Spreadsheets in Education (eJSiE)

Volume 3 | Issue 1

Article 3

August 2008

Spreadsheet Implementations for Solving PowerFlow Problems

Mark A. Lau
Universidad del Turabo, [email protected]

Sastry Kuruganty
Universidad del Turabo, [email protected]

Follow this and additional works at: http://epublications.bond.edu.au/ejsie

Recommended Citation
Lau, Mark A. and Kuruganty, Sastry (2008) Spreadsheet Implementations for Solving Power-Flow Problems, Spreadsheets in Education
(eJSiE): Vol. 3: Iss. 1, Article 3.
Available at: http://epublications.bond.edu.au/ejsie/vol3/iss1/3

This In the Classroom Article is brought to you by the Bond Business School at ePublications@bond. It has been accepted for inclusion in Spreadsheets
in Education (eJSiE) by an authorized administrator of ePublications@bond. For more information, please contact Bond University's Repository
Coordinator.

Spreadsheet Implementations for Solving Power-Flow Problems

Abstract

The solution to the power-flow problem is of fundamental importance in power system analysis and design.
Transient stability studies and fault analysis in power systems demand solutions to a power-flow problem as a
first step in the analysis. Although commercial software such as PSS/E (Power System Simulator for
Engineers) or PowerWorld can be used for power flow calculations, such specialized programs may not be
widely available in many universities or colleges. Spreadsheets, on the other hand, provide an economic
alternative for the implementation of the numerical algorithms encountered in power flows.
Keywords

power flows, Gauss-Seidel method, Newton-Raphson method, power systems

Cover Page Footnote

The authors express their appreciation to the Universidad del Turabo where this work was conducted.

This in the classroom article is available in Spreadsheets in Education (eJSiE): http://epublications.bond.edu.au/ejsie/vol3/iss1/3

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

Spreadsheet Implementations for Solving

Power-Flow Problems
Mark Lau

Department of Electrical and Computer Engineering

Universidad del Turabo
[email protected]

Sastry Kuruganty

Department of Electrical and Computer Engineering

Universidad del Turabo
[email protected]

Abstract
The solution to the power-flow problem is of fundamental importance in power system analysis and design. Transient stability studies and fault analysis in power systems demand solutions to
a power-flow problem as a first step in the analysis. Although commercial software such as PSS/E
(Power System Simulator for Engineers) or PowerWorld can be used for power flow calculations,
such specialized programs may not be widely available in many universities or colleges. Spreadsheets, on the other hand, provide an economic alternative for the implementation of the numerical
algorithms encountered in power flows.

Keywords: power flows, Gauss-Seidel method, Newton-Raphson method, power systems.

1. Introduction
The solution to the power-flow problem is of fundamental importance in power system analysis and design. In transient stability studies and fault analysis, solutions to a
power-flow problem constitute a necessary initial step in such analyses.
The objective of the power-flow problem is to calculate the voltage magnitudes and
phase angles at each bus or node in a given power system. Calculations are performed
under the assumption of balanced three-phase steady-state conditions. In addition to
voltages and angles, real and reactive power flows in equipment such as transformers
and transmission lines can be also obtained from these calculations.
The topics in this paper follow the treatment found in standard reference material on
power system analysis and design [24]. In particular, two numerical methods, namely,
the Gauss-Seidel and the Newton-Raphson methods are used to determine the power
flows in a small-scale power system. The examples are simple enough so that readers
can replicate hand calculations and reproduce the spreadsheet implementations. The
application of spreadsheets for solving power flows and other related problems has been
reported in the literature [1, 5, 6]. The emphasis of this paper is on the educational value
of spreadsheets in the analysis of power systems.
The paper is organized as follows: Section 2 presents the procedure for constructing
the bus admittance matrix of a power network; Section 3 presents the description of the
2008 Bond University. All rights reserved.
http://epublications.bond.edu.au/ejsie/

Published by ePublications@bond, 2008

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

general power-flow problem; Section 4 discusses the solution to the power-flow problem
using the Gauss-Seidel method and presents a spreadsheet implementation of this technique; Section 5 provides the solution to the power-flow problem using the NewtonRaphson method along with its spreadsheet implementation; and Section 6 presents
concluding remarks.
2. The bus admittance matrix of a power network
The bus admittance matrix offers a convenient representation of the topology of a
power network. A power network contains various components, namely, bus bars, generators, transformers, transmission lines, loads, and so on. The network is described by
the bus (or node) voltages and the equivalent admittances of components connected between buses or between buses and ground. In the following, we provide a simple constructive approach to determine the bus admittance matrix of a power network. This
network representation is referred to as the primitive representation. The relation between
the injected node currents and the (line-to-neutral) node voltages is given by
I = YbusV

(1)

where I is the vector of injected node currents, Ybus is the bus admittance matrix, and V
is the vector of node voltages.
Each component element of the interconnected network is referred to as a branch.
The branch is connected between two nodes of the network, or between a node and the
reference node (ground).
Construction of Ybus by inspection

The Ybus matrix is symmetric.

The diagonal element Yii of Ybus is the self-admittance and is equal to the sum of
the primitive admittances of all the components connected to the ith node.
The ijth off-diagonal element Yij of Ybus is equal to the negative of the primitive
admittance of all components connected between nodes i and j.

As an example, we consider the simple power system depicted in Figure 1. The system consists of three buses: a synchronous generator connected to bus 1, a load connected to bus 2, a voltage-regulated generator connected to bus 3, and transmission
lines between buses 1 and 2, 2 and 3, and 1 and 3. The parameters for the network
components are given in Tables 1 and 2. All parameters are expressed in per unit
(p.u.) of a reference base.

28

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

Bus
1

Bus
2

Bus
3

Figure 1: Three-bus power system.

Table 1: Bus input data.

Type
Slack
Load
Constant voltage

Bus
i
1
2
3

Bus voltage
|Vi| (p.u.) i (deg.)
1.0
0

1.0

Complex power (p.u.)

Pgi Qgi Pdi Qdi

0.0 0.0
0.0 0.0 2.0 0.5
1.0
0.0 0.0

Table 2: Transmission line parameters.

Line
bus i to bus k
12
23
13

Impedance
Zik (p.u.)
j0.1
j0.2
j0.4

To determine the Ybus matrix for the given system, we calculate the primitive admittance of each branch in per unit:

1
= j10,
j 0.1
1
y 23 =
= j 5,
j 0.2
1
y13 =
= j 2.5.
j 0.4
y12 =

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

29

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

With the primitive admittances just calculated, we determine the elements of Ybus:

Y11 = y12 + y13 = j10 j 2.5 = j12.5,

Y22 = y12 + y 23 = j10 j 5 = j15,
Y33 = y 23 + y13 = j 5 j 2.5 = j 7.5,
Y12 = Y21 = y12 = j10,

Y23 = Y32 = y 23 = j 5,
Y13 = Y31 = y13 = j 2.5.
Therefore, the bus admittance matrix for the network of Figure 1 is given by

j 2.5
j12.5 j10

Ybus =
j 5 .
j15
j10
j 2.5
j5
j 7.5

(2)

3. The power-flow problem

The power-flow problem discussed in this section will be presented in terms of the
Ybus matrix whose elements are of the form
Yik = |Yik|ejik = |Yik|cosik + j|Yik|sinik = Gik + jBik

(3)

for i, k = 1, 2, . . . , N. Let the voltage at bus i be denoted by

Vi = |Vi|eji = |Vi|(cosi + jsini)

(4)

for i = 1, 2, . . . , N.
As indicated in Figure 2, the net current injected into the network at bus i in terms of
the elements Yik of the Ybus is determined by
N

Ii = Yi1V1 + Yi2V2 + . . . + YiNVN =

Y V
k =1

ik

i = 1, 2, . . . , N.

(5)

Let Pi and Qi denote the net real and reactive power entering the network at bus i.
Then the complex conjugate of the power injected at bus i is
Pi jQi = Vi *

k =1

k =1

YikVk = | YikViVk | ej(ik + k - i) .

(6)

From the preceding equation we obtain the following form of the power-flow equations:
N

Pi = | Yik ViVk | cos(ik + k - i),

(7)

k =1

30

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

Qi = | Yik ViVk | sin(ik + k - i).

(8)

k =1

Now let Pgi denote the scheduled power being generated at bus i and Pdi denote the
scheduled power demand of the load at that bus. Then the net scheduled power being injected into the network at bus i is
Pi,sch = Pgi Pdi.

(9)

We define the mismatch Pi as the difference between the net scheduled power Pi,sch
and the calculated value of Pi as given by Equation (7). That is,
Pi = Pi,sch Pi = (Pgi Pdi) - Pi.

(10)

Likewise, we may define the mismatch Qi for the reactive power as

Qi = Qi,sch Qi = (Qgi Qdi) - Qi.

(11)

When the mismatches Pi and Qi are zero at bus i, we have the power-balance equations
Pi Pi,sch = Pi (Pgi Pdi) = 0,

(12)

Qi Qi,sch = Qi (Qgi Qdi) = 0.

(13)

The power-flow problem entails the computations of Pi and Qi for values of the unknown bus voltages which cause the mismatches Pi and Qi to be equal to zero at
each bus.
At each bus i two of the four quantities i, |Vi|, Pi, and Qi are specified and the remaining two are calculated. For convenience bus 1 is designated as the slack bus and the
voltage angle of the slack bus serves as reference for the angles of all other bus voltages.
The usual practice is to set 1 = 0.
4. Power-flow solution by the Gauss-Seidel method
4.1 The Gauss-Seidel method
In practice power-flow solutions are obtained by iterative methods. We consider the
three-bus system shown in Figure 1 and derive the general equations for the iterations.
With the slack bus designated as number 1, we start computations at bus 2. If P2,sch
and Q2,sch are the scheduled real and reactive power, respectively, entering the network
at bus 2, we have
P2,sch jQ2,sch = V 2 ( Y 21 V 1 + Y 22 V 2 + Y 23 V 3 )
*

and solving for the voltage at bus 2

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

31

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

V2 =

1 P2,sch jQ2,sch
(Y21V1 + Y23V3 ) .

*
Y22
V2

Similarly, at load bus 3 we have

V3 =

1 P3,sch jQ3,sch
(Y31V1 + Y32V2 ) .

*
Y33
V3

The iterations are based on the scheduled real and reactive power at buses 2 and 3,
the scheduled slack bus voltage V1 = |V1|ej1, and initial voltage estimates V2( 0 ) and

V3( 0 ) at the other buses. The corrected voltage V2(1) is calculated from
V2(1) =

1 P2,sch jQ2,sch
(0)

(
+
) .
Y
V
Y
V

21
1
23
3
Y22
V2( 0)*

The corrected voltage V2(1) is then used to calculate the value of V3(1)

V3(1) =

1
Y33

P3,sch jQ3,sch

(Y31V1 + Y32V2(1) ) .

( 0 )*
V3

The procedure is repeated until the amount of correction in voltage at every bus is
less than some predetermined precision index. This process of solving the power-flow
equations is known as the Gauss-Seidel iterative method.
For an N-bus system the general equation for the calculated voltage at any bus I
where P and Q are scheduled is given by

Vi ( m ) =

1 Pi ,sch jQi ,sch i 1

(m)

Y
V
Yik Vk( m 1) .

ik k
( m 1)*
Yii Vi
k =1
k = i +1

(14)

In the following we show the calculations for the first iteration of the Gauss-Seidel
method for the three-bus system of Figure 1. The bus admittance matrix Ybus for this system is given by Equation (2). The input data and unknowns at each bus can be inferred
from Table 1 and are summarized in Table 3.
Table 3: Input data and unknowns for the power system of Figure 1.

Type

Bus
i

Slack

Load

Constant voltage

Input data
|V1| = 1.0 p.u., 1 = 0
Psch,2 = Pg2- Pd2 = -2.0 p.u.,
Qsch,2 = Qg2 - Qd2 = -0.5 p.u.
|V3| = 1.0 p.u.,
Psch,3 = Pg3 - Pd3 = 1.0 p.u.

Unknowns
Psch,1 = Pg1 - Pd1 = Pg1,
Qsch,1 = Qg1 - Qd1 = Qg1
|V2|, 2
3,
Qsch,3 = Qg3 - Qd3 = Qg3

32

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

We assume initial guesses for the voltages V2( 0 ) = 1.0ej0 = 1.0 and V3( 0 ) = 1.0ej0 = 1.0
at buses 2 and 3, respectively. Using Equation (14) we determine an improved value for
V2:

V2(1) =

1 2 + j 0.5

j 7.853o
(
j
10
)(
1
.
0
)
(
j
5
)(
1
.
0
)
=
0
.
9758
e
.

j15 1.0

With V2(1) at hand, we then proceed to find Q3(1) and V3(1) . Since bus 3 is a voltagecontrolled bus, we can either use Equation (6) or (8) to compute the reactive power at
that bus. In the spreadsheet implementation of the Gauss-Seidel method discussed in the
next section, we use Equation (6) to compute both the real and reactive power at the
buses of interest. We observe that the real power is the real part of the right-hand side of
Equation (6), while the reactive power is given by the negative of the imaginary part of
the right-hand side of that equation. Here we show how to use Equation (6) to compute
the reactive power at bus 3:

Q3(1) = Im 1.0e j 0 ( j 2.5)(1.0e j 0 ) + ( j 5)(0.9758e j 7.853 ) + ( j 7.5)(1.0e j 0 )

o

]}

= 0.1667 p.u.
With the preceding value of reactive power, we use Equation (14) once again to calculate an improved value of V3:

V3(1) =

o
o
1 1 j 0.1667
( j 2.5)(1.0) ( j 5)(0.9758e j 7.853 ) = 1.0010e j 2.545 .

1.0
j 7.5

As indicated in Table 3, |V3| = 1 at bus 3 and, therefore, we set V3(1) = 1.0e j 2.545 for next
o

iteration. This concludes the first iteration of the Gauss-Seidel method. The calculations
are repeated in a similar manner with updated values until convergence can be discerned. We next show how to implement the above calculations using a spreadsheet.
4.2 Spreadsheet implementation of the Gauss-Seidel method
We begin by inputting the entries that form the bus admittance matrix given in Equation (2) . Also, we input the known data shown in Table 3. This step is shown in the
screen capture of Figure 2. The actual Microsoft Excel commands used to generate the
input data is presented in Figure 3.

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

33

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

Figure 2: Bus admittance matrix and input data for the power system of Figure 1.

Figure 3: Microsoft Excel commands used to generate the input data.

The calculations of the Gauss-Seidel method are shown in Figure 4 as the cell range
H1:T11 in the spreadsheet. The entries in row 3 extending from cell H3 to T3 correspond
to initial estimates for the numerical algorithm. The basic formulas are found in row 4
and extend from H4 to T4. The formulas in row 4 are copied to the rows below as many
times as needed until convergence is reached. We note that some columns were omitted
in Figure 4 for clarity; these columns simply contain intermediate calculations. The calculations reveal that convergence is reached after 8 iterations for a precision index of 10-5
for the bus voltages and 10-4 for the angles. A final calculations section is also shown in
Figure 4; the section covers the cell range V1:AC3, but columns V and W were left out as
these columns contain intermediate calculations. The solution to the power-flow problem at hand is summarized in Table 4.
To conclude this section, we provide in Table 5 all the formulas used in the implementation of the Gauss-Seidel method. Formulas corresponding to hidden columns are
also included for completeness.

34

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

Figure 4: Gauss-Seidel iterations showing relevant quantities in the power system of Figure 1.
Table 4: Steady-state power flows and bus voltages for the system in Figure 1. Except for the angles, all other quantities are expressed in per unit (p.u.).

Bus i
1
2
3

Input data (given)

|V1| = 1.0, 1 = 0
Psch,2 = -2.0, Qsch,2 = -0.5
|V3| = 1.0, Psch,3 = 1.0

Unknowns (calculated)
Psch,1 = 1.0, Qsch,1 = 0.1935
|V2| 0.9549, 2 = -6.902
3 = 3.383, Qsch,3 = 0.3066

Table 5: Microsoft Excel formulas for implementing the Gauss-Seidel method.

Cell
H3:H11
H2:T2
I3, J3
K3
L3
M3
N3:P3

Formula
List of numbers from 0 to 8.
Labels for various quantities.
Blank cells.
=COMPLEX(1,0,"j")
=IMABS(K3)
=180*IMARGUMENT(K3)/PI()

Blank cells.

Q3

=COMPLEX(1,0,"j")

R3

=COMPLEX(1,0,"j")

S3

=IMABS(R3)

T3

=180*IMARGUMENT(Q3)/PI()

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

Comments
Iteration number.
Gauss-Seidel section.
No calculations.
Initial guess for V2.
Magnitude of initial V2.
Angle of initial V2 in deg.
No calculations.
Initial value for calculated
V3.
Initial value for correctedV3;
magnitude must be constant.
Magnitude of initial corrected V3.
Angle of initial calculated V3
in deg.

In The Classroom

35

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

Table 5 (continued): Microsoft Excel formulas for implementing the Gauss-Seidel method.

Cell
I4

Formula

Comments

=IMDIV(COMPLEX($E$10,-$F$10,"j"),
IMCONJUGATE(K3))

The term

P2,sch jQ2,sch
V2( m 1)*

of

Equation (14).
J4

=IMSUM(IMPRODUCT($B$4,$C$9),
IMPRODUCT($D$4,R3))

K4

=IMDIV(IMSUB(I4,J4),$C$4)

L4
M4

=IMABS(K4)
=180*IMARGUMENT(K4)/PI()

The term Y21V1 + Y23 V3( m 1)

of Equation (14).
Completes the calculation of
V2( m ) in Equation (14) using
intermediate results in cells
I4 and J4.
Magnitude of updated V2.
Angle of updated V2 in deg.
N

N4

O4

=IMSUM(IMPRODUCT($B$5,$C$9),
IMSUM(IMPRODUCT($C$5,K4),
IMPRODUCT($D$5,R3)))

=IMDIV(COMPLEX($E$11,-P4,"j"),
IMCONJUGATE(R3))

The term

Y
k =1

3k

Vk of Equa-

tion (6).
The term

P3,sch jQ3,sch
V3( m 1)*

of

Equation (14).
Q3 computed as the negative
of the imaginary part of
P4

=-IMAGINARY(IMPRODUCT(
IMCONJUGATE(R3),N4))

V3* Y3k Vk as indicated in

k =1

Equation (6).
Updated value V3( m ) calculated from Equation (14).
Observe that in the cell formula Y33V3 was added back
Q4

since this term is not present

explicitly in Equation (14).
The reason is that cell N4
uses Y33V3 when computing

=IMDIV(IMSUM(IMSUB(O4,N4),
IMPRODUCT($D$5,R3)),$D$5)

Q3 according to Equation (6),

but this term has to be excluded when computing V3 .
Corrects V3( m ) to match the
R4

=COMPLEX(S4*COS(IMARGUMENT(
Q4)),S4*SIN(IMARGUMENT(Q4)),"j")

S4

=S3

T4

=180*IMARGUMENT(Q4)/PI()

voltage magnitude at bus 3.

The angle (argument) is the
same as in cell Q4.
Copies the constant voltage
magnitude at bus 3.
Angle of updated V3 in deg.

36

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

10

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

Table 5 (continued): Microsoft Excel formulas for implementing the Gauss-Seidel method.

Cell
I5:T11
V2:AC2

Formula
Copies of cell range I4:T4.
Labels for various quantities.

Comments
Replicates cell formulas in
rows below until convergence is achieved.
Final calculations section.
N

V3

=IMSUM(IMPRODUCT($B$4,$C$9),
IMSUM(IMPRODUCT($C$4,K11),
IMPRODUCT($D$4,R11)))

The term

k =1

2k

Vk of Equa-

tion (6).
N

W3

=IMSUM(IMPRODUCT($B$5,$C$9),
IMSUM(IMPRODUCT($C$5,K11),
IMPRODUCT($D$5,R11)))

The term

=IMREAL(IMPRODUCT(
IMCONJUGATE(K11),V3))

of V2*

Y
k =1

3k

Vk of Equa-

tion (6).
P2 computed as the real part
N

X3

Y
k =1

2k

Vk according to

Equation (6).
Q2 computed as the negative
of the imaginary part of
Y3

=-IMAGINARY(IMPRODUCT(
IMCONJUGATE(K11),V3))

V2* Y2 k Vk according to Eqk =1

uation (6).
P3 computed as the real part
N

Z3

=IMREAL(IMPRODUCT(
IMCONJUGATE(R11),W3))

of V3*

Y
k =1

3k

Vk according to

Equation (6).
Q3 computed as the negative
of the imaginary part of
AA3

=-IMAGINARY(IMPRODUCT(
IMCONJUGATE(R11),W3))

V3* Y3k Vk according to Eqk =1

AB3

=-X3-Z3

AC3

=-Y3-AA3

uation (6).
Real power P1 at slack bus
(balance of real power).
Reactive power Q1 at slack
bus (balance of reactive
power).

5. Power-flow solution by the Newton-Raphson method

5.1 The Newton-Raphson method
To set up the Newton-Raphson numerical method, we employ the power-flow expressions given by Equations (7) and (8). These equations are more convenient than Equation (6) during the computation of the Jacobian matrix, as we will show later.
eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

37

11

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

For simplicity of discussion, we will assume that we have N buses and that all buses,
except the slack bus (i = 1), are load buses with prescribed demands Pdi and Qdi. Denoting the specified values |V1| and 1 for the slack bus, then each of the remaining buses
in the network has the two state variables |Vi| and i to be determined by the powerflow solution. The objective of the Newton-Raphson method is to produce values for
|Vi| and i that will match the prescribed Pdi and Qdi as determined from Equations (7)
and (8).
At each iteration of the method, new estimates of |Vi| and i for the non-slack buses
(i = 2, 3, , N) are generated. At the end of each iteration, the power mismatch is given
by
Pi = Pi,sch - Pi,

(15)

Qi = Qi,sch - Qi.

(16)

To understand the formulation of the method, we consider a three-bus system. Bus 1

is the slack bus, while buses 2 and 3 are load buses. The basic idea is to view the power
mismatches at the load buses as first-order variations resulting from a Taylor series expansion. In this case, the mismatch equations for the real power at the load buses (i = 2,
3) are

Pi =

Pi
P
P
Pi
| V2 | +
| V3 | + i 2 + i 3 .
| V2 |
| V3 |
2
3

(17)

Similarly, the reactive power mismatch for i = 2, 3 are

Qi =

Qi
Q
Q
Qi
| V2 | +
| V3 | + i 2 + i 3 .
| V2 |
| V3 |
2
3

(18)

Equations (17) and (18) can be written in matrix form,

P2
| V
2
P2 P3
P
3 = | V2
Q2 Q2

Q3 | V2
Q3

| V2

|
|
|
|

P2
| V3
P3
| V3
Q2
| V3
Q3
| V3

|
|
|
|

P2
2
P3
2
Q2
2
Q3
2

P2
3

P3 | V2

3 | V3
Q2 2

3 3
Q3

|
|
.

(19)

The 4x4 matrix in Equation (19) is the Jacobian and it will be denoted by J. We can
summarize the numerical algorithm as follows:
1. Provide initial estimates for | Vi | ( 0) and i( 0 ) , i = 2, 3, , N.
2. Use these estimates to calculate Pi ( 0 ) , Qi( 0 ) , Pi ( 0 ) , Qi( 0 ) , and the elements of the Jacobian matrix J.
38

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

12

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

3. Solve for the initial corrections | Vi | ( 0 ) and i( 0 ) using Equation (19). This step requires matrix inversion of J.
4. Obtain new estimates

| Vi | (1) =| Vi | ( 0 ) + | Vi | ( 0) ,

(20)

i(1) = i( 0 ) + i( 0 ) .

(21)

5. Use the new values of | Vi | (1) and i(1) as starting values for iteration 2 and repeat the
above steps. In more general terms, calculate

| Vi | ( m +1) =| Vi | ( m ) + | Vi |( m ) ,

(22)

i( m +1) = i( m ) + i( m )

(23)

until a prespecified precision index is satisfied.

We now apply the procedure outlined above to the power system of Figure 1. Since
bus 3 is a constant voltage bus and not a load bus, we need to modify the preceding
formulation slightly. For this power system, the variables of interest are |V2|, 2, and 3.
Taking into account that |V1| = 1.0 p.u., 1 = 0, |V3| = 1.0 p.u., and the Ybus matrix of
Equation (2), we write out the equations that need to be solved in accordance with Equations (7) and (8):
2 = (10)(|V2|)(1.0)cos(90 + 0 2) + (15)(|V2|)(|V2|)cos(90 + 2 2)
+ (5)(|V2|)(1.0)cos(90 + 3 2),
0.5 = (10)(|V2|)(1.0)sin(90 + 0 2) (15)(|V2|)(|V2|)sin(90 + 2 2)
(5)(|V2|)(1.0)sin(90 + 3 2),
1 = (2.5)(1.0)(1.0)cos(90 + 0 3) + (5)(1.0)(|V2|)cos(90 + 2 3)
+ (7.5)(1.0)(1.0)cos(90 + 3 3).
Simplifying the right-hand sides of the preceding expressions yields

P2 = 10 | V2 | sin(2) 5 | V2 | sin(3 2),

Q2 = 10 | V2 | cos(2) + 15 | V2 | 2 5 | V2 | cos(3 2),

P3 = 2.5sin(3) 5 | V2 | sin(2 3) .
Thus, the elements of the Jacobian matrix are given by
eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

39

13

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

P2
= 10 sin( 2 ) 5 sin( 3 2 ) ,
| V2 |
P2
= 10 | V2 | cos( 2 ) + 5 | V2 | cos( 3 2 ) ,
2
P2
= 5 | V2 | cos( 3 2 ) ,
3
Q2
= 10 cos( 2 ) + 30 | V2 | 5 cos( 3 2 ) ,
| V2 |
Q2
= 10 | V2 | sin( 2 ) 5 | V2 | sin( 3 2 ) ,
2
Q2
= 5 | V2 | sin( 3 2 ) ,
3
P3
= 5 sin( 2 3 ) ,
| V2 |
P3
= 5 | V2 | cos( 2 3 ) ,
2
P3
= 2.5 cos( 3 ) + 5 | V2 | cos( 2 3 ) .
3
The above partial derivatives form the Jacobian matrix

P2

| V2 |
P3
J =
| V |
Q 2
2

| V2 |

P2
2
P3
2
Q2
2

P2

3
P3
.
3
Q2
3

For the initial estimates | V2 | = 1.0 p.u., 2 = 0 o , and 3 = 0 o , the initial Jacobian
matrix becomes

40

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

14

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

J(0)

AND

S. KURUGANTY

0 15 5
= 0 5 7.5 .
15 0
0

The initial power mismatches are

P2 = P2 + 2.0 = 0 + 2.0 = 2.0 ,

Q2 = Q2 + 0.5 = 0 + 0.5 = 0.5 ,

P3 = P3 1.0 = 0 1.0 = 1.0 .

Consequently, the corrected values after the first iteration are
1

| V2 | (1) 1.0 0 15 5 2.0 0.9667

(1)

2 = 0 0 5 7.5 1.0 = 0.1143 .
3(1) 0 15 0
0 0.5 0.0571

Although we assumed that 2 = 0 o , and 3 = 0 o for simplicity, these angles must be

converted to radians before performing the necessary matrix operations. The procedure
is repeated until the variables of interest satisfy a prescribed precision index. Once this is
achieved, we use the values of |V2|, 2, and 3 given by the algorithm to compute Q3
from Equation (8). This completes the solution process.

5.2Spreadsheet implementation of the Newton-Raphson method

In this section we present a spreadsheet implementation of the Newton-Raphson
method. Again, we take the power system of Figure 1. The input data section is identical
to that of the Gauss-Seidel method discussed in Section 4.2 (see Figures 2 and 3).
The power-flow solution is shown in Figure 5. The calculations reveal that acceptable solutions are attained after 3 iterations for a precision index of 10-5. As expected, the
solutions agree with those obtained by the Gauss-Seidel method (see Table 4). In general, the Newton-Raphson method converges to the solution faster than the Gauss-Seidel
method. In some instances, ill-conditioned problems lead to divergence by either method.
As can be seen in Figure 5, the Newton-Raphson method covers the cell range
H1:T15. A complete list of Microsoft Excel formulas is given in Table 6.

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

41

15

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

Figure 5: Spreadsheet implementation of the Newton-Raphson method for determining the power flows in the network of Figure 1.
Table 6: Microsoft Excel formulas for the Newton-Raphson method.

Cell

Formula

H3:H14

List of numbers from 0 to 3.

H2:T2

Labels for various quantities.

Labels to highlight angles in
radians and degrees.
Labels for variables of interest.
Numerical values corresponding to variables of interest.
Numerical values corresponding to variables of interest.
Jacobian matrices.
Inverse of Jacobian matrices.
Labels for power mismatches.
Numerical values corresponding to power mismatches.

I15:K15
I3:I14
J3:J14

K3:K14
L3:N14
O3:Q14
R3:R14
S3:S14

Comments
Iteration number. Some cells
are blank.
Newton-Raphson section.
No calculations.
|V2|, 2, 3.
2 and 3 in radians.

2 and 3 in degrees.
All matrices are 3x3.
All matrices are 3x3.
P2, P3, Q2.
P2, P3, Q2.

Correction terms.

| V2 | , 2 , 3 .

Initial estimates for variables

of interest.

| V2 | ( 0 ) , 2

K3

=J3

Copies the value of | V2 | ( 0 ) .

K4

=J4*180/PI()

Converts 2

(0)

to degrees.

=J5*180/PI()

Converts 3

(0)

to degrees.

T3:T14
J3:J5

K5

(0)

, 3

(0)

42

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

16

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

Table 6 (continued): Microsoft Excel formulas for the Newton-Raphson method.

Cell

Formula

Comments

L3

=10*SIN(J4)-5*SIN(J5-J4)

Jacobian element

P2
.
| V2 |

M3

=10*J3*COS(J4)+5*J3*COS(J5-J4)

Jacobian element

P2
.
2

N3

=-5*J3*COS(J5-J4)

Jacobian element

P2
.
3

L4

=-5*SIN(J4-J5)

Jacobian element

P3
.
| V2 |

M4

=-5*J3*COS(J4-J5)

Jacobian element

P3
.
2

N4

=2.5*COS(J5)+5*J3*COS(J4-J5)

Jacobian element

P3
.
3

L5

=-10*COS(J4)+30*J3-5*COS(J5-J4)

Jacobian element

Q2
.
| V2 |

M5

=10*J3*SIN(J4)-5*J3*SIN(J5-J4)

Jacobian element

Q2
.
2

N5

=5*J3*SIN(J5-J4)

Jacobian element

Q2
.
3

S3

=10*J3*SIN(J4)-5*J3*SIN(J5-J4)-$E$10

Inverse of Jacobian matrix in

cell range L3:N5.
Power mismatch P2 .

S4

=2.5*SIN(J5)-5*J3*SIN(J4-J5)-$E$11

Power mismatch P3 .

S5

=-10*J3*COS(J4)+15*J3^25*J3*COS(J5-J4)-$F$10

Power mismatch Q2 .

O3:Q5

=MINVERSE(L3:N5)

Computes correction terms

| V2 | ( 0 ) , 2
T3:T5

( 0)

, 3

=J3-T3

New estimate | V2 | (1) .

J7

=J4-T4

New estimate 2(1) .

J8

=J5-T5

New estimate 3(1) .

K6:T8

Copies of the formulas in cell

range K3:T5.

J9:T14

Copies of the formulas in cell

range J6:T8.

Published by ePublications@bond, 2008

Multiplies the inverse of the

Jacobian in O3:Q5 by the
vector of power mismatch in
S3:S5.

=MMULT(O3:Q5,S3:S5)

J6

eJSiE 3(1) 2745

( 0)

Completes calculations for

iteration 1.
Replicate formulas as
needed until solutions are
within some precision index.

In The Classroom

43

17

Spreadsheets in Education (eJSiE), Vol. 3, Iss. 1 [2008], Art. 3

SPREADSHEET IMPLEMENTATIONS FOR SOLVING POWER-FLOW PROBLEMS

6. Classroom experience and pedagogical insights

The Gauss-Seidel and the Newton-Raphson methods are among the most popular
numerical techniques encountered in power system courses. The authors introduce these
techniques to their undergraduate students when discussing the power-flow problem.
To reinforce conceptual understanding of these methods, the students are asked to analyze the power flows in a small three-bus system, such as the one given in Figure 1. To
facilitate computer implementation, the authors make the spreadsheets described in this
paper available to their students. This approach has pedagogical advantages since students are relieved from the burden of learning new software. By simply modifying the
necessary spreadsheet cells, the students can easily determine solutions to their assigned
problem.
From the authors experience throughout the years, students express positive attitudes toward the spreadsheet implementation of their power-flow project. Ease of implementation; widespread availability of spreadsheets; convenient tracking and displaying of numerical results; transparency of results that are often obscured by specialized
power system analysis programs, are among the most cited comments.
The authors recommend the use of the spreadsheet approach to solve only small
power systems, perhaps up to three buses. Large systems are more suitably analyzed using specialized software such as PSS/E or PowerWorld. The spreadsheet approach is effective in other respects such as gaining insight into the numerical techniques, making
sense of the convergence or divergence of computer-generated solutions, and developing intuition about well- or ill-conditioned systems.
7.

Conclusions

In this paper we have presented spreadsheet implementations of two widely used

methods for solving power-flow problems. The Gauss-Seidel and the Newton-Raphson
numerical methods are introduced to students in power system analysis courses. Although the mathematical underpinnings are found in courses such as numerical analysis, power systems provide a suitable real-world application upon which constructivist
activities can be designed by instructors. Spreadsheets provide students with an easily
accessible tool with which mathematical models of real systems can be built and analyzed. Furthermore, spreadsheets lend themselves to answering what if questions
when quantities such as real power, reactive power, or bus voltage change to new quiescent operating conditions.
References
[1]

Acarnley, P. A power system load flow analysis using an Excel workbook. Online
available: http://www.engsc.ac.uk/downloads/awards/load.pdf. Accessed: August
13, 2008.

[2]

Bergen, A. R. and Vittal, V. (1999). Power systems analysis, 2nd edition. Prentice Hall.

[3]

Glover, J. D. and Sarma, M. S. (2001). Power system analysis and design, 3rd edition.
Thomson-Engineering.

[4]

Grainger, J. and Stevenson, W. D. (1994). Power systems analysis. McGraw-Hill Publishing Co.

[5]

Rao, N. D. and Haddad, N. Y. (1990). Typical applications of new generation spread44

http://epublications.bond.edu.au/ejsie/vol3/iss1/3

18

Lau and Kuruganty: Spreadsheet Implementations for Solving Power-Flow Problems

M. LAU

AND

S. KURUGANTY

sheets to power system problems, Proceedings of the 33rd Midwest Symposium on Circuits
and Systems, Vol. 1, 159162, August 12-14, 1990.
[6]

Zeljkovi, C. and Gaanovi, M. An example of using Microsoft Excel Solver for power
network calculations. Online available: http://phd.etfbl.net/files/Works_PDF/ Zeljkovic%20Cedomir.pdf. Accessed: August 13, 2008.

eJSiE 3(1) 2745

Published by ePublications@bond, 2008

In The Classroom

45

19