UNIT 2 POWER FLOW ANALYSIS

1
FORMULATION OF POWER FLOW PROBLEM

Power flow analysis is the most fundamental study to be performed in a

power system both during the Planning and Operational phases. It
constitutes the major portion of electric utility. The study is concerned with
the normal steady state operation of power system and involves the
determination of bus voltages and hence the power flows for a given
network configuration and loading condition.

The results of power flow analysis help to know

1 the present status of the power system, required for continuous

monitoring.
2 alternative plans for system expansion to meet the ever
increasing demand.

The mathematical formulation of the power flow problem results in a

system of non-linear algebraic equations and hence calls for an
iterative technique for obtaining the solution. Gauss-Seidel method and
Newton Raphson ( N.R.) method are the commonly used to get the power
flow solution.

2
With reasonable assumptions and approximations, a power system may be
modelled as shown in Fig. 2.1 for purpose of steady state analysis.

G PD1  jQD1 PD2  jQD2

1 Static Capacitor
V1  1 V2  2 2
PG1  jQG1

5 3
V5  V4  4 V3  3
PG3  jQG3
5 4
PD4  jQD4
G PG5  jQG5 G

Fig. 2.1 Typical power system network

3
The model consists of a network in which a number of buses are
interconnected by means of lines which may either transmission lines or
power transformers. The generators and loads are characterized by the
complex powers flowing into and out of buses respectively. Each
transmission line is characterized by a lumped impedance and a line
charging capacitance. Static capacitors or reactors may be located at
certain buses either to boost or buck the load-bus voltages at times of
need.

The scope of Power Flow problem may be stated as follows:

Given the network configuration and the loads at various buses,

determine a schedule of generation so that the bus voltages and hence
line flows remain within security limits.

4
A more specific statement of the problem will be made subsequently
after taking into consideration the following three observations.

1 For a given load, we can arbitrarily select the schedules of all the
generating buses, except one, to lie within the allowable limits of
the generation. The generation at one of the buses, called as the
slack bus, cannot be specified beforehand since the total generation
should be equal to the total demand plus the transmission losses,
which is not known unless all the bus voltages are determined.

2 Once the phasor voltages at all the buses are known, all other
quantities of interest such as line flows, transmission losses and
generation at the buses can easily be determined. Hence the
foremost aim of the power flow problem is to solve for the bus
voltages.

5
 3 It will be convenient to use the Bus Power Specification which is
defined as the difference between the specified generation and load at
a bus. Thus for the kth bus, the bus power specification Sk is given by

Sk = PIk + j QIk

= (PGk + j QGk ) - (PDk + j QD k )

= (PGk - PDk ) + j (QGk - j QD k ) (2.1)

In view of the above three observations Power Flow Problem may be

defined as that of determining the phasor voltages at all the buses, given
the network configuration and the bus power specifications at all the
buses, except the slack bus.

6
CLASSIFICATION OF BUSES

There are four quantities associated with each bus. They are PI, QI, │V│ and δ.

Here PI is the real power injected into the bus

QI is the reactive power injected into the bus

ΙVΙ is the magnitude of the bus voltage

δ is the phase angle of the bus voltage

Any two of these four may be treated as independent variables ( i.e.

specified ) while the other two may be computed by solving the bus
power equations. Depending on which of the two variables are specified,
buses are classified into three types. There are three types of bus
classification based on practical requirements.

7
Slack bus

In a power system with N buses, power flow problem is primarily concerned

with determining the 2N bus voltage variables, namely N voltage
magnitudes and N voltage phase angles. These can be obtained by
solving the 2N bus power (Real and Reactive) equations provided there are
2N power specifications. However as discussed earlier the real and
reactive power injection at the SLACK BUS cannot be specified
beforehand.

This leaves us with no other alternative but to specify two variables │Vs │
and δs arbitrarily for the slack bus so that 2( N-1 ) variables can be
solved from 2( N-1 ) known bus power specifications.

Incidentally, the specification of │Vs │ helps us to fix the voltage level of

the system and the specification of δs serves as the phase angle reference
for the system.

Thus for the slack bus, both │V│ and δ are specified and PI and QI
are to be determined. In fact PI and QI of the slack bus shall be computed
at the end, when all the │V│ s and δs are solved for.
8
Generator bus

In a generator bus which supplies real power, it is customary to maintain

the bus voltage magnitude at a desired level which can be achieved in
practice by proper reactive power injection. Such buses are termed as
Generator Buses or P – V buses. At these buses PI and │V│ are specified
and QI and δ are to be solved for. They also referred as voltage controlled
buses.

Load bus

The buses where there is no controllable generation are called as Load

Buses or P – Q buses. At the load buses, both PI and QI are specified
and │V│ and δ are to be solved for.

│V│ δ PI QI

Slack bus Can be computed after bus voltages are solved

Load bus (P-Q bus) ? ?

Generator bus (P-V bus) ?

?

Fig. 2.2 Three types of buses

9
Development of mathematical model:
Suppose we like to determine the length and breadth of a rectangle knowing its
perimeter as 140 m and diagonal as 50 m. To determine the length x and width y,
following equations are formed.

2 (x + y) = 140

x 2  y 2 = 50

Here 140 m and 50 m are the specified values of perimeter and diagonal while the

expressions on the left-hand side of the equations, 2 (x + y) and x 2  y 2 , are

the formulas for the corresponding specified quantity.

The two equations formed are to be solved for the unknown variables x and y.

10
Development of power flow model

The power flow model will comprise of a set of simultaneous non-linear

algebraic equations. These equations relate the known complex power
injections to the unknown phasor bus voltages. The solution of this
model will yield all the bus voltages. Non-linear algebraic equations can be
solved using Gauss-Seidel or Newton Raphson method.

For the given power system it is possible to write

i) Network equations

ii) Bus power equations

These equations are used for the development of power flow model. The
solution of power flow model will yield all the bus voltages.

Once the bus voltages are known, the power flows in the lines and
transformers can be determined using the power flow equations.

11
Network equations

Network equations can be written in a number of alternative forms. Let

us choose the bus frame of reference in admittance form which is most
economical one from the point of view of computer time and memory
requirements.

The equations describing the performance of the network in the bus

admittance form is given by

I = YV (2.2)

Here I is the bus current vector,

V is the bus voltage vector and

Y is the bus admittance matrix of transformer and transmission network

In expanded form these equations are

I1   Y11 Y12  Y1N   V1 

I  Y Y  Y  V 
  = 
2 21 22 2N   2 (2.3)
        
     
 N
I  N1
Y YN2  YNN   VN 
12
 I1   Y11 Y12  Y1N   V1 
I  Y Y  Y  V 
  = 
2 21 22 2N   2 (2.3)
        
     
 
I N  YN1 YN2  YNN   VN 

Typical element of the bus admittance matrix is

Yij = |Yij |  θij = |Yij | cos θij + j |Yij | sin θij = Gij + j B ij (2.4)

Voltage at a typical bus i is

Vi = |Vi |  δi = |Vi | ( cos δi + j sin δi ) (2.5)

The current injected into the network at bus k is given by

I k = Yk1 V1 + Yk2 V2 + ...... + YkN VN

N
Thus I k = Y
m 1
km Vm for k = 1, 2, .... , N (2.6)

13
 N
Thus I k = Y
m 1
km Vm for k = 1, 2, .... , N (2.6)

Bus power equations

In addition to the linear network equations given by eqn. (2.3), bus

power equations should also be satisfied in the power flow problem.
These bus power equations introduce non- linearity into the power flow
model.

The complex power entering at the k th bus is given by

Pk  j Qk  Vk I k* (2.7)

This power should be the same as the bus power specification. Thus we
have

PIk  j QIk  Vk Ik* for k = 1, 2, .... , N (2.8)

k ≠s
As bus currents are intermediate quantities, they can be eliminated using (2.6)
in (2.8) to obtain the power flow model.

Before proceeding further, let us see how the non-linear eqs. can be solved.
14
Gauss-Seidel method

Gauss-Seidel method is used to solve a set of algebraic equations.

Consider

a 11 x 1  a 12 x 2    a 1N x N  y 1
a 21 x 1  a 22 x 2    a 2N x N  y 2
   
a N1 x 1  a N2 x 2    a NN x N  y N

Specifically

a k1 x 1  a k2 x 2    a kk x k    a kN x N  y k
N
Thus a kk x k  y k  a
m1
km xm
mk

 N

1 
This gives xk  yk   a km x m 
a kk  m 1

 mk 
for k = 1, 2, .... , N

15
  N

1 
This gives xk  y k   a km x m 
a kk  m1

 mk 

for k = 1, 2, .... , N

Initially, values of x 1 , x 2 ,, x N are assumed. Updated values are calculated

using the above equation. While calculating xk in iteration h  1 , up to m  k  1,
values of x m calculated in h  1 iteration are used and for m  k  1 to N ,
values of x m calculated in h iteration are used. Thus

1  k 1 N 
x h 1
k  yk 
a kk 
 a km x h1
m   km m 
a x h
(2.9)
m 1 m k 1 

16
Gauss-Seidel method for power flow solution

In this method, first an initial estimate of bus voltages is assumed. By

substituting this estimate in the given set of equations, a second
estimate, better than the first one, is obtained. This process is repeated
and better and better estimates of the solution are obtained until the
difference between two successive estimates becomes lesser than a
prescribed tolerance.

Let us now find the required equations for calculating the bus voltages.

From the network equations, we have

I k  Yk1 V1  Yk2 V2    Ykk Vk    YkN VN

N N
Thus Ik  Y
m 1
km Vm  Ykk Vk  Y
m 1
km Vm
(2.10)
mk

17
 N N
Thus Ik  Y
m1
km Vm  Ykk Vk  Y
m 1
km Vm (2.10)
mk

From the bus power specifications, we have

PIk  j QIk  Vk Ik* for k  1,2,,N (2.11)

ks

I k the intermediate variable, can be eliminated. Taking conjugate of Eq. (2.11)

PI k  j QI k
PI k  j QI k  Vk* I k ; Therefore Ik  (2.12)
Vk*

N
PI k  j QI k
From equations (2.10) and (2.12) we get Ykk Vk   Ykm Vm 
m 1 Vk*
mk

 
1  PI k  jQI k N
Thus Vk    Ykm Vm 
Ykk  Vk* m 1

 m k 

PI k  j QI k 1 N
Ykm (2.13)

Ykk Vk*
 Y
m 1
Vm for k = 1, 2, .... , N
k ≠s
kk
mk

18
 PI  j QI k 1 N
Ykm
Vk  k
Ykk Vk*
  Y
Vm k = 1, 2, .... , N (2.13)
m  1 kk
mk
k ≠s

A significant reduction in computing time for a solution can be achieved

by performing as many arithmetic operations as possible before initiating
the iterative calculation. Let us define

P I k  j QI k
 Ak (2.14)
Ykk

Ykm
and  Bkm (2.15)
Ykk

Having defined Ak and Bkm equation (2.13) becomes

N
A
Vk  k* 
Vk
B
m 1
km Vm k  1,2, , N ks (2.16)
mk

19
 N
A
Vk  k* 
Vk
B
m 1
km Vm k  1,2, , N ks (2.16)
mk

When Gauss-Seidel iterative procedure is used, the voltage at the k th

bus during h  1 th iteration, can be computed as

k 1 N
A
Vh1
k  hk * 
Vk
B
m1
km
h 1
V
m  B
mk 1
km Vmh

k  1,2,, N (2.17)
ks

If at every stage the latest bus voltages are going to used, then the voltage at
bus k can be calculated using Eq. (2.17)

20
Power flow equations

Once the bus voltages are determined, the power flows in the lines can be
computed as shown below.

ikm
ykm
k m
yk0 ym0

Fig. 2.3 Circuit for line flow calculation

Current i km  ( Vk  Vm ) ykm  Vk yk0

(2.18)

Power flow from bus k to bus m is Skm  Vk i km

*
(2.19)

Substituting equation (2.18) in equation (2.19)

S km  Vk [ ( Vk*  Vm* ) y km
*
 Vk* y k0
*
] (2.20)

21
Skm  Vk [ ( Vk*  Vm* ) y km
*
 Vk* y k0
*
] (2.20)

Similarly, power flow from bus m to bus k is

S mk  Vm [ ( Vm*  Vk* ) y km
*
 Vm* y m0
*
] (2.21)

The line loss in the transmission line k  m is given by

S L k m  S km  S mk (2.22)

Total transmission loss in the system is SL  S

k m
L k m
(2.23)
over all
the lines

It is to be noted that while we calculate the bus voltages, parameters used in eqs.
(2.14) and (2.15) are the elements of bus admittance matrix. However, when we
calculate power flow using eq. (2.20) primitive admittances are used.

22
Example 2.1

For a power system, the transmission line impedances and half line charging
admittances in p.u. on a 100 MVA base are given in Table 1. The scheduled
generations and loads on different buses are given in Table 2. Taking the slack bus
voltage as 1.06 + j 0.0 and using a flat start perform the power flow analysis and
obtain the bus voltages, transmission loss and slack bus power.

Table 1 Transmission line data:

Bus code Line Impedance

Sl. No. HLCA
k-m z km
1 1–2 0.02 + j 0.06 j 0.030
2 1–3 0.08 + j 0.24 j 0.025
3 2–3 0.06 + j 0.18 j 0.020
4 2–4 0.06 + j 0.18 j 0.020
5 2–5 0.04 + j 0.12 j 0.015
6 3–4 0.01 + j 0.03 j 0.010
7 4–5 0.08 + j 0.24 j 0.025

23
Table 2 Bus data:

Bus code Generation Load

Remark
k PGk in MW QGk in MVAR PDk in MW QD k in MVAR
1 --- --- 0 0 Slack bus
2 40 30 20 10 P – Q bus
3 0 0 45 15 P – Q bus
4 0 0 40 5 P – Q bus
5 0 0 60 10 P – Q bus

24
Solution

Flat start means all the unknown voltage magnitude are taken as 1.0 p.u. and
all unknown voltage phase angles are taken as 0.

V1  1.06  j 0
Thus initial solution is
V2(0)  V3(0)  V4(0)  V5(0)  1.0  j 0

STEP 1

For the transmission system, the bus admittance matrix is to be calculated.

Bus Line Impedance

Sl. Line admittance
code HLCA
No.
k-m
z km ykm
1 1–2 0.02 + j 0.06 5 - j 15 j 0.030
2 1–3 0.08 + j 0.24 1.25 – j 3.75 j 0.025
3 2–3 0.06 + j 0.18 1.6667 – j 5 j 0.020
4 2–4 0.06 + j 0.18 1.6667 – j 5 j 0.020
5 2–5 0.04 + j 0.12 2.5 – j 7.5 j 0.015
6 3–4 0.01 + j 0.03 10 – j 30 j 0.010
7 4–5 0.08 + j 0.24 1.25 – j 3.75 j 0.025

25
 Bus Line Impedance
Sl. Line admittance
code HLCA
No.
k-m
z km ykm
1 1–2 0.02 + j 0.06 5 - j 15 j 0.030
2 1–3 0.08 + j 0.24 1.25 – j 3.75 j 0.025
3 2–3 0.06 + j 0.18 1.6667 – j 5 j 0.020
4 2–4 0.06 + j 0.18 1.6667 – j 5 j 0.020
5 2–5 0.04 + j 0.12 2.5 – j 7.5 j 0.015
6 3–4 0.01 + j 0.03 10 – j 30 j 0.010
7 4–5 0.08 + j 0.24 1.25 – j 3.75 j 0.025

Y22 = (5 – j15) + (1.6667 – j5) + (1.6667 – j5) + (2.5 – j7.5) + j 0.03 + j 0.02 + j 0.02 + j 0.015

= 10.8334 – j 32.415

Similarly Y33 = 12.9167– j 38.695 ; Y44 = 12.9167 – j 38.695 ; Y55 = 3.75 – j 11.21

1 2 3 4 5

1 --- --- --- --- ---

2 -5 +j15 10.8334–j32.415 -1.6667 + j5 -1.6667+j5 -2.5 + j7.5

Y= 3 -1.25 + j3.75 -1.6667 + j5 12.9167–j38.695 -10 + j30 0

4 0 -1.6667 + j5 -10 + j30 12.9167-j38.695 -1.25 + j3.75

5 0 -2.5+j7.5 0 -1.25+j3.75 3.75-j11.21

26
STEP 2

Calculation of elements of A vector and B matrix.

P I k  j QI k Ykm
Ak  and Bkm 
Ykk Ykk
PI 2  j QI 2  (40  j 30)  (20  j 10)  (20  j 20) MVA
1
P I 2  j QI 2  ( 20  j 20 )  0.2  j 0.2
100
PI 2  j QI 2  0.2  j 0.2
P I 2  j QI 2 0.2  j 0.2
A2    0.0074  j 0.0037
Y22 10.8334  j 32.415

Similarly A 3 , A 4 and A 5 are calculated .

Y 21  5  j5
B21     0.46263  j 0.00036
Y22 10.8334  j 32.415

Other elements of matrix B can be calculated in a similar manner.

27
Thus

1 -----

2 0.00740 + j 0.00370

A = 3 -0.00698 - j 0.00930

4 -0.00427 - j 0.00891

5 -0.02413 - j 0.04545

--- --- --- --- ---

- 0.46263 - 0.15421 - 0.15421 - 0.23131
+ j 0.00036 --- + j 0.00012 + j 0.00012 + j 0.00018
- 0.09690 - 0.12920 - 0.77518
B = + j 0.00004 + j 0.00006 --- + j 0.00033 0
- 0.12920 - 0.77518 - 0.09690
0
+ j 0.00006 + j 0.00033 --- + j 0.00004
- 0.66881 - 0.33440
0
+ j 0.00072 0 + j 0.00036 ---

28
STEP 3

Bus voltages can be calculated using latest bus voltages. Present bus
voltages are:

V1= 1.06 + j 0.0; V 2 = V3 = V4 = V5 = 1.0 + j 0.0

Iterative computation of bus voltage can be carried out as shown.

First iteration:

A2
V2   B 21 V1  B 23 V3  B 24 V4  B 25 V5
V2*
0.00740  j 0.00370
  (  0.46263  j 0.00036 ) ( 1.06  j 0.0 0 )
1.0  j 0
 (  0.15421  j 0.00012 ) ( 1.0  j 0.0 )  (  0.15421  j 0.00012 ) ( 1.0  j 0.0 )
 (  0.23131  j 0.00018 ) ( 1.0  j 0.0 )

 1.03752  j 0.00290

This value of voltage V2(1) will replace the previous value of voltage V2(0)
before doing subsequent calculations of voltages.

29
The rate of convergence of the iterative process can be increased by
applying an ACCELERATION FACTOR α to the approximate solution
obtained. For example on hand, for bus 2, change in bus voltage

∆V2 = V2(1) – V2(0) = (1.03753 + j 0.00290) – (1.0 + j 0)

= 0.03752 + j 0.00290

The accelerated value of the bus voltage is obtained as

V2(1) = V2(0) + α ∆V2

By assuming α = 1.4 accelerated bus voltage

V2(1) = (1.0 + j 0) + 1.4 (0.03752 + j 0.00290) = 1.05253 + j 0.00406

This new value of voltage V2(1) will replace the previous value of the bus
voltage V2(0) and is used in the calculation of voltages for the remaining
buses.

30
In general

Vk h+1accld = Vk h + α (Vk h+1 – Vk h) (2.24)

where k is the bus at which the voltage is calculated and h+1 is the current
iteration count.

The process is continued for the remaining buses to complete one

iteration. Now

V1= 1.06 + j 0.0; V 2 = 1.05253 + j 0.00406 V3 = V4 = V5 = 1.0 + j 0.0

For the next bus 3

A3
V3  *
 B 31 V1  B 32 V2  B 34 V 4
V3
 0.00698  j 0.00930
  (  0.09690  j 0.00004 ) ( 1.06  j 0 )
1.0  j 0
 (  0.12920  j 0.00006 ) ( 1.05253  j 0.00406 )
 (  0.77518  j 0.00033 ) ( 1.0  j 0 )
 1.00690  j 0.00921
31
∆V3 = V3(1) – V3(0) = (1.00690 - j 0.00921) – (1.0 + j 0)

= 0.00690 - j 0.00921

The accelerated value of the bus voltage is obtained as

V3(1) = V3(0) + α ∆V3

V3(1) = (1.0 + j 0) + 1.4 (0.00690 - j 0.00921) = 1.00966 - j 0.01289

Continuing this process of calculation, at the end of first iteration, the

bus voltages are obtained as

V1 = 1.06 + j 0.0

V2 = 1.05253 + j 0.00406 Acceleration factor 1.4 is

common for both the real
V3 = 1.00966 – j 0.01289 part and imaginary parts
of voltage.
V4 = 1.01599 – j 0.02635

V5 = 1.02727 – j 0.07374

If α and β are the acceleration factors for the real and imaginary components
of voltages respectively, the accelerated values can be computed as

ek h+1 = ek h + α (ek h+1 - ek h) fk h+1 = fk h + β (fk h+1 - fk h) (2.25)

32
Convergence

The iterative process must be continued until the magnitude of change

of bus voltage between two consecutive iterations is less than a certain
level ε for all bus voltages. We express this in mathematical form as

ΔVmax = max. of |Vkh+1 – Vkh| for k = 1,2, …… , N k≠s

(2.26)
and ΔVmax < ε

If ε1 and ε2 are the tolerance level for the real and imaginary parts of
bus voltages respectively, then the convergence criteria will be

ΔVmax 1 = max. of |ekh+1 – ekh| ΔVmax 2 = max. of |fkh+1 – fkh|

for k = 1,2, …… , N k≠s (2.35)

(2.27)

ΔVmax 1 < ε1 and ΔVmax 2 < ε2

For the problem under study ε1 = ε2 = 0.0001

The final converged bus voltages obtained after 10 iterations are given below.

V1 = 1.06 + j 0.0 V2 = 1.04623 - j 0.05126 V3 = 1.02036 – j 0.08917

V4 = 1.01920 – j 0.09504 V5 = 1.01211 – j 0.10904

33
V1 = 1.06 + j 0.0 V2 = 1.04623 - j 0.05126 V3 = 1.02036 – j 0.08917

V4 = 1.01920 – j 0.09504 V5 = 1.01211 – j 0.10904

Computation of line flows and transmission loss

Line flows can be computed from

Skm  Vk [ (Vk*  Vm* ) y km

*
 Vk* y k0
*
]
S12  V1 [ ( V1*  V2* ) y 12
*
 V1* y 10
*
]
 (1.06  j 0 ) [ { ( 1.06  j 0 )  ( 1.04623  j 0.05126 ) } ( 5  j 15 )
 { ( 1.06  j 0) ( 0.0  j 0.03 ) } ]
 ( 0.888  j 0.086 )

Similarly
S 21  V2 [ ( V2*  V1* ) y 12
*
 V2* y 20
*

 ( 1.04623  j 0.05126 ) [ { ( 1.04623  j 0.05126)  ( 1.06  j 0 ) }( 5  j 15)

 { ( 1.04623  j 0.05126 ) (0.0  j 0.03 ) } ]
 (  0.874  j 0.062 )

34
Power loss in line 1 – 2 is

SL 1 2  S12  S 21  (0.888  j 0.086)  (0.874  j 0.062)

 ( 0.014  j 0.024 )

Power loss in other lines can be computed as

SL 1 3  0.012  j 0.019
SL 23  0.004  j 0.033
SL 24  0.004  j 0.029
SL 25  0.011  j 0.002
SL 34  0.0  j 0.019
S L 4  5  0.0  j 0.051

Total transmission loss SL = ( 0.045 - j 0.173 ) i.e.

Real power transmission loss = 4.5 MW

Reactive power transmission loss = 17.3 MVAR ( Capacitive )

35
Computation of slack bus power

Slack bus power can be determined by summing up the powers flowing

out in the lines connected at the slack bus and the load at the slack bus.

S gs G
S ds
1 3
S13

S12

In this case, load at slack bus is zero and hence slack bus power is
S gs  S 13  S 12

S 13  V1 [ ( V1*  V3* ) y 13
*
 V1* y 10
*
]
 (1.06  j 0 ) [ { ( 1.06  j 0 )  ( 1.02036  j 0.08917 ) } ( 1.25  j 3.75 )
 { ( 1.06  j 0) ( 0.0  j 0.025 ) } ]
 ( 0.407  j 0.011)

S gs  S13  S 12  ( 0.407  j 0.011)  ( 0.888  j 0.086 )  ( 1.295  j 0.075 )

Thus the power supplied by the slack bus:

Real power = 129.5 MW Reactive power = 7.5 MVAR (Capacitive ) 36

Voltage controlled bus

In voltage controlled bus k , net real power injection PIk and voltage

magnitude Vk are specified. Normally QI max and QI min will also be

specified for voltage controlled bus. Since QIk is not known for P-V bus,
PI k  j QI k
A k given by cannot be calculated. An expression for QIk
Ykk

can be developed as shown below.

N
We know Ik  Y
m 1
km Vm and PIk  j QIk  Vk Ik* i.e. PIk  j QIk  Vk* Ik

Denoting Ykm  Yk m θk m and Vk  Vk δk we have

N N
PIk  j QI k  Vk   δk Y
m 1
km Vm θk m  δm  
m 1
Vk Vm Yk m θ k m  δ m  δk

N
Thus QIk    Vk Vm Yk m sin θk m  δm  δk (2.28)
m 1

The value of Vk to be used in equation (2.28) must satisfy the relation

Vk  Vk specified
37
Because of voltage updating in the previous iteration, the voltage
magnitude of the voltage controlled bus might have been deviated from
the specified value. It has to be pulled back to the specified value,
using the relation

fkh
1
Adjusted voltage V  Vk
h
k specified δ where δ  tan
h
k
h
k ( h ) taking Vkh  e hk  j fkh
ek
(2.29)

Using the adjusted voltage Vkh as given in eqn. (2.29), net injected

reactive power Q Ihk can be computed using eqn. (2.28).

As long as Q Ihk falls within the range specified, Vkh can be replaced by

the Adjusted Vkh and the corresponding A k can be computed.

38
In case if Q Ihk falls beyond the limits specified, Vkh should not be

replaced by Adjusted Vkh , Q Ihk is set to the limit and A k can be

calculated. In this case bus k is changed from P – V to P – Q type.

Once the value of Ak is known, further calculation to find Vkh  1 will

be the same as that for P – Q bus.

Complete flow chart for power flow solution using Gauss-Seidel method
is shown in Fig. 2.4. The extra calculation needed for voltage controlled
bus is shown between X – X and Y – Y.

39
 START

READ LINE DATA & BUS DATA

FORM Y MATRIX
ASSUME Vk( 0 ) k  1,2,  , N k  s
COMPUTE Ak FOR P – Q BUSES
COMPUTE Bkm SET h = 0

SET k = 1 AND VMAX  0.0

k:s YES

X -------------------------------------------- NO-----------------------------------------
NO k : V.C. Bus X
YES
COMPUTE  kh ADJUSTED VOLTAGE, QI kh

 
QI kh : QI k MAX QI kh : QI k MIN
> <
h h
REPLACE QI k BY QI k MAX REPLACE QI k BY QI k MIN REPLACE Vkh BY ADJ Vkh

Y COMPUTE Ak Y
-----------------------------------------------------------------------------------------------------------------

COMPUTE Vkh  1 ; COMPUTE V  Vkh 1  Vkh

YES
V  VMAX SET VMAX  V
NO
REPLACE Vkh BY Vkh  1

SET k = k + 1

k:N IS V MAX <  YES

 > NO
SET h = h + 1

COMPUTE LINE FLOWS, SLACK BUS POWER

PRINT THE RESULTS
40
STOP
Introduction

Gauss-Seidel method of solving the power flow has simple problem formulation and
hence easy to explain. However, it has poor convergence characteristics. It takes large
number of iterations to converge. Even for the five bus system discussed in Example 2.1,
it takes 10 iterations to converge.

Newton Raphson (N.R.) method of solving power flow is based on the Newton Raphson
method of solving a set of non-linear algebraic equations. N. R. method of solving power
flow problem has very good convergence characteristics. Even for large systems it takes
only two to four iterations to converge.

Newton Raphson method of solving a set of non-linear equations

Let the non-linear equations to be solved be

f1 (x1 , x2 ,, xn )  k1

f 2 (x1 , x2 ,, xn )  k 2
(2.30)


fn (x1 , x2 ,, xn )  k n
41
Let the initial solution be x1(0) , x (0) (0)
2 , , x n

If k 1  f1 (x1(0) , x(0) (0)

2 ,  , x n ) 0

k 2  f 2 (x 1(0) , x (0) (0)

2 ,  , x n ) 0
 = 
 

k n  fn (x1(0) , x(0) (0)

2 ,  , x n ) 0

then the solution is reached. Let us say that the solution is not reached. Assume
Δx1 , Δx 2 ,, Δxn are the corrections require d on x1(0) , x (0) (0)
2 ,, x n respectively.
Then

f1 [(x 1(0)  Δx1 ), (x (0)

2  Δx 2 ),, (x n  Δx n )]  k 1
(0)

f 2 [(x 1(0)  Δx1 ), (x (0)

2  Δx 2 ), , (x n  Δx n )]  k 2
(0)

(2.31)
   

f n [(x 1(0)  Δx1 ), (x (0)

2  Δx 2 ), , (x n  Δx n )]  k n
(0)

42
Each equation in the above set can be expanded by Taylor’s theorem around
x1(0) , x (0) (0)
2 ,, x n
. For example, for the first equation

f1 [(x 1(0)  Δx1 ), (x (0)

2  Δx 2 ),, (x n  Δx n )]  k 1 expansion is
(0)

 f1 f 1 f 1
2 ,, x n ) 
f1 (x 1(0) , x (0) (0)
Δx 1  Δx 2    Δx n  φ 1  k 1
x 1 x 2 x n

where φ 1 is a function of higher derivatives of f 1 and higher powers of

Δx1 , Δx 2 ,, Δxn . Neglecting φ 1 and also following the same for other equations,
we get

 f1 f f
2 ,, x n ) 
f1 (x 1(0) , x (0) (0)
Δx1  1 Δx 2    1 Δx n  k 1
x 1 0 x 2 0 x n 0

f 2 f f
2 ,, x n ) 
f 2 (x1(0) , x (0) (0)
Δx1  2 Δx 2    2 Δx n  k 2
x 1 0 x 2 0 x n 0
(2.32)
   

f n f f
2 ,, x n ) 
f n (x1(0) , x (0) (0)
Δx 1  n Δx 2    n 0 Δx n  k n
x 1 0 x 2 0 x n
43
The matrix form of equations (2.31) is

f1 f1 f1

 Δx1 k 1  f 1 ( x1(0) , x (0) (0)
2 , , x n )
x1 0 x 2 0 x n 0

f 2 f 2 f 2
 Δx 2 k 2  f 2 ( x 1(0) , x (0) (0)
2 , , x n ) (2.33)
x 1 0 x 2 0 x n 0 =
    
    

f n f n f n
 Δx n k n  f n ( x 1(0) , x (0) (0)
2 , , x n )
x 1 0 x 2 0 x n 0

The above equation can be written in a compact form as

F' ( X(0) ) ΔX  K  F ( X(0) ) (2.34)

44
F' ( X (0) ) ΔX  K  F ( X (0) ) (2.34)

This set of linear equations need to be solved for the correction vector

 Δx 1 
Δx 
ΔX   2 
  
 
Δx
 n

In eqn.(2.34) F' ( X(0) ) is called the JACOBIAN MATRIX and the vector
K  F ( X (0) ) is called the ERROR VECTOR. The Jacobian matrix is also denoted as J.

Solving eqn. (2.34) for ΔX

ΔX =  ' (0 )
1
F ( X )  K  F ( X (0 ) )
  (2.35)

45
Then the improved estimate is

X ( 1)  X ( 0 )  ΔX

Generalizing this, for ( h  1 ) iteration

th

X ( h1)  X ( h )  ΔX where (2.36)

  
ΔX  F' ( X ( h ) ) 1 K  F ( X ( h ) )  (2.37)

i.e. ΔX is the solution of F' ( X ( h ) ) ΔX  K  F ( X ( h ) ) (2.38)

46
Thus the solution procedure to solve F ( X )  K is as follows :

Set iteration count h = 0; Assume initial solution vector as X(h)

(i) Calculate the error vector K  F ( X (h) )

If the error vector  zero, convergence is reached; otherwise formulate

F' ( X (h) ) ΔX  K  F ( X (h) )

(ii) Solve for the correction vector ΔX

(iii) Update the solution vector as

X (h1)  X (h)  ΔX

Values of the correction vector can also be used to test for convergence.

It is to be noted that Error vector = specified vector – vector of calculated values

47
Example 2.2

Using Newton-Raphson method, solve for x 1 and x 2 of the non-linear equations

4 x 2 sin x 1 = - 0.6; 4 x 22 - 4 x 2 cos x 1 = - 0.3

Choose the initial solution as x 1(0) = 0 rad. and x (0)

2 = 1. Take the precision index on

error vector as 10  3 .

Solution

Knowing the initial solution as x 1 = 0 and x 2 = 1 errors are calculated as

x1  0
- 0.6 – (4 x 2 sin x1 ) x  1 = - 0.6
2

x1  0
- 0.3 – (4 x - 4 x 2 cos x 1 ) x 2  1
2
2 = - 0.3

 0.6 
The error vector   is not small.
  0.3 
48
It is noted that f 1 = 4 x 2 sin x 1 f 2 = 4 x 22 - 4 x 2 cos x 1

 f 1 f 1 
 x x 2  4 x cos x 1 4 sin x 1 
Jacobian matrix is: J =  1  =  2
 f 2 f 2   4 x 2 sin x 1 8 x 2  4 cos x 1 
 x 1 x 2 

Substituting the latest values of state variables x 1 = 0 and x 2 = 1

4 0  0.25 0 
J=   ; Its inverse is J  1 = 
 0 4   0 0.25

Δ x 1  0.25 0   0.6  0.150

Correction vector is calculated as Δ x  =  0   0.3 =  0.075
 2  0.25    

x1  0   0.150  0.150

The state vector is updated as   =   +   =  0.925 
x 2  1   0.075  

This completes the first iteration.

49
Current solution: x 1 = - 0.15 rad. and x 2 = 0.925

Errors are calculated as

x 1   0.15 rad.
- 0.6 – (4 x 2 sin x 1 ) = = - 0.047079
x 2  0.925

x 1   0.15 rad.
- 0.3 – (4 x 22 - 4 x 2 cos x 1 ) x 2  0.925 = - 0.064047

  0.04709 
The error vector   is not small.
  0.064047

4 x cos x 4 sin x 1 
x 1   0.15 rad.
Jacobian matrix is: J =  2 1

 4 x 2 sin x 1 8 x 2  4 cos x 1 
x 2  0.925
 3.658453  0.597753
=  
 0.552921 3.444916 

Correction vector is calculated as

-1
Δ x 1   3.658453  0.597753  0.047079  0.016335
Δ x  =  0.552921 3.444916   0.064047 =  
 2      0.021214
50
The state vector is updated as

x1   0.150  0.016335  0.166335

x  =  0.925  +  0.021214 =  0.903786 
 2      

This completes the second iteration.

Errors are calculated as

x 1   0.166335 rad.
- 0.6 – (4 x 2 sin x 1 ) x  0.903786 = = - 0.001444
2

x 1   0.166335 rad.
- 0.3 – (4 x - 4 x 2 cos x 1 ) x  0.903786
2
2 = - 0.002068
2

Still error vector exceeds the precision index.

51
 4 x cos x 1 4 sin x 1  x 1   0.166335 rad.
Jacobian matrix is: J =  2
 4 x 2 sin x 1 8 x 2  4 cos x 1  x 2  0.903786

 3.565249  0.662276
=  
 0.598556 3.285495 

Correction vector is calculated as

-1
Δ x 1   3.565249  0.662276  0.001444  0.000540
Δ x  =  0.598556 3.285495   0.002068 = 
 
 2      0.000728

The state vector is updated as

x1   0.166335  0.000540  0.166875

x  =  0.903786  +  0.000728 =  0.903058 
 2      

52
The state vector is updated as

x1   0.166335  0.000540  0.166875

x  =  0.903786  +  0.000728 =  0.903058 
 2      

Errors are calculated as

x 1   0.166875 rad.
- 0.6 – (4 x 2 sin x 1 ) x  0.903058 = = - 0.000002
2

x 1   0.166875 rad.
- 0.3 – (4 x 22 - 4 x 2 cos x 1 ) x  0.903058 = - 0.000002
2

Errors are less than 10-3.

53
The final values of x 1 and x 2 are - 0.166875 rad. and 0.903058 respectively.

The results can be checked by substituting the solution in the original equations:

4 x 2 sin x 1 = - 0.6 4 x 22 - 4 x 2 cos x 1 = - 0.3

In this example, we have actually solved our first powe r flow proble m by N.R. method.
This is because the two non-linear equations of this example are the power flow model
of the simple system shown in Fig. 2.5 below.

j 0.25 p.u.
1 2
V1  δ1 V2  δ 2
G

Pd 2  Q d 2  ( 0.6  j 0.3 ) p.u.

j 6.0 p.u.

Fig. 2.5 Two bus system

Here bus 1 is the slack bus with its voltage V1  δ1 = 1.0  0 p.u. Further x 1
0

represents the angle δ 2 and x 2 represents the voltage magnitude V2 at bus 2.

54
Power flow model of Newton Raphson method

The equations describing the performance of the network in the bus admittance
form is given by

I = YV (2.39)

In expanded form these equations are

 I1   Y11 Y12  Y1N   V1 

I  Y Y22  Y 2N  V 
  =  21
2  2 (2.40)
        
     
I N  YN1 YN2  YNN  VN 

Typical element of the bus admittance matrix is

Yi j = Yi j  θ i j = Yi j cos θ i j + j Yi j sin θ i j = G i j + j B i j (2.41)

55
Voltage at a typical bus i is

Vi = Vi  δ (2.42)
i

The current injected into the network at bus i is given by

I i  Yi 1 V1  Yi 2 V2    Yi N Vn
N
(2.43)
  Yi n Vn
n 1

In addition to the linear network equations given by eqn. (2.40), bus powe r
equations should also be satisfied in the power flow problem. These bus power
equations introduce non- linearity into the power flow model. The complex powe r
entering the network at bus i is given by

Pi  j Qi  Vi I*i (2.44)

56
Bus powe r equations can be obtained from the above two equations (2.43) and (2.44) by
eliminating the intermediate variable I i . From eqn. (2.44)

N N
Pi  j Qi  V I i = V i
*
i
*

n 1
Yi n Vn = Vi   δ i 
n 1
Yi n  θ i n Vn  δ n

N
= 
n 1
Vi Vn Yi n θi n + δn - δ i

Separating the real and imaginary parts, we obtain

N
Pi =  Vi Vn Yi n cos ( θ i n + δ n - δ i ) (2.45)
n 1

N
Qi = - 
n 1
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.46)

The real and reactive powe rs obtained from the above two equations are referred as
calculated powe rs. During the power flow calculations, their values depend on the latest
bus voltages. Finally,these calculated powers should be equal to the specified powers.

57
Thus the non-linear equations to be solved in power flow analysis are


n 1
Vi Vn Yi n cos ( θ i n + δ n - δ i ) = PI i (2.47)

N
- 
n 1
Vi Vn Yi n sin ( θ i n + δ n - δ i ) = QI i (2.48)

It is to be noted that equation (2.47) can be written for bus i only if real power injection
at bus i is specified.

Similarly, equation (2.48) can be written for bus i only if reactive power injection at
bus i is specified.

In the power system, let there be

N 1 number of P-Q buses; N 2 number of P-V buses Then N  N1  N 2  1 .

58
Number of P-Q buses = N1 ; Number of P-V buses = N2

Number of SPECIFIED powers PI and QI

We can write N1  N 2 real power specification equations and N 1 reactive power

specification equations.

Thus total number of possible equations = 2 N1  N 2 .

Number of UNKNOWN variables δ and V

There are

i) N1  N 2 number of unknown phase angles δ and

ii) N 1 number of unknown voltage magnitudes V .

Thus total number of unknown variables = 2 N1  N 2

Therefore Number of equations = Number of variables = 2 N1  N 2

59
Thus in power flow study, we need to solve the equations
N


n 1
Vi Vn Yi n cos ( θ i n + δ n - δ i ) = PI i (2.49)

for i = 1, 2, …….., N

i ≠ s
N
and - 
n 1
Vi Vn Yi n sin ( θ i n + δ n - δ i ) = QI i (2.50)

for i = 1, 2, …….., N

i ≠ s

i ≠ P – V buses

for the unknown variables δ i i = 1,2,…….., N, i ≠ s and

Vi i = 1,2,…....., N, i ≠ s , i ≠ P – V buses

The unknown variables are also called as state variables.

60
Example 2.3

In a 9-bus system, bus 1 is the slack bus, buses 2,5 and 7 are the P-V buses. List the state
variables. Also indicate the specified power injections.

Solution

Buses 3,4,6,8 and 9 are P-Q buses.

δ 2 , δ 3 , δ 4 , δ 5 , δ 6 , δ 7 , δ 8 , δ 9 , V3 , V4 , V6 , V8 and V9 are the state variables.

PI 2 , PI 3 , PI 4 , PI 5 , PI 6 , PI 7 , PI 8 , PI 9 , QI 3 , QI 4 , QI 6 , QI 8 and QI 9 are the specified

power injections.

61
Power flow solution by Newton Raphson method

As discussed earlier, taking the bus voltages and line admittances in polar form, in
power flow study we need to solve the non-linear equations
N


n 1
Vi Vn Yi n cos ( θ i n + δ n - δ i ) = PI i (2.51)

N
- 
n 1
Vi Vn Yi n sin ( θ i n + δ n - δ i ) = QI i (2.52)

Knowing that Yi i = Yi i  θ i j = Yi i cos θ i i + j Yi i sin θ i i = G i i + j B i i

by separating the term with n  i we get

N
Gii + 
2
Vi Vi Vn Yi n cos ( θ i n + δ n - δ i ) = PI i (2.53)
n 1
ni

N
 Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) = QI i (2.54)
n 1
ni

62
In a compact form, the above non-linear equations can be written as

P ( δ, V )  PI (2.55)

Q ( δ, V )  QI (2.56)

On linearization, we get

 P P 
 δ  V   Δδ   ΔP 
   
Q  Δ V  ΔQ (2.57)
 Q  
 δ  V


where

ΔP  PI  computed value of P ( δ, V ) corresponding to the present solution.

ΔQ  QI  computed value of Q ( δ, V ) corresponding to the present solution.

ΔV
To bring certain symmetry in the elements of the coefficient matrix, is
V
taken as problem variable in place of Δ V .

63
Then eqn. (2.57) changes to

 P P 
 V   Δδ 
V
 δ     ΔP 
  Δ V     (2.58)
 Q     
Q ΔQ 
 V   V 

 δ V 

In symbolic form, the above equation can be written as

H N  Δδ   ΔP 
     
       (2.59)
  Δ V   
M L   V  ΔQ
 

The matrix H N is known as JACOBIAN matrix.

M L

It is to be noted that Δδ calculated using eq. (2.59) will be in RADIAN.

64
 P P 
 V H N
V
 δ   
    
 Q Q   
 V M L 
 δ V 

The dimensions of the sub-matrices will be as follows:

H (N 1  N 2 ) x (N 1  N 2 )

N (N 1  N 2 ) x N 1

M N 1 x (N 1  N 2 ) and

L N1 x N1

where N 1 is the number of P-Q buses and N 2 is the number of P-V buses.

65
Consider a 4 bus system having bus 1 as slack bus, buses 2 and 3 as P-Q buses
and bus 4 as P-V bus.

Real power injections PI 2 , PI 3 & PI 4 and reactive power injections QI 2 & QI 3 are

specified.

Noting that δ 2 , δ 3 , δ 4 , V2 and V3 are the problem variables, linear equations

that are to be solved in each iteration will be

66
 Bus 1 is slack bus. Buses 2 and 3 are P-Q buses and bus 4 is a P-V bus

δ2 δ3 δ4 V2 V3

P2 P2 P2 P2 P2

P2 V2 V3 Δδ 2 ΔP2
δ 2 δ 3 δ 4  V2  V3

P3 P3 P3 P3 P3

P3 V2 V3 Δδ 3 ΔP3
δ 2 δ 3 δ 4  V2  V3

P4 P4 P4 P4 P4

P4 V2 V3 Δδ 4 = ΔP4 (2.60)
δ 2 δ 3 δ 4  V2  V3

δQ 2 Q 2 Q 2 Q 2 Q 2 Δ V2
Q2 V2 V3 ΔQ 2
δ 2 δ 3 δ 4  V2  V3 V2

δQ 3 Q 3 Q 3 Q 3 Q 3 Δ V3
Q3 V2 V3 ΔQ3
δ 2 δ 3 δ 4  V2  V3 V3

67
The following is the solution procedure for N.R. method of power flow analysis.

1 Read the line data and bus data; construct the bus admittance matrix.

2 Set iteration count k = 0. Assume a starting solution. Usually a FLAT START

is assumed in which all the unknown phase angles are taken as zero and
unknown voltage magnitudes are taken as 1.0 p.u.

3 Compute the mis match powe rs i.e. the error vector. If the elements of error
vector are less than the specified tolerance, the problem is solved and hence go to
Step 7; otherwise proceed to Step 4.

4 Compute the elements of sub-matrices H, N, M and L . Solve

H N  Δδ   ΔP   Δδ 
       
       for  
  Δ V    Δ V 
M L  k  V  ΔQ  V 
  k  

68
4 Compute the elements of sub-matrices H, N, M and L . Solve

H N  Δδ   ΔP   Δδ 
       
       for  
  Δ V    Δ V 
M L  k  V  ΔQ  V 
  k  

5 Update the solution as

δ δ Δδ
= +
V V k
ΔV
k+1

6 Set k = k + 1 and go to Step 3.

7 Calculate line flows, transmission loss and slack bus power. Print the

results and STOP.

69
Calculation of elements of Jacobian matrix

 H N
Jacobian matrix is   where
 M L 

P P Q Q
H ; N V; M and L V
δ V δ V

We know that as given in eqns. (2.45) and (2.46)

N
Pi =  Vi Vn Yi n cos ( θ i n + δ n - δ i )
n 1

N
Qi = - 
n 1
Vi Vn Yi n sin ( θ i n + δ n - δ i )

Separating the diagonal terms, as seen in eqns. (2.53) and (2.54)

N
Gii + 
2
Pi = Vi Vi Vn Yi n cos ( θ i n + δ n - δ i ) (2.61)
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.62)
n 1
ni
70
 N
Gii + 
2
Pi = Vi Vi Vn Yi n cos ( θ i n + δ n - δ i ) (2.61)
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.62)
n 1
ni

Diagonal elements of Jacobian matrix:

P N
Hii  i =  Vi Vn Yi n sin ( θ i n + δ n - δ i ) =  Q i  Vi
2
Bii (2.63)
δ i n 1
ni

Pi N
Ni i  Vi  2 Vi G i i + 
2
Vi Vn Yi n cos ( θ i n + δ n - δ i )
 Vi n 1
ni

2
= Pi + Vi Gii (2.64)

71
Again using
N
Pi = Vi G i i +  Vi Vn Yi n cos ( θ i n + δ n - δ i )
2
(2.61)
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.62)
n 1
ni

Q i N
Mii  =  Vi Vn Yi n cos ( θ i n + δ n - δ i ) = Pi - Vi
2
Gii (2.65)
δ i n  1
ni

Q i N
Lii  Vi =  2 Vi B i i - 
2
Vi Vn Yi n sin ( θ i n + δ n - δ i )
 Vi n 1
ni

= Q i  Vi
2
Bii (2.66)

72
Off-diagonal elements of Jacobian matrix:

We know that
N
Gii + 
2
Pi = Vi Vi Vn Yi n cos ( θ i n + δ n - δ i ) (2.61)
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.62)
n 1
ni

Pi
Hi j    Vi Vj Yi j sin ( θ i j + δ j - δ i ) (2.67)
δ j

Pi
Ni j  Vj = Vi Vj Yi j cos ( θ i j + δ j - δ i ) (2.68)
 Vj

73
We know that
N
Gii + 
2
Pi = Vi Vi Vn Yi n cos ( θ i n + δ n - δ i ) (2.61)
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i ) (2.62)
n 1
ni

Q i
Mi j  =  Vi Vj Yi j cos ( θ i j + δ j - δ i ) (2.69)
δ j

Q i
Li j  Vj =  Vi Vj Yi j sin ( θ i j + δ j - δ i ) (2.70)
 Vj

74
Summary of formulae: As shown in (2.63) to (2.70)

H i i   Q i  Vi
2
Bii

2
Nii  Pi + Vi Gii

Pi  Vi
2
Mii  Gii

Q i  Vi
2
Lii  Bii
(2.71)
Hi j   Vi Vj Yi j sin ( θ i j + δ j - δ i )

Ni j  Vi Vj Yi j cos ( θ i j + δ j - δ i )

Mij   Vi Vj Yi j cos ( θ i j + δ j - δ i )

Li j   Vi Vj Yi j sin ( θ i j + δ j - δ i )

Flow chart for N. R. method of power flow solution is shown below.

75
 START

READ LINE and BUS DATA

COMPUTE Y MATRIX
ASSUME FLAT START
SET k = 0

FOR ALL P-V BUSES COMPUTE Qi

IF Qi VIOLATES THE LIMITS SET
Qi = Q i Limit AND TREAT BUS i AS A P-Q BUS

COMPUTE MISMATCH POWERS ΔP & ΔQ

NO ELEMENTS OF YES
ΔP & ΔQ < ε ?

COMPUTE ELEMENTS OF MATRICES H, N, M & L and

 Δδ   Δδ 
 H N   Δ V   ΔP   Δ V  AND UPDATE
FORM      ; SOLVE FOR
M L   V  ΔQ   
   V 
δ  δ   Δδ 
 V =  V + Δ V 
  k+1   k   k

SET k = k + 1

COMPUTE LINE FLOWS, TRANSMISSION LOSS & SLACK BUS POWER

PRINT THE RESULTS

STOP

76
Example 2.4

Perform power flow analysis for the power system with the data given below,
using Newton-Raphson method, and obtain the bus voltages.

Line data (p.u. quantities)

Line No. Between buses Line impedances

1 1 2 0  j 0.1
2 23 0  j 0.2
3 1 3 0  j 0.2

Bus data ( p.u. quantities )

Bus Generator Load

Type V δ Q min Q max
No P Q P Q

1 Slack --- --- 0 0 1.0 0 --- ---

2 P - V 1.8184 --- 0 --- 1.1 --- 0 3.5

3 P-Q 0 0 1.2517 1.2574 --- --- --- ---

77
Solution

The bus admittance matrix can be obtained as

1 2 3

1  j15 j10 j5 
  j15 j5 
Y = 2  j10
3  j5 j5  j10 

This gives

1 2 3 1 2 3

1 0 0 0  1  15 10 5 
G 2 0 0 0  and B= 2  10  15 5 
  
3 0 0 0 3  5 5  10

1 2 3 1 2 3

1 15 10 5 1  90 0 90 0 90 0 
 
Y  2 10 15 5  and θ = 2  90 0  90 0 90 0 

3  5 5 10 3  90 0 90 0  90 0 
78
Bus 1 is slack bus; bus 2 is a P – V bus and bus 3 is a P – Q bus.

PI 2  1.8184  δ2 
 
In this problem PI 3   1.2517 and unknown quantities =  δ 3 
QI 3   1.2574  V3 

With flat start V1  1.00 0

V2  1.10 0

V3  1.00 0

We know that
N
Gii + 
2
Pi = Vi Vi Vn Yi n cos ( θ i n + δ n - δ i )
n 1
ni

N
Q i =  Vi B i i  
2
Vi Vn Yi n sin ( θ i n + δ n - δ i )
n 1
ni

79
 1 2 3 1 2 3

1 15 10 5 1  90 0 90 0 90 0 
2 10 15
 
Y   5  and θ = 2  90 0  90 0 90 0 
3  5 5 10 3  90 0 90 0  90 0 

Substituting the values of bus admittance parameters, (whose values will remain same
in all iterations) expressions for P2 , P3 and Q 3 are obtained as follows

2
P2 = V2 G 2 2  V2 V1 Y2 1 cos ( θ 2 1  δ 1  δ 2 )  V2 V3 Y2 3 cos ( θ 2 3  δ 3  δ 2 )

= 0 + 10 V2 V1 cos ( 90  δ1  δ 2 )  5 V2 V3 cos ( 90  δ 3  δ 2 )

=  10 V2 V1 sin ( δ1  δ 2 )  5 V2 V3 sin ( δ 3  δ 2 )

Similarly

P3 =  5 V3 V1 sin ( δ 1  δ 3 )  5 V3 V2 sin ( δ 2  δ 3 )

Likewise

Q 3   V3 B 3 3  V3 V1 Y3 1 sin ( 90  δ 1  δ 3 ) 
2
V3 V2 Y3 2 sin (90  δ 2  δ 3 )

 5 V3 V1 cos ( δ1  δ 3 )  5 V3 V2 cos ( δ 2  δ 3 )
2
= 10 V3

80
To check whether bus 2 will remain as P-V bus, Q 2 need to be calculated.

Q 2  15 V2  10 V2 V1 cos ( δ1  δ 2 )  5
2
V2 V3 cos ( δ 3  δ 2 )

Knowing V1  1.00 0 V2  1.10 0 V3  1.00 0

Q2 = ( 15 x 1.1 x 1.1 ) – ( 10 x 1.1 x 1 x 1 ) – ( 5 x 1.1 x 1 x 1 ) = 1.65

This lies within the Q limits (0 – 3.5). Thus bus 2 remains as P – V bus.

Since δ1  δ 2  δ 3  0, we get P2  P3  0

 5 V3 V1 cos ( δ1  δ 3 )  5 V3 V2 cos ( δ 2  δ 3 )
2
Q 3 = 10 V3

= (10 x 1 x 1 ) – ( 5 x 1 x 1 ) – ( 5 x 1 x 1.1) = - 0.5

Mismatch powers are: ΔP2  PI 2  P2  1.8184  0  1.8184

ΔP3  PI 3  P3   1.2517  0   1.2517

ΔQ 3  QI 3  Q 3   1.2574  0.5   0.7574

δ2 δ3 V3
Linear equations P2 H 22 H 23 N 23 Δδ 2 ΔP2
H 32 H 33 N 33 Δδ 3 = ΔP3
to be solved are P3
Δ V3 ΔQ 3
Q3 M 32 M 33 L 33 V3
81
For this problem, since G i i are zero and θ i j are 90 0

H i i   Q i  Vi B i i H i j   Vi Vj Yi j cos ( δ j  δ i )
2

N i i  Pi ; N i j   Vi Vj Yi j sin ( δ j  δ i )
M i i  Pi
Mi j  Vi Vj Yi j sin ( δ j  δ i )
L i i  Q i  Vi B i i
2

We know P2  P3  0 ; Q 3   0.5 ; V2 = 1.1; V3 = 1.0; B 22   15 ; B 33   10

H 22   Q 2  V2 B 22  - 1.65 + (1.1 x 1.1 x 15 ) = 16.5

2

H 33   Q 3  V3 B 33  0.5  10  10.5
2

N 33  P3  0 ; M 33  P3  0

L 33  Q 3  V3 B 33   0.5  10  9.5
2

H23   V2 V3 Y2 3 cos ( δ 3  δ 2 ) =  1.1 x 1 x 5 x 1   5.5 and H 3 2 =  5.5

N 2 3   V2 V3 Y2 3 sin ( δ 3  δ 2 )  0

M 23  V3 V2 Y3 2 sin ( δ 2  δ 3 )  0

82
  16.5  5.5 0  Δδ 2  1.8184 
Thus  5.5 10.5 0  Δδ 3 =   1.2517 
   
 0 0 9.5 Δ V3  0.7574
V3

Δδ 2  0.08538 
Solving the above Δδ 3 =  0.07449
 
Δ V3   0.0797 
V3

Therefore

δ (21 )  0  0.08538  0.08538 rad.  4.89 0

δ (31 )  0  0.07449   0.07449 rad.   4.27 0

 1.0  0.0797  0.9203

(1)
V3

Thus V1  1.0 0 0

V2  1.1 4.89 0

V3  0.9203   4.27 0

This completes the first iteration.

83
Second iteration:

Q 2  (15 x 1.1 x 1.1) - (10 x 1.1 x 1.0 cos 4.89 0 ) - (5 x 1.1 x 0.9203 cos 9.16 0 )

= 2.1929
This is within the limits. Bus 2 remains as P-V bus.

P2 =  10 V2 V1 sin ( δ1  δ 2 )  5 V2 V3 sin ( δ 3  δ 2 )

P3 =  5 V3 V1 sin ( δ 1  δ 3 )  5 V3 V2 sin ( δ 2  δ 3 )

 5 V3 V1 cos ( δ1  δ 3 )  5 V3 V2 cos ( δ 2  δ 3 )
2
Q 3 = 10 V3

V1  1.00 0 V2  1.14.89 0 V3  0.9203 - 4.27 0

P2  (10 x 1.1 x 1.0 sin 4.89 0 ) + ( 5 x 1.1 x 0.9203 sin 9.16 0 ) = 1.7435

P3  - ( 5 x 0.9203 x 1.0 sin 4.27 0 ) – ( 5 x 0.9203 x 1.1 sin 9.16 0 ) = -1.1484

Q 3 = 10 x 0.9203 x 0.9203 - ( 5 x 0.9203 x 1.0 cos 4.27 0 ) - ( 5 x 0.9203 x 1.1 cos 9.16 0 )

= - 1.1163
ΔP2  1.8184 – 1.7435 = 0.0749
Error vector is
ΔP3  - 1.2517 + 1.1484 = - 0.1033 not small
enough
ΔQ 3  - 1.2574 + 1.1163 = -0.1444
84
H 22  - 2.1929 + ( 1.1 x 1.1 x 15 ) = 15.9571

H 33  1.1163 + (0.9203 x 0.9203 x 10) = 9.5858

N 33 = - 1.1484; M 33 = - 1.1484

L 33  - 1.1163 + ( 0.9203 2 x 10 ) = 7.3532

H 23  - 1.1 x 0.9203 x 5 cos 9.16 0 = - 4.9971

H 32 = - 4.9971

N 23  1.1 x 0.9203 x 5 sin 9.16 0 = 0.8058

M 32 = 0.8058

The linear equations are

 
 15.9571  4.9971 0.8058   Δδ   0.0749 
 4.9971 9.5858  1.1484  2
  
   Δδ 3  =  0.1033
 0.8058  1.1484 7.3532   Δ V3   0.1444
 
 V3 
85
  
 Δδ   0.001914 
 2
  0.012388
Its solution is  Δδ 3  =  
 Δ V3   0.021782
 
 V3 

Δ V3  - 0.9203 x 0.02178 = - 0.02

δ (22 )  0.08538  0.001914  0.08729 rad.  5.00 0

δ (32 )   0.07449  0.012388   0.08688 rad.   4.98 0

 0.9232  0.02  0.9032

(2)
V3

Thus at the end of second iteration

V1  1.00 0 V2  1.15.00 0 V3  0.9032  4.98 0

Continuing in this manner the final solution can be obtained as

V1  1.00 0 V2  1.15 0 V3  0.9  5 0

Once we know the final bus voltages, if necessary, line flows, transmission loss
and the slack bus power can be calculated as discussed in Gauss Seidel method.
86
In Newton Raphson power flow method, in each iteration the linear equations

H N  Δδ   ΔP 
     
      
  Δ V   
M L   V  ΔQ
 

are to be obtained and then solved for ∆δ and ∆|V|. When the system size is large
linear algebraic equations can be solved using factorization as explained below.

Solving large size linear algebraic equations

Using factorization method, solve the equation A x = b given by

2.0 0.2 0.4 0.6  x1   4.8 

3.0 6.3 3.0 3.9  x  16.2 
   = 
2 
4.0 7.4 12.6 10.1 x 3  25.5 
     
5.0 8.5 14.2 22.5  x 4  41.0 

87
Solution

Factor matrices of the coefficient matrix A can be obtained as:

2.0 0 0 0  1.0 0.1 0.2 0.3 

3.0 6.0 0 0   0 1.0 0.4 0.5 
L =   and U =  
4.0 7.0 9.0 0  0 0 1.0 0.6 
   
 5.0 8.0 10.0 11.0   0 0 0 1.0 

Since A = L U equation A X = b becomes L U X = b

Knowing L, U and b we need to solve for vector X.

This can be done performing one forward and one backward substitution.

Let us define a vector p as U X = p . Then L p = b

88
Let us solve L p = b i.e.

2.0 0 0 0  p 1   4.8 
3.0 6.0 0 0  p  16.2 
    = 
2 
4.0 7.0 9.0 0  p 3  25.5 
     
5.0 8.0 10.0 11.0  p 4  41.0 

p1 = 4.8 / 2.0 = 2.4

p2 = [ 16.2 – ( 3.0 x 2.4 )] / 6.0 = 1.5

p3 = [ 25.5 – ( 4.0 x 2.4 ) – ( 7.0 x 1.5 ) ] / 9.0 = 0.6

p4 = [ 41.0 – ( 5.0 x 2.4 ) – ( 8.0 x 1.5 ) – ( 10.0 x 0.6 ) ] / 11 = 1.0

89
Now let us solve U X = p i.e.

1.0 0.1 0.2 0.3   x 1  2.4 

 0 1.0 0.4 0.5   x  1.5 
    =  
2

0 0 1.0 0.6   x 3  0.6 

     
0 0 0 1.0  x 4  1.0 

The calculations are to be carried out in backward direction.

x4 = 1.0

x3 = 0.6 – ( 0.6 x 1.0 ) = 0

x2 = 1.5 – ( 0.4 x 0.0 ) – ( 0.5 x 1.0 ) = 1.0

x1 = 2.4 – ( 0.1 x 1.0 ) – ( 0.2 x 0.0 ) – ( 0.3 x 1.0 ) = 2.0

2.0 
1.0 
Thus the solution vector is X =  
 0 
 
1.0 

90
DECOUPLED / FAST DECOUPLED POWER FLOW METHOD

In Newton Raphson method of power flow solution, in each iteration, linear

equations

 P P 
H N  Δδ   ΔP  H N  V
V
         δ 
       where   =  
  Δ V       Q 
 V  Q (2.71) A
M L   ΔQ M 
L  V
   δ V 

are to be framed and solved for the correction vector. When the power
system has N1 number of P-Q buses and N2 number of P-V buses the
size of the Jacobian matrix is 2N1 + N 2 .

When power flow analysis is carried out on large power system, inversion of
large size Jacobian matrix becomes a challenging problem.

Though factored matrix method can be adopted to solve such large size linear
algebraic equations, factorization has to be carried out in each iteration
since the elements of the Jacobian matrix will change in values in each
iteration. This results in enormous amount of calculations in each iteration.

91
Some time in practice, the Jacobian matrix is recalculated once in every
few iterations and this speeds up the overall solution process. The final
solution is obtained, of course, by the allowable power mismatches at
the buses.

When solving large scale power systems, an alternative strategy for

improving computational efficiency and reducing computer storage
requirements, is the FAST DECOUPLED POWER FLOW METHOD, which
makes use of an approximate version of the Newton Raphson procedure.
The principle underlying the decoupled approach is based on a few
approximations which are acceptable in large practical power systems.

As a first step, the following two observations can be made:

92
  P P 
H N  Δδ   ΔP  H N  V
V
         δ 
       where   =  
  Δ V       Q 
 V  Q (2.71) A
M L   ΔQ M L   V
   δ V 

1 Change in voltage phase angle at a bus primarily affects the flow of

real power in the transmission lines and leaves the flow of the reactive
power relatively unchanged.

2 Change in the voltage magnitude at a bus primarily affects the flow

of reactive power in the transmission lines and leaves the flow of the
real power relatively unchanged.

The first observation states essentially that the elements of the Jacobian
sub-matrix H are much larger than the elements of sub-matrix M, which
we now consider to be approximately zero.

The second observation means that the elements of sub-matrix L are

much larger than the elements of sub-matrix N which are also
considered to be approximately zero.

Incorporation of these two approximations in equation (2.71)A yields two

separated systems of equations
93
H Δδ  ΔP (2.72)

Δ V
L  ΔQ (2.73)
V

The above two equations are DECOUPLED in the sense that the voltage
phase angle corrections Δδ are calculated using only real power
mismatches ΔP, while voltage magnitude corrections ΔΙVΙ are calculated
using only ΔQ mismatches.

Hi j   Vi Vj Yi j sin ( θ i j + δ j - δ i )
(2.71)
Li j   Vi Vj Yi j sin ( θ i j + δ j - δ i )

However, as seen by Eq. (2.71), the coefficient matrices H and L are still
interdependent because the elements of matrix H depend on voltage
magnitudes, being solved in eqn. (2.73), whereas the elements of matrix
L depend on voltage phase angles that are computed from eqn. (2.72).
Of course, the two sets of equations could be solved alternately, using in one
set the most recent solution from the other set.
94
H Δδ  ΔP (2.72)

Δ V
L  ΔQ (2.73)
V

The power flow method that uses the decoupled equations (2.72) and
(2.73) is known as DECOUPLED POWER FLOW METHOD. But this
scheme would still require FACTORIZATION of the coefficient matrices at
each iteration.

If the coefficient matrices in Eqs. (2.72) and (2.73) do not change in every
iteration, factorization need to be obtained only once and this will result in
considerable reduction in the calculations. To achieve this we introduce
further simplifications, which are justified by the physics of transmission
line power flow. This leads to FAST DECOUPLED POWER FLOW
METHOD in which the coefficient matrices become constant matrices.
These matrices are factored only once. During different iteration, only
mismatch powers are recalculated and the solution is updated easily.

95
In a well designed and properly operated power transmission system:

1 The differences ( δ p  δ q ) between two physically connected buses of

the power system are usually so small that

cos ( δ p  δ q )  1; and sin ( δ p  δ q )  ( δ p  δ q ) (2.74)

2 The line susceptances Bpq are many times larger than the line

conductances Gpq so that

Gpq sin ( δ p  δ q ) << Bpq cos (δ p  δ q ) (2.75)

3 The reactive power Qp injected into any bus p of the system during
normal operation is much less than the reactive power which would flow
if all lines from that bus were short circuited to reference bus. That is

2
Qp << Vp Bpp (2.76)

The above approximations can be used to simplify the elements of

Jacobian sub-matrices H and L.
96
As seen by eq. (2.71) the diagonal elements of sub-matrices H and L are

Hii = - Qi - ΙViΙ2 Bii Lii = Qi - ΙViΙ2 Bii

They now become

Hii = Lii = - ΙViΙ2 Bii (2.77)

Similarly, using eq. (2.71) the off-diagonal elements of sub-matrices H and L

are:

Hij = - ΙViΙ ΙVjΙ ΙYijΙ sin (θij + δj – δi)

Lij = - ΙViΙ ΙVjΙ ΙYijΙ sin (θij + δj – δi)

Knowing that

ΙYijΙ sin (θij + δj – δi) = ΙYijΙ sin θij cos (δj – δi) + ΙYijΙ cos θij sin (δj – δi)

= Bij cos (δj – δi) + Gij sin (δj – δi)

≈ Bij cos (δj – δi) ≈ Bij

Hij = Lij = - ΙViΙ ΙVjΙ Bij (2.78)

97
 Δ V
H Δδ  ΔP L  ΔQ
V

Hii = Lii = - ΙViΙ2 Bii Hij = Lij = - ΙViΙ ΙVjΙ B ij

For a 4 bus system having bus 1 as slack bus, buses 2 and 3 as P-Q
buses and bus 4 as P-V bus the decoupled equations are:

δ2 δ3 δ4

P2  V2 2 B22  V2 V3 B23  V2 V4 B24 Δδ2 ΔP2

P3  V3 V2 B32  V3 2 B33  V3 V4 B34 Δδ3 = ΔP3 (2.79)

P4  V V B  V4 V3 B43  V4 2 B44 Δδ4 ΔP43

4 2 42

|V2| |V3|
Δ V2
Q2  V2 2 B22  V2 V3 B23 ΔQ2
V2
(2.80)
Δ V3 =
Q3  V3 V2 B32  V3 B33
2
ΔQ 3
V3

98
 δ2 δ3 δ4

P2  V2 2 B22  V2 V3 B23  V2 V4 B24 Δδ2 ΔP2

P3  V V B  V3 2 B33  V3 V4 B34 Δδ3 = ΔP34 (2.79)

3 2 32

P4  V V B  V4 V3 B43  V4 2 B44 Δδ4 ΔP4

4 2 42

We wish to keep the coefficient matrix in Eq. (2.79) independent of voltages. As

a first step, Eq. (2.79) be rearranged as

δ2 δ3 δ4
ΔP2
P2  V2 B22  V3 B23  V4 B24 Δδ2
V2

ΔP3
P3  V2 B32  V3 B33  V4 B34 Δδ3 = (2.81)
V3

ΔP4
P4  V2 B42  V3 B43  V4 B44 Δδ4
V4

As a next step, bus voltage magnitude, V2 , V3 and V4 in the coefficient

matrix are set to 1.0 per unit. Then

99
 δ2 δ3 δ4
ΔP2
P2  B22  B23  B24 Δδ2
V2

ΔP3
P3  B32  B33  B34 Δδ3 = (2.82)
V3

ΔP4
P4  B 42  B 43  B 44 Δδ4
V4

This can be written in a compact form as

ΔP
B' Δδ  (2.83)
V

It is to be noted that Δδ computed using eq. (2.83) will be in RADIAN.

100
Now equation (2.80)

|V2| |V3|
Δ V2
Q2  V2 2
B22  V2 V3 B23 ΔQ2
V2
can be rearranged as
=
Δ V3
Q3  V3 V2 B32  V3 2
B33 ΔQ 3
V3

|V2| |V3|
Q2  V2 B22  V2 B23 Δ V2 ΔQ2
= Thus
Q3  V3 B32  V3 B33 Δ V3 ΔQ 3

|V2| |V3|
ΔQ 2
Q2  B 22  B 23 Δ V2
V2
= (2.84)
ΔQ 3
Q3  B32  B33 Δ V3
V3

Eq. (2.84) can be written in a compact form as

ΔQ
B" Δ V  (2.85)
V
101
Let us see how to get matrices B ’ and B ”

In a large power network, the bus admittance matrix is symmetrical and

sparse. Separating the real and imaginary parts, it can be written as

Y  G  jB

The constant matrix B’ is obtained from matrix B

(1) deleting the row and column corresponding to the slack bus and

(2) changing the sign of all the elements.

The constant matrix B” is obtained from matrix B’ by deleting the rows

and columns corresponding to all the P-V buses.

102
One typical solution strategy FDPF solution is to:

1 Read line and bus data.

2 Compute bus admittance matrix and form B ’ and B” and their factor matrices.

3 Assume flat start.

4 Calculate the mismatches ΔP / V for all buses except slack bus

' ΔP
5 Solve eqns. B Δδ  for Δδ
V

6 Update the angles δ. Use the updated bus voltages to calculate mismatches
ΔQ / V for all P-Q buses

" ΔQ
7 Solve eqns. B Δ V  for Δ V and update the magnitudes V and
V

return to step 4 to repeat the iteration until all mismatches are within
specified tolerances.

103
The Fast Decoupled Power Flow method uses the constant matrices B’

and B” that are factored only once. During different iterations repeat
solution is obtained corresponding to the present mismatch power
ΔP ΔQ
vectors and . Thus tremendous amount of computational
V V

simplifications are achieved in Fast Decoupled Power Flow method and

hence it is ideal for large scale power systems.

104
Example 2.5

Consider the power system described in Example 2.4. Determine the bus
voltages at the end of second iteration, employing Fast Decoupled
Power Flow method.

Solution

Susceptance matrix of the power network is

1 2 3
1  15 10 5
B = 2 10  15 5
3 5 5  10

The constant matrices are

2 3
2 15 5
and B” = 10
B'  3  5 10

V1  1.0 0 0
Initial solution is V2  1.1 0 0
V3  1.0 0 0

105
As in example 2.4, corresponding to the assumed initial solution
ΔP2  1.8184 and ΔP3   1.2517

ΔP2 ΔP3
Therefore  1.6531;   1.2517
V2 V3

ΔP  15  5  Δδ2   1.6531 
Thus B' Δδ  yields  5 10  Δδ  =  1.2517 
V    3  

On solving the above

Δδ2  1 10 5   1.6531   0.08218 

Δδ  =
125  5 15   1.2517 
=  0.08408 
 3  

δ (21)  0  0.08218  0.08218 rad.  4.710

δ (31)  0  0.08408   0.08408 rad.   4.82 0

V1  1.0 0 0
This gives V2  1.1 4.710
V3  1.0   4.82 0

106
  5 V3 V1 cos ( δ1  δ 3 )  5 V3 V2 cos ( δ 2  δ 3 )
2
Q 3  10 V3

V1  1.0 0 0 V2  1.1 4.710 V 3  1.0   4.82 0

Reactive power at bus 3 is calculated as

Q 3  (10 x 1.0 x 1.0) - {5 x 1.0 x 1.0 cos (  4.82 0 )} - {5 x 1.0 x 1.1 cos ( 9.53 0 )}

=  0.4064

ΔQ3  QI 3  Q 3   1.2574  0.4064   0.851

ΔQ
Thus B" Δ V  yields
V

10 Δ V3 = - 0.851 i.e. Δ V3 = - 0.0851

This gives V3 ( 1)
 1.0  0.0851  0.9149

At the end of first iteration, bus voltage

V1  1.0 0 0
V2  1.1 4.710
V3  0.9149   4.82 0
107
V1  1.0 0 0 V2  1.1 4.710 V3  0.9149   4.82 0

Second iteration:

Q 2   {( 10 x 1.1 x 1.0 cos 4.710 )  ( 15 x 1.12 ) + ( 5 x 1.1 x 0.9149 cos 9.53 0 )}

= 2.2246

This is within the limits. Bus 2 remains as P-V bus.

P2  ( 10 x 1.1 x 1.0 sin 4.710 )  0  ( 5 x 1.1 x 0.9149 sin 9.53 0 ) = 1.7363

P3  { 5 x 0.9149 x 1.0 sin (  4.82 0 )} + { 5 x 0.9149 x 1.1 sin (  9.53 0 )}  0

= - 1.2175

ΔP2  1.8184  1.7363  0.0821

ΔP3   1.2517  1.2175   0.0342

ΔP2 ΔP3
 0.07464 ;   0.03738
V2 V3

 15  5  Δδ2   0.07464 
Equation with Δδ2 and Δδ3 as variables are  5 10  Δδ  =  0.03738 
   3  
108
 Δδ2  0.004476
On solving this Δδ  =  
 3   0.0015 

δ (22 )  0.08218  0.004476  0.08666 rad.  4.97 0

δ (32 )   0.08408  0.0015   0.08558 rad.   4.90 0

This gives V1  1.0 0 0 V2  1.1 4.97 0 V3  0.9149   4.90 0

Reactive power at bus 3 is calculated as

Q 3   [{ 5 x 0.9149 x 1.0 cos (  4.90 0 )} +{ 5 x 0.9149 x 1.1 cos (  9.87 0 )}

 ( 10 x 0.9149 2 )] =  1.1448

ΔQ 3
ΔQ 3   1.2574  1.1448   0.1126 ;   0.1231
V3

ΔQ
Thus B Δ V  yields 10 Δ V3 = - 0.1231 Δ V3 = - 0.01231
"
i.e.
V

This gives V3 (2)

 0.9149  0.01231  0.9026

At the end of second iteration, bus voltages are

V1  1.0 0 0 V2  1.1 4.97 0 V3  0.9026   4.90 0 109