series, we have
ÈÊ ∂ f ˆ 0 Ê ∂ fi ˆ
0 ˘
f 0i (x01, ..., x0n) + ÍÁ n ˙ + higher order terms = 0
0 0
i
D x + + D x (6.51)
ÍË ∂ x1 ˜¯ ÁË ∂ x ˜¯
1
˙
Î n ˚
0 0
Ê∂ f ˆ Ê∂f ˆ
where Á i ˜ , … , Á i ˜ are the derivatives of fi wrt x1, x2, ..., xn evaluated at x10, ..., x0n.
Ë ∂ x1 ¯ Ë ∂ xn ¯