MPPT

J.
Electromagnetic Analysis & Applications, 2009

Published Online September 2009 in SciRes (www.SciRP.org/journal/jemaa)
Copyright 2009 SciRes JEMAA
CONTENTS
Volume 1 Number 3 September 2009

Current Distortion Evaluation in Traction 4Q Constant Switching Frequency Converters
M. Brenna, F. Foiadelli, M. Roscia, D. Zaninelli129

Rotating Capacitor and a Transient Electric Network
H. Sarafian, N. Sarafian138
Delta Modulation With PI ControllerA Comparative Study
A. I. Maswood, S. Anjum145
Energy Comparison of Seven MPPT Techniques for PV Systems
A. Dolara, R. Faranda, S. Leva152
Optimal Power System Restoration and Reconfiguration in Distribution Circuit Using BFAM
and BPSO
K S. Kumar, T. Jayabarathi163
Islanding Detection Method for Multi-Inverter Distributed Generation
A. Cardenas, K. Agbossou, M. L. Doumbia170
Effect of Warm Ionized Plasma Medium on Radiation Properties of Mismatched Microstrip
Termination
A. Al-Sawalha181
Fault Diagnosis Based on ANN for Turn-to-Turn Short Circuit of Synchronous Generator
Rotor Windings
H. Z. Ma, L. Pu187
An Approach to Harmonic State Estimation of Power System
N. C. Zhou, L. Lin, J. Z. Zhu192

Journal of Electromagnetic Analysis and Applications (JEMAA)
Journal Information

SUBSCRIPTIONS
The Journal of Electromagnetic Analysis and Applications (Online at Scientific Research Publishing, www.SciRP.org)
is published quarterly by Scientific Research Publishing, Inc., USA.
E-mail: jemaa@scirp.org

Subscr iption r ates: Volume 1 2009
Print: $50 per copy.
Electronic: free, available on www.SciRP.org.
To subscribe, please contact Journals Subscriptions Department, E-mail: jemaa@scirp.org
Sample copies: If you are interested in subscribing, you may obtain a free sample copy by contacting Scientific
Research Publishing, Inc at the above address.

SERVICES
Adver tisements
Advertisement Sales Department, E-mail: jemaa@scirp.org
Repr ints (minimum quantity 100 copies)
Reprints Co-ordinator, Scientific Research Publishing, Inc., USA.

COPYRIGHT
Copyright2009 Scientific Research Publishing, Inc.
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in
any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as
described below, without the permission in writing of the Publisher.
Copying of articles is not permitted except for personal and internal use, to the extent permitted by national
copyright law, or under the terms of a license issued by the national Reproduction Rights Organization.
Requests for permission for other kinds of copying, such as copying for general distribution, for advertising or
promotional purposes, for creating new collective works or for resale, and other enquiries should be addressed to
the Publisher.
Statements and opinions expressed in the articles and communications are those of the individual contributors and
not the statements and opinion of Scientific Research Publishing, Inc. We assume no responsibility or liability for
any damage or injury to persons or property arising out of the use of any materials, instructions, methods or ideas
contained herein. We expressly disclaim any implied warranties of merchantability or fitness for a particular
purpose. If expert assistance is required, the services of a competent professional person should be sought.

PRODUCTION INFORMATION
For manuscripts that have been accepted for publication, please contact:
J. Electromagnetic Analysis & Applications, 2009, 3: 129-137
doi:10.4236/jemaa.2009.13021 Published Online September 2009 (www.SciRP.org/journal/jemaa)
129
Cur r ent Distor tion Evaluation in Tr action 4Q
Constant Switching Fr equency Conver ter s
M. BRENNA
1
, F. FOIADELLI
1
, M. ROSCIA
2
, D. ZANINELLI
1

1
The Politecnico di MilanoDepartment of Energy, Milano, Italy;
2
Department of Tecnologie e Progettazione, Universit di Bergamo,
Bergamo, Italy.
Email: {morris.brenna, federica.foiadelli, dario.zaninelli}@polimi.it, cristina.roscia@unibg.it

Received April 14
th
, 2009; revised July 13
th
, accepted August 2
nd
, 2009.
ABSTRACT
The paper deals with the power quality analysis of interlaced four quadrant (4Q) converters with constant switching
frequency. These are in fact the input stages of the locomotives and high speed trains supplied by 25 kV, 50 Hz and 15
kV, 16.7 Hz lines. Due to the high power needed for the trains circulation, the 4Q converter can absorb distorted cur-
rents, whose harmonic content can affect the signaling systems and communication devices. The presence of more con-
verters gives the opportunity, using dedicated commutation strategy, to interlace them in order to reduce the harmonic
content in the absorbed current. In the paper a suitable model of more 4Q converters is developed. The control logic
implemented in the model allows the evaluation of the harmonic contribution of both single converter and the interlaced
configuration. The analysis is carried out through electromagnetic transient simulations.
Keywords: Four Quadrant (4Q) Converter, Interlacing, Traction Systems, Power Quality Analysis, Commutation
Strategy
1. Intr oduction
The four quadrant converter (4Q) is the actual best choice
to supply the DC voltage link from AC power contact
line. A typology of distributed power high speed electric
trains having as input stages more 4Q converters [1] has
been considered.
The presence of more converters is necessary to guar-
antee a good redundancy in case of failure and gives the
opportunity, using dedicated control logics, to interlace
them in order to reduce the harmonic content of the ab-
sorbed current. The high power requested by the train for
its acceleration in starting phase and for the auxiliary
services needs high power converters that do not allow
high switching frequencies. Consequently the absorbed
current presents a high ripple value characterized by high
harmonic current components that cannot be tolerated by
the system. Indeed the track circuit used for signaling and
communication for the traffic management and safety
employs signal currents overlapped with the power ones.
These currents can have low frequencies (50 Hz and 178
Hz) in the traditional signaling system, or they are in the
audio frequencies range for the new European
ERTMS/ETCS one. Therefore, the harmonics produced
by the 4Q converter can disturb the communications of
the track circuit, degrading the safety of the trains circu-
lation. On the other hand the 4Q converter has the main
benefit to give a nearly sinusoidal line current in both
directions of energy flow and the mitigation of reactive
power drawn from the line. In fact the 4Q converter is
based on the use of forced commutation switches (GTO,
IGBT) and presents a sinusoidal current absorption in
phase with the contact line voltage. Moreover, this con-
verter is intrinsically bidirectional and then it can be used
both for traction and regenerative braking phases.
The aim of this paper is an analysis of the current ab-
sorbed by the high speed trains through a suitable model
of more 4Q converters. Thanks to a control logic applied
in this work, it is possible to interlace two or more con-
verters in order to evaluate the harmonic contribution of
both single converter and the interlaced configuration.
The simulation results obtained with an electromagnetic
analysis will be presented.
2. Mathematical Model of the 4Q Conver ter
The principle scheme reported in Figure 1 shows how the
four-quadrant converter is structurally equal to a single
phase voltage source inverter and employs the same
switches used for the motor drives. This is an advantage
Current Distortion Evaluation in Traction 4Q Constant Switching Frequency Converters 130
because the various branches can have a modular con-
struction [2]. The system includes:
he transformer secondary side;
four switches (GTO or IGBT), T
1
T
4
;
four freewheeling diodes D
1
D
4
;
a DC link with capacitive middle circuit C be-
tween the terminals AK and working at the im-
posed voltage V
d
=cost;
a second harmonic filter L
2
C
2
placed downstream
the main bridge, tuned to a frequency f
2
=2f
1
dou-
ble to the line one.
The system has two main purposes. The first one is the
absorption from the contact line, at a voltage e
1
and fre-
quency f
1
, of a current having the fundamental harmonic
i
1
in phase with e
1
and with low harmonic content, in
order to respect the following conditions:
1
~1 distortion factor
cos 1 power factor =
(1)
The second purpose is the absorption from the line of a
power with a mean value P
1
, pulsing at the frequency 2f
1

and the supply of the three-phase motor drive inverters
connected to the dc link with a continuous power P
d
.
In order to study the behaviour of this converter, the
modelling process starts by a mathematical representation
of the discrete operation modes of the converter. The
discrete model describes each working mode through
separate equations. Figure 2 shows a simplified repre-
senttation of the two-level converter from which the discrete
model is derived.
i
1
e
1
e
2
T
1
T
2
T
4
T
3
D
1
D
2
D
3
D
4
M
K
N
A i
i
s
L
2
C
2
C V
d
+
-
I
d

Figur e 1. Pr inciple scheme of a four -quadr ant conver ter .
e
1
=line voltage at a fr equency f
1
; i
1
=line cur r ent

Figur e 2. Equivalent cir cuit of a four -quadr ant conver ter
In the equivalent circuit of Figure 2, the transformer
secondary side is represented by:
an ideal voltage generator (e
2
=e
1
/h, where h is the
transformer ratio);
an inductance L, equivalent to the leakage trans-
former one;
a resistances R
s
mainly due to the switches;
load resistances R
0
;
DC link capacitance C.
Considering the state variables i
s
and v
c
, the converter
state equations are the following:
2
2 2
sin( )
s c
di
L R i v E t
2
0
C C
dt
dv v
C i
dt R
e = +
=
(2)
where is the angular frequency. Rewriting these equa-
tions in matricial format, it comes out:
2 2 2
0
1 0 s
1 1/ 0 0
s
C C
i R i L E d
v R v C dt
e | | ( | | | |
= +
| | | ( (
\ . \ . \ .
in( ) t (
(3)
or E x A x Z + = where x is the state vector and is
its time derivative.
x
Taking into account that the converter is a two level
type, there is another possible operation mode, described
by the following matrix A:
0
1
1 1/
s
R
A
R
(
=
(

(4)
Looking at the matrix A, it is possible to note that in
these two operation modes the first element of the first
row and the second one of the second row are the same.
Therefore, the state equation can be rewritten to:
2
0
0 s
1/ 0 0
s
R L E
x x
R C
e
( | |
= +
| (
\ .
in( ) t (
(
1
(5)
where depends from the operation mode and can as-
sumes the value 1 or -1.
In this analysis, the switches resistance R
s
has been ne-
glected.
The transformer leakage reactance depends by the
contact line frequency:
1
2 X L f L e t = = (6)
and it causes a lagging phase shift of an angle between
the converter voltage
2
V
and the supply one 2
E
.
Therefore the first harmonic component of voltages and
current become:
( ) ( ) ( )
2 2 1 2 2 1 2 2 1
2 sin , 2 sin , 2 sin i I t e E t v V t e e e = = = (
where the rms value of the AC voltage v
2
is related to th
7)
e
DC one V
d
through a proportional coefficient:
(8)
The
a
whe

DC link voltage of the single-phase converter has
significant ripple component at twice the supply fre-
quency. In fact, as it is possible to note in Figure 1 and
reported in [3], the current i given by the 4Q converter to
the middle circuit is composed by two components:
s d
i I i =
re:
( ) ( )
2
2
cos sin
d
k E
I k I
X

= (9)
is the direct component of the i(t) r
)
=
esponsible of the
power absorption, and
(
2 1
cos 2
s
i k I t e = (10)
is an harmonic at the frequency f
2
=2f
1
, with the follow-
ing rms value:
2
s
k I
I
2
tan
2 2
k E
X

= =

(11)
The i
s
(t) is absorbed by the d
at
orption the input current
edicated filter L
2
C
2
tuned
frequency f
2
so that into the dc section that feeds the
motor inverter flows the only continuous component I
d
.
Referring to the control, a smart modulation [4] has
been applied, where the input current follows a suitable
sinusoidal reference in order to have a sinusoidal absorp-
tion, as explained in the following paragraph.
3. Cur r ent Modulation
In order to have a sinusoidal abs
follows a suitable sinusoidal reference.
The AC reference current is obtained by multiplying
the AC line voltage with a suitable equivalent conduc-
tance, in accordance with
1
( ) ( ) i t G v t = (12)
where i is the phasor of the AC phase current and 1
v
is
the fun amental component of the AC contact line volt-
age.
The AC voltage contains the fundamental component
an
d

d the component v" corresponding to the perturbations
present in the AC line (i.e. harmonics), the Park vector of
the AC voltage being
''
1
( ) ( ) ( ) v t v t v t = + (13)
Thus, the instantan
ve
eous real power, expressed as Park
ctor, is given by
( ) ( )
( )
* *
( ) R
a a
p t v i = =
1
" 2 *
1 1
e Re
Re
v i G v v
G v G v v
=
= +
(14)
where is the complex conjugated value of
*
1
v
1
v .
id th e sin-
gle phase converter, the AC - DC power balance is given
by
Cons ering null power losses associated wi th

( )
*
1
( ) Re
DC DC
p t v v v i = = (15)
and the DC current, neglecti
pressed with
ng the ripple, can be ex-
( )
*
1
1
Re
DC
v v
v i G
DC
v
= (16)
Considering the curren
a non-linear relation between the DC voltage v , the con-
tro
ts flowing through the DC bus,
dc
l variable G and the load current i
load
is obtained
( )
( )
*
1
Re
2
DC
load
v v
dv C
i G
dt
= =
" *
2
1
1
Re
DC
load
DC DC
v
v v
v
i G G
v v
=
(17)
Since V
1
can be considered constant, the (17) can be
linearized as follows:
( )
2 2
1 1
2
DC
d v V V C
dt V
A
2
" *
1
Re
DC
DC DC
load
DC
G G v d
V
v v
where d i G
V
= A + A +
= A
(18)
d represents a perturbation, due to the AC line distur-
bances and the DC ripple.

Its transfer function is
For the DC voltage control, a proportionalintegral (PI)
controller has been chosen.
( )
I
P DC
k
G k F s v
s
| |
A = + A
|
\ .
(19)
where F(s) is the transfer functio
required for reducing the DC bus ripple.
puting the val-
ue
n of the low pass filter
The Equations (18) and (19) constitute the closed loop
of the DC voltage control that allows com
s of the PI parameters. In a first approximation the
transfer function F(s) can be neglected, obtaining
( )
2 2
DC
s
v s d
V V C G
2 1 1
2
P I
DC DC DC
s s k k
V V V
A =
| |
(20)
+ +
|
\ .
The denominator, considering a damping ratio of 0.707,
constrains the PI controller parameters to respect the fol-
lowing relations:
2
1
1
a
P
V G G
k
V
| |
> nd
I P
DC DC DC
k k
V V C
=
|
\ .
(21)
4. Model of the System
retical analysis above de-
he system has been imple-
In order to validate the theo
scribed, a suitable model of t
mented in the EMTP-ATP dynamic simulation tool. The
data employed for the modelization refer to a real High
Speed Train operating in Italy in 25kV50Hz lines. This
2
2
cos
d
V k V
= =
E
132

Figur e 3. Tr action cir cuit schematic diagr am of the Italian
high speed tr ain

Figur e 4. Block diagr am of the implemented model of the
system
ach. Every motor drive is supplied by double 4Q
um peak power equal to 1500kW.
Th
l one. The power circuit has been modeled with
th
the two input stage 4Q
co
s can be identified. The first one is the DC
lin
Inter lacing of the Conver ter s
requency
[5] lets an opportunely shifting of the switching time for
l absorbed
cu
s
re
f the angle . Therefore the Fourier se-
rie
vehicle is constituted by two locomotives with two motor
drives e
converters input stage through the main transformer as
depicted in Figure 3.
Each 4Q converter is sized for a rated power equal to
900kW and a maxim
ese are relevant power values for a switching converter
that has to be small and light enough to be installed on-
board.
The 4Q converters are constituted by a power part and
a contro
e circuital elements already available in EMTP-ATP.
The MODELS language has been used for implementing
the proposed converter control.
The final schematic model is reported in Figure 4,
where it is shown only one of
nverters.
In the block diagram represented in Figure 4 three
control loop
k voltage control loop constituted by the voltage
measurement, the low-pass filter explained in the previ-
ous paragraph, the comparison with the reference value
V
0ref
and the PI controller. Its output is the value of the
equivalent conductance G that keeps the DC link voltage
constant varying the power requested or injected by the
traction motors and auxiliary services. The second loop is
the reference current generator. It considers the input
voltage measurement followed by a filter dedicated to the
high frequency disturbances. The obtained value multi-
plied with the equivalent conductance G gives the refer-
ence current that the converter, through the switching
modulation, has to generate in order to balance the input
and output powers. The third loop is related to the DC
component compensation in the AC input current. In fact
its output value is a constant current that, algebraically
added to the reference one, allows to cancel the DC
component avoiding the saturation of the input trans-
former.
5. The
The use of modulation techniques at constant f
the different 4Q converters onboard the train.
The final goal is to interlace more converters in order
to diminish the harmonic content in the tota
rrent. The interlacing operation gives a shifting through
the current waves coming out from the converters that
have to be the one that minimize the harmonic content.
In order to preliminary study the benefit of the inter-
lacing practice, two boost converters are considered. A
ported in Figure 5, their current waves are shifted of a
generic angle .
The two waves have the same duty cycle and they
are only shifted o
s f and f' of the two converters are the following:
| |
0
1
( ) cos( ) ( )
n n
f t A A n t B sen n t e e
= + +

| |
0
1
)
n n
t n '( ) cos( ) ( f t A A n t n B sen n e e
= + + +

b
+
Applying the known trigonometric formula, the f' ca
e rewritten as:
n

Figur e 5. Waves shifting of the two inter laced boost con-
ver ter s cur r ents

Figur e 6. Behavior of the THD, var ying the duty cycle and the shifting between the cur r ents absor bed by the two boost
Applying the known trigonometric formula, the f' can
)
)
conver ter s

be
here:
rewritten as:

w
'
n
cos( ) (
' cos( ) (
n n
n n n
A A n B sen n
B B n A sen n

+
=

=
Considering that the current absorbed by the train is
th
| )
e sum of the currents absorbed by the two converters,
the resulting function to consider for the THD calculus is
given by the sum of the two functions, that means:
( ) ( ) '( )
t
f t f t f t
|
0
1
' '' cos( ) '' (
n n
A A n t B sen n t e e
= + +

= + =
with
n
'' '
n n
A A A = + and
n
'' '
n n
B B B = + .
his case the ab HD is define following: In t solute T d as
2
1
n
THD Z =

(22)
with
2
it is possible to no e THD calculation the
first harmonic has been also considere referred to the
sw
cycle , her
ha
THD is null for =0.5
he
|
2 2
( '') ( '')
n n n
Z A B = + .
As te, in th
d ,
itching frequency, because all the ripple is an undesir-
able component. The THD so calculated is function of the
shifting angle between the two waves and of the duty
e supposed the same for the two converters.
In order to underline the THD dependence from the
parameters and , a simulation in Matlab environment
|
0
1
' cos( ) ' ( )
n n
s been carried out. The (22) has been considered,
varying the duty cycle between 0.1 and 0.9 and the shift-
ing between 0 and 2 and supposing the current ripple
amplitude equal to the 1% of the continuous component.
The result is reported in Figure 6.
From Figure 6 it is possible to make some considera-
tions. First of all, as expected, the
and =, because it is the only case in which the two tri-
angle waves are symmetric and in opposite phase, there-
fore all harmonics are cancelled.
From the optimal point, the THD initially grow fast,
both varying and , underlining the fact that also small
dissymmetry in the two converters switching diminish the
advantages given by the interlacing. The optimal shifting
is always equal to and in this case the THD is always
minor than the case of not interlaced converters (=0).
Finally, the THD is directly proportional to the ripple
amplitude respect to the DC component, and for this t
'( ) f t A A n t B sen n t e e +
= +

Figur e 7. Shifting of the four switching fr equencies
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
-400
-300
-200
-100
0
100
200
300
400
500
t [s]
V

[
V
]
line voltage

Figur e 8. Line voltage wavefor m measur ed on the secondar y side of an auxiliar y tr ansfor mer
effective value
account the different typologies of inter-
mally running
on
s coming out from the practical cases are operable locomotives and electro trains nor
higher than the ones reported in Figure 6, considering the
relevant current undulation due to the low switching fre-
quencies.
Taking into
the railway lines, it is possible to note that after each
input transformer there are 2 or 4 4Q converters. Each
converter is supplied by a dedicated winding of the main
transformer.
ch converter switching frequency has to be
re
ain
fr
g frequencies.
mutation instants are calculated
st
Model for Har monic
Analysis and Simulation Results
the 4Q
conve configu-
low short circuit level of the rail-
w
due to the resonance phenomena
al
je
dur ing a measur ements sur vey
tude [%]
It comes out that, in order to have a good interlacing,
the shift of ea
spectively equal to or of the switching period.
In this way the ripple of the current absorbed by the
transformer due to all the 4Q converters has the m
equency equal to two or four times the switching one
with the advantages of a minor amplitude and easier fil-
tration.
In Figure 7 is possible to note the shifting of the four
switchin
Analyzing the smart modulation here applied, it is pos-
sible to note as the com
arting from points on the reference current that have the
same time distance. The interlacing can be easily ob-
tained intercalating these time instants referred to the
different converters.
6. 4Q Conver ter s
In order to evaluate the harmonic contribution of
rters in single operation and in interlaced
ration, some numerical simulations are carried out, using
ATP/EMTP program.
These converters are often supplied by a distorted in-
put voltage, due to the
ay lines and a greater harmonic content allowed by the
dedicated standards. A typical waveform measured on the
secondary side of the transformer for auxiliary services
ins reported in Figure 8.
The most significant components are reported in Table 1.
This distortion is mainly
ong the line, also enhanced by the harmonic current in-
cted by the trains. In order to have a correct operation of
the 4Q converters it is necessary to have a cleaner input
voltage obtained installing LC filters in the input stage.
Moreover, the greater components (>50
th
) are carefully
Table 1. Har monic content of the line voltage r ecor ded
Harmonics Order Frequency
[Hz]
Harmonic Ampli-
1 50 100
3 150 7.14
15 750
1
25 1250 5.1
6.12
19 950 5.1
21 1050 2.76
23 1150 3.06
filtered to oid interferen signaling sy
In this a alysis, each onverter has be
consideri the following characteristics:
Hz;
equency: 500 Hz.
e converter have been
, while the DC load with
presents the current
ab
mines the input cur-
re
is
f the switches. This DC compo-
ne
its
ach half period.
av ce with stem.
n
ng
4Q c en studied
rated voltage of the contact line: 25 kV, 50
transformation ratio of the onboard transformer:
25/0.7 kV;
DC link voltage: 1800 V;
rated power: 900 kW;
switching fr
The power components of th
modeled with ideal switches
current generator. A negative value re
sorbed during the traction phase, while a positive one
represents the regenerative braking.
In the controller are implemented two control loops.
The first one gives the value of conductance G that
multiplied by the input voltage deter
nt reference. The value of the equivalent conductance G
obtained through PI controller comparing the measure
of the DC link voltage with its reference in order to keep
the DC voltage constant at its nominal value of 1800 V.
The function of this controller is to guarantee the equiva-
lence between the input power and the one absorbed by
the load. In order to assure a uniform power absorption
among the various converters, there is only one regulator
for all the converters.
The second loop is necessary to cancel the DC com-
ponent in the AC current that can be generated by the low
switching frequency o
nts can be dangerous for the onboard transformer be-
cause can saturate the magnetic core. Indeed this trans-
former is not oversized due to the need to reduce its
weight and volume. Therefore it is really sensible to the
direct current component. Because the four 4Q converters
are supplied by four independent transformer windings, it
is necessary to adopt one of these regulators for each
converter. In Figure 9 the current absorbed by one 4Q
converter and its Fourier analysis are reported. It is pos-
sible to note a high ten order components corresponding
to the switching frequency, but also an appreciable third
harmonic. This last one is due to the difficulty to follow
up the sinusoidal reference having a so low switching
frequency.
The interlacing of the four converters has really re-
duced the harmonic content, as it is well represented in
Figure 10, where the current absorbed at pantograph and
Fourier analysis are reported. In particular there are no
more harmonics at the switching frequency, but there is
still the third harmonic, due, as told before, to the low
switching frequencies.
What told above is confirmed by Figure 11, where it is
possible to note that there is a difficulty of the converter
to switch at the end of e

0 5 10 15 20 25 30
0,0
0,2
0,4
0,6
0,8
1,0
harmonic order

Figur e 9. Cur r ent absor bed by one 4Q conver ter and its
Four ier analysis

cing in improving the current waveform,
e out of service of one converter has been considered.
Th
es out again. In fact it
is
e end of each
ha
als with a power quality analysis regarding
the modern interoperable high speed
ore four quadrant (4Q) converters.
In fact, in the
In order to better evaluate the contribution of the four
converters interla
th
is condition can occur during the train operation.
The current absorbed at pantograph in this second case
with only three 4Q converters working and its Fourier
analysis are reported in Figure 12.
It is possible to note the worst harmonic content re-
spect to the previous case and, most of all, the contribu-
tion of the switching frequency com
evident the presence of the ten order component. Re-
garding the third harmonic, the situation is not changed,
always due to the low switching frequency.
In Figure 13 is reported the shifting of the three
switching frequencies and also in this case it is evident
the difficulty of the converter to switch at th
lf period.
7. Conclusions
The paper de
the input stage of
trains constituted by m
In fact, the use of the 4Q converter allows to overcome
the limits imposed by traditional rectifier.
However, the high power needed for the train accelera-
tion, in the order of 6000 kW per locomotive, does not
allow to have high switching frequencies.

0 5 10 15 20 25 30
0,0
0,2
0,4
0,6
0,8
1,0
harmonic order

Figur e 10. Cur r ent at pantogr aph in case of four 4Q con-
ver ter s wor king and its Four ier analysis

Figur e 11. Cur r ents absor bed by all the four 4Q conver ter
inter laced

quently the absorbed current presents a high
rip le value not always tolerable by the system.
rrent at
th
of the interlacing in improving the cur-
re
recent realization they reach the maximum value of 500
Hz. Conse
p
The presence of more converters gives the opportunity,
using dedicated control logics, to interlace them in order
to reduce the harmonic content of the absorbed cu
e pantograph.
Computer simulations have been carried out using a
suitable model of more 4Q converters, for determining
the contribution
nt waveform.
First, the case of four interlaced 4Q converters has
137

0 5 10 15 20 25 30
0,0
0,2
0,4
0,6
0,8
1,0
harmonic order

Figur e 12. Cur r ent at pantogr aph in case of thr ee 4Q
ver ter s wor king and its Four ier analysis

con-

Figur e 13. Cur r ents absor bed by thr ee 4Q conver ter
laced

ed operation allows the reducing of the absorbed
urrent ripple, since the equivalent switching frequen
erformances of the
REFERENCES
[1] G. W. Chang, en, Modeling
ic systems for transportation, (in
, F. Foiadelli, G. C. Lazaroiu, and D. Zaninelli,
Carmeli, F. Castelli Dezza, and G. Superti Furga,
urga,
Zaninelli, Harmonic disturbances in
ics due to
asso, R. Lamedica, R. Manigrasso, G. Sani, G.
chi, F. Foiadelli, and D. Zaninelli, Developments
Allan, B. Mellitt, and J. Taufiq, A power
inter -
Superti Furga, and E. Tironi, Reference power network
for the harmonic propagation analysis, European Trans-
actions on Electrical Power Engineering ETEP, Vol. 2,
1992.
[10] G. Bur
been analyzed. In particular the analysis shows that the
nterlac
c cy is of power quality studies in electric transportation system,
Electrical Power Quality and Utilisation, Journal, Vol. 6,
No. 2, 2005.
[11] S. Burdett, J.
four times the single converter one. However the low
harmonic components, such as the third one, cannot be
canceled by this solution, therefore they have to be fil-
tered by traditional LC passive filters.
The second case analyzed regards a condition that can
occur during the train operation such as the out of service
of one converter. In this situation the p
fact
train are guarantee by the other three converters, even if it
is not possible to completely cancel the harmonic com-
ponent at the switching frequency.

H.-W. Lin, and S.-K. Ch
characteristics of harmonic currents generated by
high-speed railway traction drive converters, Transac-
tions on Power Delivery, IEEE, Vol. 19, No. 2, pp.
766773, April 2004.
[2] F. Perticaroli, Electr
Italian), Casa Editrice Ambrosiana, Milano (Italy), Janu-
ary 2001.
[3] M. Brenna
Four quadrant converter analysis for high speed trains,
12th International Conference on Harmonics and Quality
of PowerICHQP06, Cascais, Portugal, 16 October
2006.
[4] M. S.
Smart modulation: A new approach to power converter
control, EPE01, Graz, Austria, 2729 Aug. 2001.
[5] M. S. Carmeli, F. Castelli Dezza, and G. Superti F
Constant frequency current modulation algorithm based
on linkage flux, PESC03, Power Electronics Specialists
Conference, Acapulco, Mexico, 1519 June 2003.
[6] F. Foiadelli, G. C. Lazaroiu, and D. Zaninelli, Probabil-
istic method for harmonic analysis in railway system,
2005 IEEE PES General Meeting, San Francisco (USA),
1217 June 2004.
[7] P. Pinato and D.
electric traction system overhead lines, IEEE X Interna-
tional Conference on Harmonics and Quality of Power,
Rio De Janeiro (Brasil), 69 October 2002.
[8] W. Runge, Control of line harmon
four-quadrant-converter in AC tractive stock by means of
filter and transformer, in Proc. EPE97, pp. 3.4593.464,
1997.
[9] A. Cap
or and harmonic comparison of AC railway power
electronic traction converter circuits, Fifth European
Conference on Power Electronics and Applications, pp.
235240, 1316 Sep. 1993.
Rotating Capacitor and a Tr ansient Electr ic
Networ k
Haiduke SARAFIAN
1
, Nenette SARAFIAN
2

1
The Pennsylvania State University, York, PA 17403;
2
Nenette Sarafian, Hershey Medical Center.
Emall:
1
has2@psu.edu,
2
nsarafian@hmc.psu.edu

Received May 16
th
th
, 2009; accepted July 16
th
, 2009.
ABSTRACT
This paper presents a rotating parallel-plate capacitor; one of the plates is assumed to turn about the common vertical
axis through the centers of the square plates. Viewing the problem from a purely geometrical point of view, we evaluate
the overlapping area of the plates as a function of the rotated angle. We then envision the rotation as being a mechani-
cal continuous process. We consider two different rotation mechanisms: a uniform rotation with a constant angular
velocity and, a rotation with a constant angular accelerationwe then evaluate the overlapping area as a continuous
function of time. From the electrostatic point of view, the time-dependent overlapping area of the plates implies a
time-dependent capacitor. Such a variable, a time-dependent capacitor has never been reported in literature. We insert
this capacitor into a series with a resistor, forming a RC circuit. We analyze the characteristics of charging and dis-
charging scenarios on two different parallel tracks. On the first track we drive the circuit with a DC power supply. We
study the implications of the rotation modes. We compare the response of each case to the corresponding traditional
constant capacitor of an equivalent RC circuit; the quantified results are intuitively just. On the second track, we drive
the circuit with an AC source. Similar to the analysis of the first track, we generate the relevant electrical characteris-
tics. In the latter case, we also analyze the sensitivity of the response of the circuit with respect to the frequency of the
source. The analyses of the circuits encounter nontrivial differential equations. We utilize Mathematica [1] to solve
these equations.
Keywords: RC-Series Transient Electric Network, Rotating Parallel-Plate Capacitor, Mathematica
1. Intr oduction and Motivation
It is a far-fetched concept to think about a conventional
transient electrical circuit and incorporate its signal char-
acteristics into a discrete and abstract geometry problem.
The authors have even taken the initiative one step fur-
ther relating these two basic concepts to kinematics. This
article shows how these three discrete concepts are
brought together and molded into one coherent and
unique research project. A thorough literature search of
the standard undergraduate and graduate physics texts
and reference books reveals the lack of any similar
analysis [2]. In the course of quantifying the signal
analysis we encounter challenging nontrivial differential
equations. In our experience in analyzing comparable
technical problems it is proven that Mathematica is the
most convenient and efficient tool of choice. Utilizing its
coherent and simultaneously applicable symbolic, nu-
meric and graphic features assists the authors to focus on
the physics issues of interest. Mathematicas tools also
suppress the challenges of developing basic computer
programs from the ground-up.
The flow chart of this article including the Introduction
contains three additional sections. In Section I, as a pure
geometry problem we apply Mathematica to evaluate the
overlapping area of the two rotated squares about their
common vertical axis. We then incorporate the rotational
kinematics and view the rotation as a mechanical process
and consider two different scenarios: 1) a symmetric,
uniform rotation; and 2) an asymmetric, accelerated rota-
tion. In Section 3, we view the overlapping squares as
being two parallel metallic plates that are separated by a
gap forming a parallel-plate capacitor. Following the tra-
ditional textbook approach [2], we assume the gap is
much smaller than the dimensions of the square. This
suppresses the fringe effects. Since the area of the over-
lapping plates evaluates the capacitance of the capacitor,
the rotating plates make the capacitor variable. Technical
literatures, particularly Mathematica-based reference
books such as [3] contain a wealth of references within
Rotating Capacitor and a Transient Electric Network 139
themselves, yet lack our novel view. It is the ultimate
objective of this project to analyze the electric signal
characteristic response of the electrical network to the
kinematics of the rotating plates. Specifically, in this arti-
cle, we address the modifications of the characteristic
signal responses of the electrical circuits composed of a
resistor connected in a series with our designed,
time-dependent capacitor. In particular, we analyze the
characteristics of the RC circuits driven with DC as well
as AC sources. The third and conclusion section, in con-
junction with our recent analysis, descriptively outlines
the results of more generalized cases.
{ , , } {{ , tan },
2 2 2
1 - tan
L
2
{ ( ), },
2 2
1 tan
2
L L
{- tan , }}
2 2 2
L L
a b c
L
u
u
u
u
=
+
2. Analysis
2.1 Geometr y
Figure 1 shows two identical overlapping squares. The
bottom square designated with non-prime vertices is fas-
tened to the xy coordinate system. The top square, desig-
nated with prime vertices is rotated counter clockwise
about the common vertical axis through the common ori-
gin O by an angle . The squares have the side length of
L and the rotation angle is the angle between the semi
diagonals
1
op
and
'
1
op
To evaluate the overlapping area of these two squares
we evaluate the area of trapezoid oabco; the overlapping
area then equals four times the latter. The intersecting
points of the rotated sides of the top square with the sides
of the bottom one are labeled a, b, and c. Utilizing the
coordinates of these points, the area of the trapezoid is
the sum of the areas of two triangles abc, and oac.

Figur e 1. Display of two r otated squar es. The bottom squar e
is fastened to the xy coor dinate system, the top squar e is
r otated counter clockwise by r adian
To evaluate the coordinates of a, b, and c we write the
equations for the slanted lines,
' '
4 1
p p and
' '
1 2
p p and
intersect them with the sides of the bottom square. Inter-
section of the former with the 4 1
p p

and 1 2
p p
gives
the coordinates of a and b, respectively. Similarly, the
intersection of the latter with
1 2
p p

yields the coordi-
nates of c. Theses are:

{0, 0,1}; o
(1)

To evaluate the areas of the needed triangles, we con-
vert the above coordinates into Mathematica code. The
inserted 1's in the third position of the coordinates of the
origin and the intersecting points are for further calcula-
tions.
[ , ] { , [ ],1};
2 2 2
L L
a L Tan
u
u

= =
1 [ ]
2
Tan
L L
u
[ , ] { ( ), ,1};
2 2
1 [ ]
2
b L
Tan
u
u

=
+

[ , ] { [ ], ,1};
2 2 2
L L
c L Tan
u
u

=
We define two auxiliary lists,
[ , ] { [ , ], [ , ], abc L a L b L c[ , ]}; L u u u

u =
[ , ] { , [ , ], [ , ]}; oac L o a L c L u u u

=

Figur e 2. The nor malized values of the over lapping ar ea of
the squar es as a function of the r otation angle u

Figur e 3. The gr aphs ar e the nor malized values of the over lapping ar eas for : a) a unifor m r otation with e=t/2r ad/s, (the left
gr aph) and b) a unifor m angular acceler ation with o=t r ad/s
2
(the r ight gr aph)
The needed areas are,
1
[ , ] [ [ , ]];
2
areaABC L Det abc L u u

=
1
[ , ] [ [ , ]];
2
areaOAC L Det oac L u u

= therefore,
[ , ] [ , ] [ , ]; areaOABCO L areaOAC L areaABC L u u u

= +
We divide the value of the overlapping area by the area
of the square, L
2
, and plot its normalized values as a
function of the rotation angle u. Figure 2 shows the nor-
malized area begins and ends at the same values. The
value of the area after a t/4 radian turn drops to about
83% of the maximum value. The plot as one anticipates is
symmetric about t/4. The Mathematica codes follow,

Plot[4 areaOABCO[L, u]/L
2
,{ t,0, t/2},Ticks{Range
[0, t/2, t/16],Automatic},PlotRange{0,1},AxesLabel
{"u ,rad","normalizedarea"},GridLines{Range[0, t/2,
t/32],Automatic}];
2.2 Modes of Mechanical Rotations
In this section we extend the analysis of Subsection 2.1.
Here, instead of viewing the rotation as being a discrete
and purely geometrical concept, we view it as a kinematic
process. We set the rotation angle ( ) t t u e = ; that is, we
introduce the continuous time parameter t. For
2
=
T
t

Figure 4. The schematics of a DC driven RC circuit. Throwing
the DPDT switch onto as charges the capacitor, while throw-
ing the switch onto bs discharges the charged capacitor
with the period 4 T s = , we explore the uniform rotation.
For an asymmetrical case, we consider a rotation with a
constant angular acceleration. According, for the latter, to
rotate the square by 90
in one second we set

2
1
2
t u o = ,with
2
rad
s
o t = . The corresponding normal-
ized overlapping areas are displayed in Figure 3.
The Mathematica codes follows,
UniformRotation = Plot [4 areaOABCO [L,u]/L
2
/.u
t/2t, {t, 0, 1}, Ticks{Range[0,1,1/8],Automatic},
PlotRange{0,1},AxesLbel{"t,s","area"},GridLines
{Range[0,1,1/16], Automatic}];
AcceleratedRotation=Plot [4areaOABCO[L,u]/L
2
/.u
t/2t
2
,{t,0,1},Ticks{Range[0,1,1/8],Automatic},
PlotRange{0,1},AxesLabel {"t,s","area"}, GridLines
{Range[0,1,1/16], Automatic}];
Show [GraphicsArray[{UniformRotation, Accelerated-
Rotation}]];
3. Electr ical Networ ks
Now we consider a RC series circuit. One such circuit
driven by a DC power supply is shown in Figure 4. The
circuit is composed of two loops. Throwing the DPDT
(Double-Pole Double-Throw) switch to as position
charges the capacitor, while setting the switch to a bs
discharges the charged capacitor.
As we pointed out in the Introduction, in this section
we view the overlapping squares as being two parallel
metallic plates that are separated by a gap forming a par-
allel-plate capacitor. Since the capacitance of a paral-
lel-plate capacitor is in proportion to the overlapping area
of the plates, the continuous rotation of the plates makes
the capacitor time-dependent. It is the objective of this
section to analyze the characteristics of the electric re-
sponse of one such time-dependent capacitor in the
charging and discharging processes.
3.1 Char acter istics of Charging and Dischar ging a
DC Driven RC Cir cuit with Time-Dpendent
Unifor mly Rotating Plates
1
t
e
t
table C0Discharge=Table [{t,N[ /.t1./6]}, {t,0,1,

1/16}]; listC0Discharge=ListPlot [tableC0D ischarge, Pl-
otStyle{PointSize[0.02], GrayLevel [0]}, GridLines->
{Range [0,1,1/8], Automatic}, Ticks {R ange [0,1,1/8],
Automatic}];
For the charging process we apply Kirchhoff circuit law
[2], this gives
tableCtUniformDis-
charge=Table[{k,
0
( ) 1 1
( ) 0
( )
A dQ t
Q t
dt A t t t
+ = (2)
2
1
[ /. ,{
4 [ , ] 2
L
NIntegrate k
areaOABCO L
e
t
u
t u
,0, }]
/.t1
./6},{k,0,1,1/16}];
For the sake of convenience, we assume V C
0
=1,
where C
0
is the capacitance of the parallel-plate with the
plates completely overlapped, Q(t) and A(t) are the ca-
pacitor's charge and the overlapping area at time t, re-
spectively; A
0
is the area of one of the squares; and
0
RC t = is the time-constant of the circuit. For a con-
stant, time independent capacitor,
0
( ) A t A = , and Equa-
tion (2) yields the standard solution ( ) Q t
listCtUniformDis-
charge=ListPlot[tableCtUniformDischarge,PlotStyle{P
ointSize[0.02],GrayLevel[0.5]}];
s12=Show[{listC0Discharge,listCtUniformDischarge},
A-xesLbel{"t,sec","Q"},PlotRange{Automatic,{0.,
1.02}},Plot Label->"Dis charging"];
1
t
e
t
Show[GraphicsArray[{s34,s12,UniformRotation}],Ima
geSize600];
= . In
this equation the maximum charge is normalized to unity.
For the rotating plates, however, Equation (2) does not
have an analytic solution. We apply Mathematica
NDSolve along with an appropriate initial condition and
solve the equation numericallythis yields Q(t).
Graphically, we compare its characteristics vs. the char-
acteristics of an equivalent RC circuit, this is shown Fig-
ure 5. The Mathematica code follows,
It is interesting to note that the charging and discharg-
ing signals respond differently to the time-varying ca
pacitors; the impact of the time-dependent capacitor is
more pronounced for the former. Moreover, for the value
of the chosen time-constant of the circuit, t=1/6 s, al-
though the time-independent capacitor reaches its plateau
within one second, the variable capacitor requires a
longer time span.
solQUniformRotation=With[{t=1./6}, NDSolve[{Eva-
luate[(Q'[t]+1/t L
2
/(4 areaOABCO[L,u]/.ut/2 t) Q[t]
-1/t)]==0,Q[0]==0},Q[t],{t,0,1}]];
3.2 Char acter istics of Char ging and Dischar ging
DC Dr iven RC Cir cuit with Time-Dependent
Acceler ated Rotating Plates
1
t
e
t
s34=Plot [{1- /.t1./6,Q [t]/. solQUniformRota-

tion},{t,0,1},PlotStyle{GrayLevel[0],GrayLevel [0.5]},
AxesLabel{"t,sec","Q"}, PlotLabel->"Charging",Plot-
Range{Automatic,{0,1.02}},GridLines {Range [0,1,
1/8],Automatic}];
Similarly, we analyze the characteristics of the dis-
charging process. Equation (2) for the corresponding dis-
charging process is,
0
( ) 1
( ) 0
( )
A dQ t
Q t
dt A t t
+
One may comfortably also apply the analysis of Subsec-
tion 3.1 to generate the characteristic curves associated
with the uniformly accelerated rotating plates. The
Mathematica codes may easily be modified to yield the
needed information. The codes along with the associated
graphic outputs are:
solQAcceleratedRotation=With [{t=1./6},NDSolve [{-
Evaluate [(Q'[t]+1/tL
2
/(4areaOABCO [L,u]/.ut/2 t
2
) Q
[t]-1/t)]==0,Q[0]==0},Q[t],{t,0,1}]];
= . This equa-
tion for a constant, time independent capacitor,
0
( ) A A t = , yields
( ) 1
( ) 0
dQ t
Q t
dt t
+ = , and gives
( )
t
Q t e
t
=
. For the rotating capacitor, however, its solu-
tion is
0
0
1
( )
( ) Q t
1
t
e
t
s56=Plot[{1- /.t1./6,Q[t]/.solQAcceleratedRotat
ion},{t,0,1},PlotStyle{Hue[0.7],{Dashing[{0.02}], Gr-
ayLevel[0.5]}},AxesLabel{"t,sec","Q"},PlotLabel->
"Charging",PlotRange{Automatic,{0,1.02}},GridLines
t
A
d
A
e
{Range[0,1,1/8],Autmatic},Ticks{Range[0,1,1/8],A
}
= . To solve the latter we apply
Mathematica NIntegrate. This yields the needed values.
The results are displayed in Figure 5. The Mathematica
code follows,
utomatic}];
tableCtAcceleratedDischarge=Table[{k,
2
,0, }]
2
1
[ /. ,{
4 [ , ] 2
L
NIntegrate k
areaOABCO L
e
t
u
t u

/.t1./6},{k,0,1,1/16}];

Figur e 5. Display of char ging (the left gr aph) and dischar ging (the r ight gr aph) signals of unifor mly r otating plates. In each
gr aph the black and the gr ay cur ves/dots ar e the signals of the time-independent and time dependent capacitor s, r espectively.
For a compr ehensive under standing these gr aphs ar e to be incor por ated with the left gr aph of Figur e 3

Figur e 6. Display of char ging (the left gr aph) and dischar ging (the r ight gr aph) signals of unifor mly acceler ated plates. The
color codes ar e the same as Figur e 5. For a compr ehensive under standing these gr aphs ar e to be incor por ated with the r ight
gr aph of Figur e 3
listCtAcceleratedDischarge=ListPlot[tableCtAccelerat-
edDischarge,PlotStyle{PointSize[0.02],GrayLevel
[0.5]},GridLines{Range[0,1,1/8],Automatic},Ticks
{Range [0,1,1/8],Automatic}];
s78=Show[{listC0Discharge,listCtAcceleratedDischar
ge},AxesLabel{"t,sec","Q"},PlotRange{Automatic,
{0.,1.02}},PlotLabel->"Discharging"];
Show[GraphicsArray[{s56,s78,AcceleratedRota-
tion}]];
To form an opinion about the characteristics of the
charging signal for the variable capacitor, one needs to
view it together with the far right graph of Figure 3. The
rotating plates in this case are accelerated, meaning that
for identical time intervals, the overlapping area at the
beginning is greater than the overlapping area at the end
of the interval. The effects of the asymmetrical rotation
are most clearly visible at the tail of the signal. Similar to
the uniform rotation (see the lift graph of Figure 5) the
impact of the non-uniform rotation for the discharge sig-
nal is negligible.
3.3 Char acter istics of Char ging and Dischar ging
an AC Dr iven RC Cir cuit with Time
Dependent Capacitor
In this section we analyze the charging and the discharg-
ing characteristics of an RC series circuit driven with an
AC source. Schematically speaking, this implies in Fig-
ure 4 we replace the DC power supply with an AC source.
For this circuit, Kirchhoff's law yields
0
( ) 1 1
( ) sin(2 ) 0,
( )
A dQ t
Q t ft
dt A t

(3)
In this equation f is the frequency of the signal and the
voltage amplitude is set to one volt.
Equation (3) is a non-trivial differential equation. To
solve this equation, we apply NDSolve along with the
corresponding initial condition. The response of the cir-
cuit is compared to the equivalent circuit with a constant
capacitor. The Mathematica code follows,
solQac=NDSolve[{Evaluate[(Qac'[t]+1/ L
2
/L
2
Qac[t]
-1/ Sin[2 f t])/.{1./6,f0.6}]0,Qac[0]0},Qac[t],
{t,0,1}];
plotQac=Plot[Qac[t]/.solQac,{t,0,1},PlotStyleGrayL
evel[0],AxesLabel{"t,sec","Q"},PlotLabelAC Driver,
DisplayFunctionIdentity];
solQactUniform Rotation=NDSolve [{Evaluate [(Qact'
[t]+1/L
2
/(4area OABCO [L,]/./2t) Qact[t]-1/ Sin-
[2ft])/.{1./6,f0.6}]0,Qact[0]0},Qact[t], {t,0,1}];

Figur e 7. Display of the char ge on the capacitor vs. time.
The solid black cur ve is the signal of the time-independent
capacitor . The solid and dashed gr ay cur ves ar e the signals
of the time-dependent capacitor s for unifor m and acceler -
ated r otations, r espectively

Figur e 8. Display of two r otated r ectangles. The bottom
r ectangle is fastened to the xy coor dinate system; the top
r ectangle is r otated counter clockwise by u r adian

Figur e 9. Display of char ging (the left gr aph) and dischar ging (the r ight gr aph) signals of unifor mly acceler ated plates. In
both gr aphs the light gr ay cur ves ar e the over lapping ar ea; the black cur ves ar e the time-independent capacitor s signals,
while the medium gr ay cur ves ar e the cor r esponding time-dependent signals
plotQactUniformRotation=Plot[Qact[t]/.solQact Unifo-
rmRotion,{t,0,1},PlotStyleGrayLevel[0.5],AxesLabel
{"t,sec","Q"}, PlotLabelAC Driver];solQact Accel-
erated Rotation=NDSolve [{Evaluate [(Qact'[t]+1/t L
2
/(4
areaOABCO[L,u]/.ut/2 t
2
) Qact[t]-1/t Sin[2t f t])/.
{t 1./6,f0.6}]==0,Qact[0]==0},Qact[t],{t,0,1}];
plotQactAcceleraedRotion=Plot[Qact[t]/.olQactAcce-
leratedRotation,{t,0,1},PloStyle{Dashing[{0.02}],Gr-
ayLevel[0.5]},AxesLabel{"t,sec","Q"},PlotLabelAC
Driver];Show[plotQac,plotQactUniformRotation,plotQa-
ctAcceleratedRotation,ImageSize300];
Utilizing the Mathematica code, one may analyze the
frequency sensitivity response of the circuit. As the result
of one such analysis, we observe that the differences be-
tween the time-independent vs. the time-dependent sig-
nals are more pronounced with a frequency domain of
less than 1 Hz.
4. Conclusions
Since the inception of the first version of this project,
the square plates the authors have extended the scope
of their investigation. The idea of rotating the plates of a
parallel-plate squared capacitor is sound; however, as we
addressed in the aforementioned text the area of the
overlapping plates at its best is reduced to only 83% of
the maximum area. The square geometry puts a limit to
the overlapping area of the rotated plates, limiting the
impact of the corresponding time-dependent capacitor on
the characteristic signals. To enhance the impact, a po-
tential remedy is to replace the square plates with less
symmetric flat objects, e.g. a rectilinear or a curvilinear
shape such as a rectangle or an ellipse, respectively. For
the rectangular plates, for instance, by adjusting the
length and the width of the rectangle one is able to reduce
the value of the overlapping area at will. This, in turn,
more effectively impacts the capacitance of the
time-dependent capacitor. The unevenness of the sym-
metry of the rectangle about its perpendicular axis
through the center of the rectangle results a host of
mathematical challenges. The evaluation of the overlap-
ping area of the rotated rectangles, contrary to the square
plate case, is composed of three different configurations;
one such configuration is shown in Figure 8. The

Figur e 10. Display of two r otated ellipses. The bottom ellipse
is fastened to the xy coor dinate system; the top ellipse is
r otated counter clockwise by u r adian
other two configurations are shown in [4]. The character-
istics of the associated charging and discharging signals
for uniformly accelerated plates with dimensions L/W =
8/2 are shown in Figure 9. The changes in the character-
istic signals in comparison to the counterparts, Figure 6,
are drastic. Mathematical analysis and the detailed solu-
tion of the differential equations describing the intricacies
of the rectangular plates are discussed in length in [4].
The authors also have extended their investigations to
consider curvilinear plates, such as a pair of elliptical
plates. Figure 10 is borrowed from [5]. Mathematical and
detailed calculations resulting in the value of the overlap-
ping area of the plates are discussed in [5]. Reference [5]
also includes the characteristic charging and discharging
signals as well. Given the comparable dimensions of the
rectangles in [4] vs. the dimensions of the ellipses in [5]
we reasoned the similarity of their output signals. To
avoid the repetition of showing their similar transient
electric signals we omit displaying the latter. An inter-
ested reader is encouraged to review reference [5].
As indicated in the Introduction, the authors have pro-
posed a unique research project that has brought together
three different subject areas: Geometry, Mechanics, and
Electrical Network. Mathematica, with its flexible and
easy to use intricacies, is chosen as the ideal tool to ana-
lyze the project and address the what-if scenarios. As
pointed out in the text, some of the derived results are
intuitively just. And for the hard to predict cases, we ap-
plied Mathematica to analyze the problem. As an
open-ended question and research oriented project, one
may attempt to modify the presented analysis along with
the accompanied codes to investigate the response of
parallel RC circuits. It would also be complimentary to
our theoretical analysis to manufacture a rotating capaci-
tor to supplement the experimental data.
REFERENCES
[1] S. Wolfram, The mathematica book, 5th Ed., Cam-
bridge University Publication, 2003.
[2] D. Halliday, R. Resnick, and J. Walker, Fundamentals of
physics extended, 8
th
Ed, John Wiley and Sons, 2007; J.
D. Jackson, Classical Electrodynamics, 3
rd
Ed, John
Wiley and Sons, 1998.
[3] M. Trott, The mathematica guidebook for graphics,
Springer, 2004.
[4] H. Sarafian, Rotating rectangular parallel-plate capacitor
and a transient electric circuit, International Mathematica
Symposium, IMS, 2008.
[5] H. Sarafian, Rotating elliptical parallel-plate capacitor
and a transient electric circuit, International Conference
on Computational Science and its Application, ICCSA08,
pp. 291296, Springer 2008.

doi: 10.4236/jemaa.2009.13023 Published Online September 2009 (www.SciRP.org/journal/jemaa)
145
Delta Modulation with PI Contr oller A
Compar ative Study
A. I. MASWOOD, S. ANJUM

School of Electrical & Electronic Engineering, Nanyang Technological University, Nanyang Avenue, Singapore.
Email: Eamaswood@ntu.edu.sg

Received March 7
th
, 2009; revised May 25
th
, 2009; accepted June 12
th
, 2009.
ABSTRACT
The paper discusses the application of PWM delta modulation with PI controller as an alternative to the standard PWM
techniques for providing gating signals to the voltage source inverters. Its inherent characteristics of constant
volts/Hertz control without feedback complexity and boosting of the fundamental voltage makes it an excellent choice
for motor drive applications. This paper discusses the comparison between a basic delta modulation, the advanced
Delta modulation with PI controller and an optimized DMPI. It is shown that in addition to the aforementioned advan-
tages, VSI with optimized DMPI produces superior load current/voltage waveforms compared to simple DMPI, when
IGBT is employed as a switching device for the inverter. Since the tacho feedback or other traditional speed sensing
means is not permissible in sealed motor or the pump, a novel method is used to monitor the motor speed from the ter-
minal quantities like voltage, current, and motor input power factor.
Keywords: Delta Modulation, PI Controller, Voltage Source Inverter, Constant v/f Control
1. Intr oduction
The fundamental aim of any PWM switching is to elimi-
nate the lower order harmonics at relatively minimal
commutation of inverter switches. Attraction of the delta
modulation (DM) technique is that it guarantees that the
on and off time of the inverter switches will never fall
below a given minimum value. DM technique is an es-
tablished alternative to the traditional sinusoidal PWM
switching used in Commercial voltage source inverters
driving AC motors. Previous work on DM [14] has dis-
cussed the objectives, qualities and advantages of delta
modulation (DM) technique already from a very specific
point of view.
This work gives an overview of the DM from its basic
version to the latest state of the art form, i.e. DM with
built-in proportional Integral controller.
The adaptive or the rectangular DM has the inherent
ability to track the reference signal within a well defined
hysteresis threshold level and provides the desired V/f
characteristic. The optimal switching frequency and the
harmonic control can be easily achieved by:
Adjusting the integrator R, C parameters.
Changing the reference signal VR amplitude.
Controlling the hysteresis threshold level.
Changing the time constant
In the circuit of Figure 1, if the reference Sine wave
signal V
R
is maintained at the same level of modulation,
the ratio of the fundamental voltage of the modulated
wave to frequency remains constant at all frequencies.
This is true for PWM mode of operation. The phenome-
non is only valid up to the base frequency, which is the
fundamental frequency of the reference wave VR. To
investigate these aforementioned important properties of
the delta modulator, initially a basic DM circuit in PSIM
simulation environment is developed.

Figur e 1. Basic delta modulator
Delta Modulation with PI ControllerA Comparative Study 146
V/f Characteristics
In delta modulation, finding the base frequency i.e. fre-
quency up to which the DM retains constant V/f charac-
teristic, is quite important. Above the base frequency the
inverter enters constant voltage region (CVR). The CVR
may lead to the saturation of the motors magnetic core
and is to be avoided [2]. Figure 2(a) shows the V/f rela-
tionship of fundamental voltage versus the operating fre-
quency of the input source V
r
.
2. Per for mance Par ameter s of DM
Various performances of DM are discussed by:
1) Changing the time constants of DM
2) Changing the hysteresis gap of the DM
3) Changing voltage amplitude of the input source Vr.
4) Using PI controller with DM.
2.1 Changing the Time Constant ( )
In delta modulation circuit shown above the positive and
negative slops of carrier wave V2 depends on the time
constant =R3C of inverter circuit, using some variations
in the time constant by changing R3, different ripple fre-
quency of the carrier waveform can be achieved. Fourier
analysis for the harmonic variation of V1 with the
changes in time constant is carried out for the various
values of operating frequencies of the source VR. The
results are shown in Figures 2 and 3.
It has been observed, that by making time constant of
inverter circuit to /3 by changing R3 to 27k, the linearity
stage of V/f characteristic is extended to 300Hz. This is
three times the standard base frequency which is 100Hz,
and the fundamental voltage amplitude is reduced and
also there is change in harmonic amplitude of V1.
Using a time constant of 3 (R3=204k) harmonic am-
plitude of switching output voltage of DM (V1) is re-
duced, however the base frequency has been reduced to
50 Hz, this is two times smaller than base fre-
quency(100Hz).
Therefore, it is found from DM operation that, the base
frequency is inversely proportional to the inverter time
constant . A range of R3 value from 27k to 68k is attrac-
tive for optimal performance of the DM circuit.
2.2 Changing the Hyster esis Gap of Delta
Modulator
The hysteresis gap can be changed by changing feed
forward resistor R2 having VR and the as constants.
Thus by using hysteresis 5 % ,10% and 20 % by changing
resistor R2, Fourier analysis for the harmonic variation of
V1 was carried out for various frequencies of VR.
V/f characteristic with variation of hysteresis gap from 5%
to 20% is shown in Figure 4. The actual harmonic
0
0.2
0.4
0.6
0.8
1
1.2
10 20 30 40 50 60 70 80 90 100 110 120 130
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E

(a)

0
0.2
0.4
0.6
0.8
1
1.2
10 20 30 40 50 70 90 110
frequency
n
o
r
m
a
l
i
s
e
d

v
o
l
t
a
g
e

(b)

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
2
0
3
0
4
0
5
0
6
0
7
0
8
0
9
0
1
0
0
1
1
0
1
2
0
1
3
0
1
4
0
1
5
0
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E

(c)
Figur e 2. (a) Nor malized voltage ver ses fr equency distr ibu-
tion with time constant , (b) V/f char acter istic with time
constant 3, (c) V/f Char acter istic with time constant /3

spectrum is not shown as the aim of Delta modulation is
to boost the fundamental output voltage (V), which is
shown in its normalized form.
It is found from the DM performance that, a lower hys-
teresis gap provides for a wider base frequency, but the
changes in the base frequency are minimal for the various
values of hysteresis gap. R2 range of 100k to 200k is
optimal for optimum performance of the DM.
0
0.2
0.4
0.6
0.8
1
1.2
1
0
3
0
5
0
7
0
9
0
1
1
0
1
3
0
2
0
0
3
0
0
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E

Figur e 3. Fundamental component of nor malized voltage
ver ses fr equency for differ ent time constant

HYSTERESIS COMPARISON
0
0.2
0.4
0.6
0.8
1
1.2
20 30 40 50 60 70 80 90 100 110 120 130
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E
10%hysteresis 5% hysteresis 20% hysteresis

Figur e 4. Effect of Constant V/f char acter istic with var ia-
tion of Hyster esis gap, 5%, 10%, and 20%

0
0.2
0.4
0.6
0.8
1
1.2
10 20 30 40 50 60 70 80 90 100 110 120
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E

Figur e 5. Nor malized fundamental voltage vs. fr equency
char acter istics for var iation of VR fr om 2V to 8V
2.3 Changing the Refer ence Signal Amplitude
(VR)
Under this mode the time constant and the hysteresis gap
are kept constant. In normal DM operation, VR is held
constant. However, if in some applications the base fre-
quency needs to be varied without changing the time
constant and the hysteresis gap, it can be achieved by
changing VR only. It is found that VR is inversely related
to the base-frequency fb. This is shown in Figure 4. A
range of 3 to 5 volts for the input source VR is suitable
for optimum performance of DM.
3. Pr oper ties of DM with PI Contr ol
A PI controller is used to boost up the fundamental volt-
age of the delta modulator [6,8].The PI controller is inte
grated to the delta modulator (DMPI) as shown in Figure
6(a). Fourier analysis was done for three different values
of the feedback resister Rf. The respective fundamental
output voltages are shown in Figure 6(b).
As observed from the test that, PI controller helps to
boost the fundamental voltage widening the gap between
the fundamental and harmonic voltages, and it is also
observed that the PI introduces a disturbance to the line-
arity of the V/f characteristics.
The low voltage percentage boost is calculated from
the above figures for each value of Rf based on the gen-
eral DM circuit in Figure 1 & Figure 6(a) means Rf value
equals to infinity. It is clear from the above tabulated

Figur e 6(a). Cir cuit schematic of Delta-modulator with PI
contr oller

FUNDAMENTAL VOLTAGE AT VARIOUS " Rf" VALUES
0
0.2
0.4
0.6
0.8
1
1.2
20 30 40 50 60 70 80 90 100 110 120
FREQUENCY
N
O
R
M
A
L
I
S
E
D

V
O
L
T
A
G
E
FOR Rf=68K FOR Rf=100K FOR Rf=150K

Figur e 6(b). A compar ison of fundamental nor malized volt-
age for differ ent values of the Rf
Table 1. Low voltage boosting thr ough R
f

Rf ()
Fundamental
Value at fm
50Hz
Low Voltage
Boost
Infinity 6.5370 -
150K 7.009 7.220%
100K 7.4114 13.376%
68K 8.1785 25.1109%
45K (optimal) 9.5 35.12

results that with proper selection of Rf, the fundamental
voltage amplitude can be boost up to 35.10% as long as
the operating frequency is within the base frequency. A
range of 45k to 100k is suitable for Rf value for the op-
timum performance of DMPI.
3.1 Optimized Delta Modulator with PI Contr ol
A final and optimum performance based Delta Modulator
with PI controller is suggested for inverter (VSI) with the
following optimal values: R2=125K, R3=45K, Rf=45k.
The harmonic spectrum of switching output voltage of
DM (V1) is obtained with suggested optimized DMPI
with modulating frequency of 50Hz and 100Hz are
shown in Figure 7.

(a)

(b)
Figur e 7. Har monic spectr um of switching output voltage of
DM i.e. V1 at modulating fr equency (a) 50Hz (b) 100Hz
As one can see in the above figures, the optimized
DMPI with modified component values shifts the domi-
nant harmonics towards higher frequencies preferably
from the 11th harmonic position. This optimized DMPI
model is now used as a driver circuit for the voltage
source inverter.
4. Delta Modulator as the Dr iver
We have selected the above discussed delta modulation
(basic DM, DM with PI) techniques for generating the
switching gate signal for the IGBTs of inverter. Analysis
of the performance of proposed rectifier-inverter topol-
ogy [4] carried out for resistive-inductive load.
4.1 Per for mance of Inver ter with Basic DM
Rectified DC supply is applied given to an IGBT inverter
with the basic delta modulator and it is tested for a RL
load(R=18Ohms, L=95mH) at 50Hz and 100Hz operating
Frequencies (see in Figures 8 and 9).
4.2 Per for mance of Inver ter with DMPI
Study is conducted with the delta modulator inverted
with PI controller, and it is observed for same resistive
inductive load at 50 and 100Hz operating frequencies as
shown in Figures 10 and 11.
It can be seen in the Figure 10(b), there is an 11%
boost in the amplitude of fundamental component of
output phase voltage of inverter for 50 Hz operation of
DMPI with respect to basic DM characteristics shown in
Figure 8(b).
4.3 Per for mance Compar ison of Inver ter with
Basic DM and with DMPI
The output phase voltage of the inverters has been stud-
ied for both the delta modulation (DM and DMPI) tech-
niques for a range of frequencies with R-L load. The
normalized fundamental voltage versus frequency char-
acteristics is plotted and it is shown below. It is seen from
the Figure 12, that introduction of PI controller helps to
boost up the fundamental voltage compared to that of
basic DM. The PI controller also introduces slight dis-
turbance to the V/f characteristic but the linearity is still
maintained up to the range of base frequency of 100 Hz.
5. Optimized DMPI with IGBT Inver ter
Rectifier-inverter performance is studied with the opti-
mized DMPI, which we have discussed earlier. Opti-
mized DMPI is used to drive the driver circuit for the
overall three phase inverters and the output harmonic
spectrum generated by the inverter circuit is shown below.
It is observed that improved version of DMPI is able to
shift the more dominant harmonics to the higher frequen-
cies.

(a) (b)
Figur e 8. (a) Phase voltage wavefor m with basic DM at 50 Hz. (b) Phase voltage fr equency spectr um at 50 Hz

(a) (b)
Figur e 9. (a) Phase voltage wavefor m with basic DM at 100 Hz. (b) Phase voltage fr equency spectr um at 100 Hz

(a) (b)
Figur e 10. (a) Phase voltage wavefor ms with DM with PI, at 50 Hz. (b) Phase voltage fr equency spectr um at 50 Hz

(a) (b)
Figur e 11. (a) Phase voltage wavefor m with DM with PI, at 100 Hz. (b) Phase voltage fr equency spectr um at 100 Hz
5.1 Experimental Results & Drive Application
Figure 14 shows the time domain sample of DM output
waveform captured from oscilloscope (Fluke spectrum
analyzer). The voltage is normalized with respect to the
inverter input dc bus voltage. The spectrum shows the
absence of the low frequency harmonics and a high
(96.8%) fundamental DM output voltage. Traces of 3
rd

harmonic still is present due to imperfection in gating
signal timing. The output characteristics of the experi-
mental prototype Delta-Modulator is studied at different
modulating signal Vr frequency and magnitude levels in
order to achieve maximum inverter fundamental output
voltage. The results are compared and confirmed with the
ones from simulation.
In submersible motor pump application, the DMPI
method uses either the voltage or the current variation to
maintain the constant slip operation. This is known as the
speed variation and is used because the pump production
and the efficiency are dependent on the speed of the
pump. The operational characteristic of the overall sys-
tem is shown in Figure 15.
The experimental motor employed is a 1.5 HP squirrel
cage submersible induction motor. The motor parameters
are given in the Appendix. The variation of inverter
voltage, frequency and the resulting motor slip are ob-
tained for a sudden change in load, as shown in Figure 15.
When the motor load is increased, the slip increases re-
sulting in a slowdown. The inverter responds by increas-
ing the terminal voltage and frequency. The motor slip
eventually decreases and settles to the steady state value
after few oscillations. These characteristics also reflect
the constant v/f operation of the DM inverter.
6. Conclusions
From the above study of DM techniques, and various
comparisons between them, it has been concluded that
delta modulation with PI controller helps to provide con-
stant V/F characteristics and boosting of the fundamental
voltage especially at low operating frequency. The opti-
FUNDAMENTAL HARMONIC COMPARISON BETWEEN
BASIC DM AND DMPI
0
50
100
150
200
250
1
0
2
0
3
0
4
0
5
0
6
0
7
0
8
0
9
0
1
0
0
1
1
0
1
2
0
1
3
0
1
4
0
1
5
0
FREQUENCY
F
U
N
D
A
M
E
N
T
A
L

V
O
L
T
A
G
E
DMPI BASIC DM

Figur e 12. The nor malized fundamental voltage ver sus fr e-
quency char acter istics
Figur e 13. Har monic spectr um for VSI output voltage using
optimized DM with PI

Figur e 14. Inver ter output wavefor m fr om oscilloscope

Figur e 15. Change of slip, fr equency and fundamental
voltage for an incr ease in load at 50Hz oper ation
mized DM brings about a significant decrease in the
harmonic contents towards the 13th harmonic. The DMPI
inverter with higher voltage boosting capability and con-
stant V/f characteristic without the feed-back complexity
leads to lighter and more economical inverter and the
optimized DM inverter with attenuation of low-order
harmonics lead to reduced filter size.
It was observed that the motor torque and speed can be
instantaneously computed solely from its terminal elec-
Delta Modulation with PI ControllerA Comparative Study
151
trical parameters. It was shown how the DMPI does
eliminate the problems existing in SMM. The results of
the experimental setup show that the proposed method
handles load change overshoots and oscillations with ease,
especially where frequency variation with load change is
not warranted.
REFERENCES
[1] P. D. Ziogas, The delta modulation techniques in static
PWM inverters, IEEE Transactions on Ind. Applications,
pp.19904, Mar/Apr. 1981.
[2] M. A. Rahman, J. E. Quacioe, and M. A. Chowdhury,
Performance analysis of delta modulated PWM, IEEE
Transactions in Power Electronics, Vol. PE2, No. 3, pp.
227233, July 1987.
[3] T. C. Green, J. C. Salmon and B. W. Willams, Investiga-
tion of delta modulation and of subharmonic elimination
techniques, IEEE PESC Record, Vol. 1, April 1988.
[4] M. A. Rahman, J. E. Quacioe, and M. A. Chowdhury,
Harmonic minimization in Delta modulated inverters
using Tuned filters, IEEE PESC Record, Vol. 1, April
1988.
[5] A. I. Maswood, P. D. Ziogas, and G. Joos, Problems and
solutions associated with the operation of phase controlled
rectifiers under unbalanced input voltage conditions,
IEEE Transactions on Industry Applications, U.S.A, Vol.
27, No. 4, pp. 765772, July/August 1991.
[6] A. I. Maswood and M. A. Rahman, A survey of delta
modulation techniques, characteristics & sub harmonic
elimination for VSI, Electric Machines and Power Sys-
tems Journal, U.S.A., Vol. 26, No. 6, pp. 435448, July
1998.
[7] A. I. Maswood and S. Wei, A novel current source
PWM drive topology with specific harmonic elimination
switching patterns, IEEE Canadian Conference on Elec-
trical & Computer engineering, CCECE, Halifax, Nova
Scotia, May 810, 2000.
[8] A. I. Maswood and M. A. Rahman, A PWM voltage
source inverter with PI controller, performance parame-
ters under non-ideal conditions, published in the Electric
Power Systems Research [EPSR] Journal, U.S.A, Vol. 38,
No. 1, pp 1924, 1996.
[9] A. I. Maswood and M. H. Rashid, A novel method of
harmonic assessment generated by 3Phase AC-DC con-
verters under unbalanced supply conditions, IEEE Trans-
actions on Industry Applications, U.S.A, Vol. 24, No. 4,
pp. 590597, July/August 1988.
[10] M. H. Kheraluwala and D. M. Divan, Delta modulation
strategies for resonant link inverters, IEEE Transactions
on Power Electronics, Vol. 5, No. 2, pp. 220227, April
2000.

Ener gy Compar ison of Seven MPPT Techniques
for PV Systems
A. DOLARA, R. FARANDA, S. LEVA

Department of Energy of Politecnico di Milano, Via la Masa 34, 20156, Milano, Italy.
Email: alberto.dolara@mail.polimi.it, {roberto.faranda, sonia.leva}@polimi.it

Received May 14
th
rd
, 2009; accepted July 12
th
, 2009.
ABSTRACT
In the future, solar energy will be a very important energy source. Several studies suppose that more than 45% of the
energy in the world will be generated by photovoltaic array. Therefore it is necessary to concentrate our forces to re-
duce the application costs and to increment their performance. In order to reach the last aspect, it is important to note
that the output characteristic of a photovoltaic array is nonlinear and changes with solar irradiation and cells tem-
perature. Therefore a Maximum Power Point Tracking (MPPT) technique is needed to maximize the produced energy.
This paper presents a comparative study of seven widely-adopted MPPT algorithms; their performance is evaluated
using, for all the techniques, a common device with minimum hardware variations. In particular, this study compares
the behaviors of each technique in presence of solar irradiation variations.
Keywords: Maximum Power Point Tracking (MPPT), Photovoltaic (PV), PV Performance Comparison, Renewable
Energy, DC-DC Converter.
1. Intr oduction
Solar energy is one of the most important renewable en-
ergy sources. As opposed to the conventional not renew-
able sources such as gasoline, coal, etc. solar energy is
clean, inexhaustible and free. The main applications of
photovoltaic (PV) systems are in either stand-alone (wa-
ter pumping, domestic and street lighting, electric vehi-
cles, military and space applications) [1] or grid-con-
nected configurations (hybrid systems, power plants) [2].
Unfortunately, PV generation systems have two major
problems: the conversion efficiency in electric power
generation is low (in general less than 17%, especially
under low irradiation conditions), and the amount of
electric power generated by solar arrays changes con-
tinuously with weather conditions.
Moreover, the solar cell V-I characteristic is nonlinear
and changes with irradiation and temperature. In general,
there is a point on the V-I or V-P curve only, called the
Maximum Power Point (MPP), at which the entire PV
system (array, inverter, etc.) operates with maximum
efficiency and produces its maximum output power. The
location of the MPP is not known, but can be located,
either through calculation models or by search algorithms.
Maximum Power Point Tracking (MPPT) techniques are
used to maintain the PV arrays operating point at its
MPP.
Many MPPT techniques have been proposed in the lit-
erature; examples are the Perturb and Observe (P&O)
method [25], the Incremental Conductance (IC) method
[26], the Artificial Neural Network method [7], the
Fuzzy Logic method [8], etc.. The P&O and IC techniques,
as well as variants thereof, are the most widely used.
Because of the large number of methods for MPPT, in
the last years researchers and practitioners in PV systems
have presented survey or comparative analysis of MPPT
techniques. As a matter of fact, some papers present
comparative study among only few methods [5,6] and
one paper presents a survey and a discussion of several
MPPT methods [10]. Another paper [11] presents a
ranking of ten widely adopted MPPT algorithms (P&O,
modified P&O, Three Point Weight Comparison [12],
Constant Voltage, IC, IC and CV combined [13], Short
Current Pulse [14], Open Circuit Voltage [15], the Tem-
perature Method and methods derived from it [16]),
based on simulations, under the energy production point
of view. The MPPT techniques are evaluated considering
different types of insolation and solar irradiance varia-
tions and calculating the energy supplied by a complete
PV array.
In this paper, the attention will be focused on experi-
mental comparisons between some of these techniques,
considering several irradiation conditions. Therefore, the
Energy Comparison of Seven MPPT Techniques for PV Systems 153

Figur e 1. Block diagr am of the whole exper imental system

aim of this work is to compare several widely adopted
MPPT algorithms between them in order to understand
which technique has the best performance. The evalua-
tion of the algorithms performance is based on the power
measurement valuating the total energy produced by the
panel during the same test cycle. In this work, respect to
the MPPT algorithm compared by simulations, the
methods that need temperature or irradiance measure-
ments are not considered for sake of simplicity. Indeed,
as described in [11], these techniques do not have very
high performance and they are too expensive. In the
simulations, the considered MPPT techniques have been
implemented strictly following the description indicated
in the references: no MPPT algorithm is preferred and no
MPPT techniques have been realized with more attention
respect to the others.
In particular, without lack of generality, we will focus
our attention on a stand-alone photovoltaic system con-
structed by connecting the DC-DC converter between the
solar panel and a dc load.
2. Exper imental System
The experimental comparison among the different MPPT
techniques has been performed realizing the whole sys-
tem in the Power Quality Laboratory of Department of
Energy at the Politecnico di Milano.
The experimental system is constituted by three main
elements (Figure 1): the DC-DC converter, the PV-panels
and the solar simulator.
2.1 The DC-DC Conver ter
It has been realized a single device constituted by a
DC-DC converter [17] and other components able to im-
plement all the different MPPT techniques here analyzed,
including Open Circuit Voltage (OV) [14] and Short
Current Pulse (SC) [13] which required to insert further
static switches to open the circuit or to create the
short-circuit condition, in order to compare the results.
All the MPPT techniques here described are easily ob-
tained changing the software compiled in the microcon-
troller. In this way the differences in the measured energy
load depend mainly on the software used for the imple-
mentation of the particular MPPT technique.
The choice of a stand-alone system, and hence the
choice of using a DC-DC converter, reflects some indus-
trial configurations composed by a first DC-DC conver-
sion stage, in which usually the control of MPPT tech-
niques is implemented, a second filter stage, and eventu-
ally a DC-AC conversion stage.
The DC-DC converter developed includes the power
and control boards as shown in Figure 2.
The control board is constituted by all the components
that need for the implementation of the various MPPT
algorithms already illustrated in [1016]. The microcon-
troller, in this case a Microchip dsPIC30f4012, is the core
of the control board.
The command connection to the power board is pro-
vided by means of driver circuits which allow the valves
commutation.
The interface between control and power circuits is re-
alized with optoinsulators and Hall effect transducers to

Figur e 2. (a) Stand-alone PV system analyzed. DC-DC con-
ver ter s (b) power and (c) contr ol boar ds
guarantee the necessary metallic insulation required be-
tween these boards. Such connection allows not only to
drive the valve in PWM mode and hence to implement
the different MPPT techniques without modifying the
power components, but also to acquire the PV voltage
and current signals.
In particular, the voltage and current measurements are
made by Hall effect transducers; they are perfectly suit-
able for this application indeed they are able to detect
continuous components, furthermore they can guarantee
very low losses during the measurement and insulation
between the control board and the power one, and finally
they have a wide enough bandwidth.
There are a lot of DC-DC conversion circuits. In the
present work the boost configuration is chosen. It is very
spread thanks to its high reliability respect to other more
complex configurations, to the reduced number of com-
ponents and also to the high-minded experience in its
operation. The complete power device scheme is shown
in Figure 3.
The boost section is realized by two accumulation
units, L and C
out
, the static switch T
1
and the diode D
3
.
Moreover, the diode D
1
is put into the circuit to protect
the PV-panel against negative current which could dam-
age it.
The equivalent measures of the PV-panel voltage, V
PV
,
and current, I
PV
, are obtained by inserting the voltage
transducer V and the current one A in the circuit as re-
ported in Figure 3.
Figure 3 shows the circuit elements T
v0
, T
sc
, K
1
, K
2
,
C
in
and D
2
; that have been inserted to:
measure the PV-panel open circuit voltage, that is
necessary in OV technique, through the opening
of T
v0
valve, in this case D
2
is short-circuited
through K
2
;
measure the PV-panel short-circuit current, that is
necessary in the SC technique, through the closure
of T
sc
valve, in this case T
v0
is short-circuited
through K
1
.
During the tests of other MPPT techniques, the valve
T
sc
is kept open, while T
v0
and D
2
are short-circuited, re-
spectively through K
1
and K
2
switches, to increase con-
verter efficiency removing their power losses. It is impor-

Figur e 3. Scheme of the power device
tant to underline that, in each MPPT algorithms, the
DC-DC converter power losses do not influence the MPP
because the system acquires the PV voltage and current.
It is important to note that in the SC MPPT technique it
is necessary to insert the D
2
diode to avoid, during the
short-circuit test, the discharging of C
in
placed at boost
input. Such capacitor is always inserted in each tech-
niques analysed to limit the high frequency harmonic
components.
The prototype converter has been sized for the voltage
of 3 in-series modules and the current of 3 in-parallel
modules. In particular, in correspondence of the Standard
Test Condition (STC), therefore at 1000 W/m
2
and 298 K,
we have:
a maximum open circuit voltage equal to 21.8 V
and a maximum short-circuit current equal to
13.05 A with the modules in parallel configura-
tion;
a maximum open circuit voltage equal to 65.4 V
and a maximum short-circuit current equal to
4.35 A with the modules in series configuration.
The DC-DC converter is designed to work at the MPP
with a duty cycle of 25%. The DC-DC converter sizing,
with a security margin, leads to the following data:
switching frequency of 20 kHz, nominal current of 15 A,
and nominal voltage of 150 V.
The IGBT IRG4PC30KD electronic valves are chosen.
These components integrates an ultrafast recycling diode
and present small switching losses also in presence of
high switching frequency.
2.2 PV Panel
The PV panels here considered are the poly-crystalline
70 W PV-module by Helios Technology. Its main speci-
fications are shown in Table 1.
2.3 Solar Simulator
The sunlight simulator have to guarantee low spatial
non-uniformity and low temporal instability of irradiance,
moreover it have to generate a significant power output
from PV-system and finally it have to allow different
irradiance levels on the PV-panel.
The solar simulator used in the present tests is realized by
using both incandescent and halogen lamps. The maxi-
mum power of the solar simulator is 2.8 kW and its size
is 1200 mm long and 600 mm wide.
Combining the lamps, it is possible to have, with ade-
quate uniformity, four different irradiation levels equal to
0 W/m
2
, 272 W/m
2
, 441 W/m
2
and 587 W/m
2
.
3. MPPT Contr ol Algor ithm
There are many MPPT methods available in the literature;
the most widely-used techniques are described in
Table 1. Electr ical char acter istics of PV panel in STC
Symbol Quantity Value
P
MPP
Maximum Power 70 W
V
MPP
Voltage at P
MPP
17 V
I
MPP
Voltage at I
MPP
4.11 A
I
SC
Short-Circuit Current 4.35 A
V
OV
Open-Circuit Voltage 21.8 V
NOCT Nominal Operating Cell Temperature 432 C

the following sections, starting from the simplest method.
3.1 Constant Voltage Method
The Constant Voltage (CV) algorithm is the simplest
MPPT control method. The operating point of the PV
array is, each n
th
step, kept near the MPP by regulating
the array voltage and matching it to a fixed reference
voltage V
REF
equal to the V
MPP
of the characteristic PV
module (see Table 1) or another pre-valuated best voltage
value [16]. The CV method assumes that insulation and
temperature variations on the array are insignificant on the
V
MPP
, and that the constant reference voltage V
REF
is an
adequate approximation of the real MPP. Therefore, the
operating point is never exactly at the MPP and different
data have to be adopted for different geographical regions.
The CV method needs the measurement of the PV ar-
ray voltage V
PV
in order to set up the duty-cycle of the
DC/DC boost converter as shown in Figure 4.
It is important to observe that when the PV panel is in
low insulation conditions, the CV technique is, generally,
more effective than either the P&O method or the IC
method (analyzed below), as shown in [11]. Thanks to
this characteristic, the CV method is often combined to-
gether with other MPPT techniques.
Measurement of V
PV
V
PV
= V
REF
YES
YES
V
PV
> V
REF
(n+1) = (n) + (n+1) = (n) -
NO
NO

Figur e 4. Flow char t of the CV method
3.2 Shor t-Cur r ent Pulse Method
The Short-Current Pulse (SC) method achieves the MPP
by giving a reference current I
REF
to the power converter
controller. In fact, the optimum operating current for
maximum output power is proportional to the sho-
rt-circuit current I
SC
under various conditions of irradi-
ance level S as follows [12]:

REF 1 SC
I S k I S (1)
where k
1
is a proportional constant.
This control algorithm requires the measurement of
the current I
SC
. To obtain this measurement, it is neces-
sary to introduce a static switch in parallel with the PV
array, in order to create the short-circuit condition. It is
important to note that when V
PV
=0 no power is supplied by
the PV system and consequently no energy is generated.
The SC method needs the measurement of the PV ar-
ray current I
PV
in order to set up the duty-cycle of the
DC/DC boost converter (see Figure ).
3.3 Open Voltage Method
The Open Voltage (OV) method is based on the observa-
tion that the voltage V
MPP
is always close to a fixed per-
centage of the open-circuit voltage V
OV
. Production
spread, temperature, and solar insulation levels change
the position of the MPP within a 2% tolerance band. This
technique uses 76% of V
OV
as reference value V
REF
(at
which the maximum output power can be obtained); in
general, this value is very close to the V
MPP
.
This control algorithm requires measurements of the
voltage V
OV
when the circuit is opened. Here again it is
necessary to introduce a static switch into the PV system;
for the OV method the switch must be used to open the
circuit. When I
PV
=0 no power is supplied by the PV sys-
tem and consequently no energy is generated. Also in this
method measurement of the PV array voltage V
PV
is re-
quired by the regulator (see Figure 6).
3.4 Per tur b and Obser ve Methods
The P&O algorithms operate by periodically perturbing
(i.e. incrementing or decrementing) the array terminal
voltage and comparing the PV output power with that of
the previous perturbation cycle. If the PV array operating
Measurement of
I
PV
NO
NO
PV short-circuit
condition
PV work
condition
I
PV
< I
REF
(n+1) = (n) - (n+1) = (n) +
I
PV
= I
REF
Refresh
reference?
Measurement of
I
SC
I
REF
= k
1
I
SC
NO YES
YES
YES

Figur e 5. Flow char t of the SC method

Measurement of
V
PV
NO
NO
PV short-circuit
condition
PV work
condition
V
PV
> V
REF
(n+1) = (n) - (n+1) = (n) +
V
PV
= V
REF
Refresh
reference?
Measurement of
V
OV
V
REF
= k
2
V
OV
NO YES
YES
YES

Figur e 6. Flow char t of the OV method

voltage changes and power increases (dP/dV
PV
>0), the
control system moves the PV array operating point in that
direction; otherwise the operating point is moved in the
opposite direction. In the next perturbation cycle the al-
gorithm continues in the same way.
A common problem in P&O algorithms is that the ar-
ray terminal voltage is perturbed every MPPT cycle;
therefore when the MPP is reached, the output power
oscillates around the maximum, reducing the generable
power by the PV system. This is mainly true in constant
or slowly-varying atmospheric conditions but also under
rapidly changing atmospheric conditions [17].
There are many different P&O methods available in
the literature. In this paper we consider the classic, the
optimized and the three-points P&O algorithms.
In the classic P&O technique (P&Oa), the perturba-

1
0
1
PV PV
P n P n
n n

Figur e 7. Flow char t of the P&Oa method

tions of the PV operating point have a fixed magnitude
(see Figure 7). In our analysis, the magnitude of pertur-
bation is 0.35% of PV array V
OV
. In the optimized
P&O technique (P&Ob), an average of several samples of
the array power is used to dynamically adjust the magni-
tude of the perturbation of the PV operating point. This
adjusting is realized by valuating M
OV
and calculating the
parameter a(n) as reported in Figure 8. In the three-point
weight comparison method (P&Oc), the perturbation
direction is decided by comparing the PV output power
on three points of the P-V curve, and valuating the pa-
rameter M as shown in Figure 9. These three points are
the current operation point A, a point B perturbed from
point A, and a point C doubly perturbed in the opposite
direction from point B.
All these three algorithms require the measurement of
the PV array voltage V
PV
and of the PV array current I
PV
.
3.5 Incr emental Conductance Method
The Incremental Conductance (IC) algorithm is based on
the observation that the following equation holds at the
MPP [2]:
(dI
PV
/dV
PV
)+(I
PV
/V
PV
)=0 (2)
where I
PV
and V
PV
are the PV array current and voltage,
respectively. When the operating point in the P-V plane is
to the right of the MPP, it is verified (dI
PV
/dV
PV
)+
(I
PV
/V
PV
)<0, whereas when it is to the left of the MPP this
(dI
PV
/dV
PV
)+(I
PV
/V
PV
)>0.
The MPP can thus be tracked by comparing the in-
stantaneous conductance I
PV
/V
PV
to the incremental con-
ductance dI
PV
/dV
PV
. Therefore, if the quantity (dI
PV
/dV
PV
)
+(I
PV
/V
PV
) is more than , its sign means a power produc-
tion decrement and indicates the correct direction of per-
turbation leading to the MPP. Once MPP has been
reached, the operation of PV array is maintained at this
point and the perturbation stopped unless a change in dI
PV

is noted. In this case, the algorithm decrements or incre-
ments the PV array voltage V
PV
to track a new MPP. The
increment size determines how fast the MPP is tracked.
The IC method offers good performance under rapidly
changing atmospheric conditions. The classic IC algo-
rithm requires the measurement of the PV array voltage
V
PV
and current I
PV
in order to determine the correct per-
turbation direction.
4. Numer ical Results
The measurements have been performed several times in
order to cut off deviations caused by interferences and/or
environmental factors in this system. The most important
environmental factor, that hardly influence the PV-panel
behaviour, is its temperature. In order to maintain the
PV-panel temperature equal in all tests and to preserve
this parameter into a little range during tests, all experi-
ments are made starting from the same PV-panels tem-
perature, and the duration of tests has been reduced as
short as possible avoiding overheating.
Due to energy absorbed from the network, available
space constraints and especially economic constraints
associated to the dimensions of the solar simulator, the test
campaign involved a single module described in Table 1.
In order to realize a precise analysis of the perform-
ance of the different MPPT techniques, they are experi-
mentally compared taking into account two different ir-
radiation diagrams. The first one, Case 1 (Figure 11), is
characterized by medium and medium-high irradiation
levels of 441 W/m
2
and 587 W/m
2
with a time of 180 s
and the second one, Case 2 (Figure 12), with low, lo-
w-medium, medium-high irradiation levels of 0 W/m
2
,
272 W/m
2
, 441 W/m
2
and 587 W/m
2
, with a time of 160 s
(Case 2 include a 10 s interval without irradiation).
Every MPPT technique analysis starts when the initial
steady state condition of each case is reached.
A couple of samples of voltage and current is available
every 10 ms, and P&Oa, P&Ob, IC, CV, OV and SC al-
gorithms can perform an iteration for each couple of val-
ues; only P&Oc needs 3 measurement of power instead
of 1.
Duty-cycle variation amplitude is 0.5% for all
techniques except P&Ob, in witch is proportional to
the ratio and it ranges between 0.5% and 2.7%. A
reduced duty-cycle variation value decreases the speed of
the algorithm dynamic behaviour but it increases the pre-
cision in reaching MPP.
/ dP dV
The P&Oa technique performs very well with low ra-
diance values: in this condition the P-V curve is very
smooth near the maximum. The P&Ob logic with vari-
able step is able to reduce steady state oscillations and, at
the same time, to provide higher response speeds at me-

Figur e 8. Flow char t of the optimized P&Ob method

Measurement of V
b
(n), I
b
(n);
P
b
(n) = V
b
(n) I
b
(n)
P
b
(n) P
a
(n)
M = M + 1
P
a
(n) > P
c
(n)
M= 2
M= - 2
M = M - 1
YES NO
M = M + 1 M = M - 1
YES NO
V(n+1) = V
a
(n)
(n+1) = (n)
V(n+1) = V
b
(n)
(n+1) = (n) +
V(n+1) = V
c
(n)
(n+1) = (n) - 2
YES NO
YES NO
(n)
b
= (n)
a
+
Measurement of V
a
(n), I
a
(n);
P
a
(n) = V
a
(n) I
a
(n)
Measurement of V
c
(n), I
c
(n);
P
c
(n) = V
c
(n) I
c
(n)
(n)
c
= (n)
b
2
(n)
a
= (n)
Refresh
reference?
M = 0

Figur e 9. Flow char t of the optimized P&Oc method
Energy Comparison of Seven MPPT Techniques for PV Systems
159

Figur e 10. Flow char t of the IC method

Figur e 11. Ir r adiation diagr am of Case 1 Figur e 12. Ir r adiation diagr am of Case 2

dium-high irradiance level with respect to the P&Oa ap-
proach with fixed . This technique is very slow in
reaching MPP when irradiance level is low because
is small. / dP dV
Case 2 shows the main disadvantage of the IC tech-
nique: for low radiance values the technique works on a
P-V curve with a derivative close to zero in a large inter-
val around the maximum value, therefore it is not able to
properly identify the MPP. It results in oscillations
around the MPP with a reduced output energy value.
The P&Oc technique compare the power of three dif-
ferent working points as described in Fig. 9. The algo-
rithm modifies the duty-cycle, in function of the obtained
results, to reach the MPP value as described in [12]. The
increment (or decrement) of the duty-cycle amplitude is
constant and the algorithm performs an iteration every
10 ms.
The CV technique is optimized for a single radiance
value; the performance of this technique is strongly re-
lated with the voltage set point. It provides satisfying
results, but they are not as good as the ones provided by
P&O and IC techniques.
MPPT Technique Case 1 Case 2
P&Oa

P&Ob

P&Oc

IC

CV

OV

SC

Figur e 13. Power gener ated by the PV ar r ay in the Case 1 and Case 2 by differ ent optimized MPPT methods (solid line) and
ideal (dot-dashed line) MPPT method
Further considerations may regard OV and SC tech-
niques. They require additional valves for, respectively,
the measurement of the PV open circuit voltage and of
the PV short-circuit current.
Concerning the OV technique, it refreshes the voltage
reference value every 3 s through the open voltage meas-
urement (for this measurement is necessary 10 ms with-
out power generation). The ratio of the open voltage and
the MPP voltage is not strictly constant with temperature,
and the technique can be optimized only for a single
temperature value. For this reason the converter per-
formance with OV technique is in general better than CV,
but is not as good as the ones provided by P&O and IC
techniques because of the voltage drop on T
v0
and the
necessary measurement time. It is important to underline
that the voltage drop on T
v0
is equal to about 1.5 V, which
implies a significant reduction of the load voltage, and
hence of the output power. This is a significant feature in
the evaluation of the converter efficiency.
Analogous considerations regard the SC technique. In
this case the voltage drop that reduces converters output
power is due to the D
2
valve and it is about 0.6 V. This
technique refreshes the reference current value once
again every 3 s through the short-circuit current meas-
urement (for this measurement is necessary 10 ms with-
out power generation). The voltage applied to the
PV-panel during the measurement step is the voltage
drop on T
sc
. In this condition the measured current can be
approximated to the real short-circuit current.
Figure 13 shows the power generated from the
PV-panel with the same converter configuration and dif-
ferent MPPT techniques in the two cases. The diagrams
also show the ideal power, obtained by using an ideal
MPPT technique that is equal to the maximum power that
the PV-panel can produce. These values are measured
directly on the PV-panel under test in STC. For this rea-
son MPPT ideal curve must be considered only as a
qualitative reference to compare tests results. It is im-
portant to observe that the uncertain in the PV voltage
and current measurements and the small difference in the
environmental conditions between each tests suggest that
the results cannot be a good reference to calculate the
efficiency of the single MPPT algorithm. In these condi-
tions, even an uncertain of 0.5% in the measurements
could produce an uncertain in the relative power losses
that could be more than 10%.
Table 2 summarizes the performances of the different
techniques in the two radiation cases and the differences
respect to the ideal MPPT algorithm.
5. Conclusions
This paper has presented a comparison among some of
the more diffused Maximum Power Point Tracking tech-
niques in relation to their energy performance. In par-
ticular, different types of solar insolation characterized by
low and medium irradiation level are considered, and the
energy supplied by a complete PV array is experimen-
tally evaluated. The whole systemincluding the
DC-DC converter and the lighting systemis arranged in
the Power Quality Laboratory of Department of Energy
of the Politecnico di Milano. The different MPPT tech-
niques have been implemented following the directions
indicates in the papers listed in the references; no one has
been preferred or better improved respect to the others.
The results show that the best MPPT technique is the
modified P&O (P&Ob). The logic turned out to be effec-
tive in both the situations here considered, providing al-
ways the highest efficiency. P&Ob technique shows its
limit in the response to the irradiance variation at low
irradiance level.
The IC technique has an efficiency lower than the
P&O techniques, but its response time is quite independ-
ent to the irradiation values and its efficiency increase
with the irradiance level. This technique can be a good
alternative to the P&O techniques in applications charac-
terized by high, fast and continuous radiance variations,
e.g. the PV applications in transportation.
The two techniques are also equivalent concerning the
costs and the software complexity; in particular both the
techniques require a microcontroller with medium/higher
performances than the ones required by other techniques,
due to the necessity of high computation capability.
Among the other hill climbing techniques, the P&Oa
method presents acceptable results: this algorithm can be
a good alternative to the two previous techniques. Instead
the P&Oc method, even if characterized by output energy
values analogous to the P&Oa, has a more complex algo-
rithm and a lower reactivity, with no benefit in terms of
performances. Furthermore, given the features required
by the controller, the P&Ob technique is better than the
P&Oc one.
The P&Oa technique requires a microcontroller which
has lower computational capability constraints with re-
spect to the best technique here considered. It is therefore
necessary to evaluate if the cost gap between the two
microcontrollers can justify the lower performances of
the technique.
It is necessary to underline that the maximum irradi-
ance level obtained from solar simulator is about half
than the real irradiance from the sun. In these conditions
the performance of IC are quite less than the P&O tech-
niques ones because the MPP in the PV power character-
istic has a derivative close to zero for a quite large volt-
age variation.
In the present analysis the CV, OV and SC techniques
turned out to be the worst ones. Their performances are
lower than the ones obtained with P&Ob techniques es-
pecially in case of conditions very different from the ra-
diance value in correspondence of which these techniques
have been modeled. Moreover, OV and SC techniques
Table 2. Ener gy gener ated as a function of MPPT technique and ir r adiance input
MPPT
Technique
Case 1 Case 2
Energy [J] Rank
Difference respect to
the ideal case
Energy [J] Rank
Difference respect
to the ideal case
Ideal 4493 - - 3298 - -
P&Oa 4282 2 -4,7% 3144 2 -4,7%
P&Ob 4346 1 -3,3% 3212 1 -2,6%
P&Oc 4278 3 -4,8% 3135 3 -4,9%
IC 4215 4 -6,2% 3117 4 -5,5%
CV 4201 5 -6,5% 3100 6 -6,0%
OV 4200 6 -6,5% 3104 5 -5,9%
SC 4088 7 -9,0% 2942 7 -10,8%

requires additional valves in the converter that decrease
its efficiency and the output power.
The CV technique is still a very simple logic which
provides a very good efficiency for radiance values
closed to 700 W/m
2
, with low costs. Hence, generally this
technique can be selected only if there is the necessity to
minimize the control system cost.
However the cost of a microcontroller currently low,
so that the implementation of the P&O type techniques is
anyway preferred.
REFERENCES
[1] J. Schaefer, Review of photovoltaic power plant per-
formance and economics, IEEE Trans. Energy Convers.,
EC-5, pp. 232238, 1990.
[2] N. Femia, D. Granozio, G. Petrone, G. Spaguuolo, and M.
Vitelli, Optimized one-cycle control in photovoltaic grid
connected applications, IEEE Trans. Aerosp. Electron.
Syst., Vol. 42, pp. 954972, 2006.
[3] W. Wu, N. Pongratananukul, W. Qiu, K. Rustom, T. Kas-
paris, and I. Batarseh, DSP-based multiple peack power
tracking for expandable power system, in Proc. APEC,
pp. 525530, 2003.
[4] C. Hua and C. Shen, Comparative study of peak power
tracking techniques for solar storage system, in Proc.
APEC, pp. 679685, 1998.
[5] D. P. Hohm and M. E. Ropp, Comparative study of
maximum power point tracking algorithms using an ex-
perimental, programmable, maximum power point track-
ing test bed, in Proc. Photovoltaic Specialist Conference,
pp. 16991702, 2000.
[6] K. H. Hussein, I. Muta, T. Hoshino, and M. Osakada,
Maximum power point tracking: an algorithm for rapidly
chancing atmospheric conditions, IEE Proc.-Gener.
Transm. Distrib., Vol. 142, pp. 5964, 1995.
[7] X. Sun, W. Wu, X. Li, and Q. Zhao, A research on
photovoltaic energy controlling system with maximum
power point tracking, in Power Conversion Conference,
pp. 822826, 2002.
[8] T. L. Kottas, Y. S. Boutalis, and A. D. Karlis, New
maximum power point tracker for PV arrays using fuzzy
controller in close cooperation with fuzzy cognitive net-
work, IEEE Trans. Energy Conv., Vol. 21, pp. 793803,
2006.
[9] I. S. Kim, M. B. Kim, and M. Y. Youn, New maximum
power point tracking using sliding-mode observe for es-
timation of solar array current in the grid-connected
photovoltaic system, IEEE Trans. Ind. Electron., Vol. 53,
pp. 10271035, 2006.
[10] Y. T. Hsiao and C. H. Chen, Maximum power tracking
for photovoltaic power system, in Proc. Industry Appli-
cation Conference, pp. 10351040, 2002.
[11] G. J. Yu, Y. S. Jung, J. Y. Choi, I. Choy, J. H. Song, and
G. S. Kim, A novel two-mode MPPT control algorithm
based on comparative study of existing algorithms, in
Proc. Photovoltaic Specialists Conference, pp. 15311534,
2002.
[12] T. Noguchi, S. Togashi, and R. Nakamoto, Short-current
pulse-based maximum-power-point tracking method for
multiple photovoltaic-and-converter module system,
IEEE Trans. Ind. Electron., Vol. 49, pp. 217223, 2002.
[13] D. Y. Lee, H. J. Noh, D. S. Hyun, and I. Choy, An im-
proved MPPT converter using current compensation
method for small scaled PV-applications, in Proc. APEC,
pp. 540545, 2003.
[14] M. Park and I. K. Yu, A study on optimal voltage for
MPPT obtained by surface temperature of solar cell, in
Proc. IECON, pp. 20402045, 2004.
[15] T. Takashima, T. Tanaka, M. Amano, and Y. Ando,
Maximum output control of photovoltaic (PV) array, in
Proc. 35th Intersociety Energy Convers. Eng. Conf. Ex-
hib., pp. 380383, 2000.
[16] P. C. M. de Carvalho, R. S. T. Pontes, D. S. Oliveira, D. B.
Riffel, R. G. V. de Oliveira, and S. B. Mesquita, Control
method of a photovoltaic powered reverse osmosis plant
without batteries based on maximum power point track-
ing, in Proc. IEEE/PES Transmiss. Distrib. Conf. Expo.:
Latin America, pp. 137142, 2004.
[17] T. Esram and P. L. Chapman, Comparison of photo-
voltaic array maximum power point tracking techniques,
IEEE Trans. Energy Conv., Vol. 22, pp. 439449, 2007.

163
Optimal Power System Restor ation and
Reconfigur ation in Distr ibution Cir cuit
Using BFAM and BPSO
K. Sathish KUMAR, T. JAYABARATHI

School of Electrical Sciences, VIT, Vellore, Tamil Nadu, India.
Email: k_sathishkr@sify.com

Received June 19
th
th
, 2009; accepted August 7
th
, 2009.
ABSTRACT
This paper approaches the problem of restoring a faulted area in an electric power distribution system after locating
and isolating the faulted block and reconfiguring the system. Through this paper we are going to explain the power
system restoration technique using brute-force attack method (BFAM) and binary particle swarm optimization (BPSO).
This is a technique based on the possible combination in mathematical analysis which is explained in the introduction.
After isolating the fault, main concentration will be towards the reconfiguration of the restored system using BPSO.
Here due to fault in the system near-by agent will be affected and become useless and will go in the non-working mode.
Now in order to restore these near-by loads we will give a new connection called NO (Normally Open. Using these
switch system will be restored with power availability. After restoration using the BFAM, the BPSO will be used in or-
der to provide the stable configuration. The output of the BFAM will be used as input for the BPSO and then we will
reconfigure our system in order to provide the stable configuration. The effectiveness of the proposed BFAM and BPSO
is demonstrated by simulating tests in a proposed distribution network and verified the results using the Matlab and C
programming.
Keywords: Brute-Force Attack Method, Power System Restoration and Particle Swarm Optimization
1. Intr oduction
The key elements of power system are its continuity and
reliability. If these two elements are deviated from their
normal condition then it may leads to the system in ab-
normal conditions which may be either alert state or
emergency state. Both shouldnt be entertained in the
system. If condition reaches to the extreme state then we
may have to shut down our system, this leads to a heavy
loss in terms of money and customer satisfaction. This is
the area on which many real time researches are going on
and still future cant be predicted. So up to 99%, continu-
ity of the system can be maintained but due to fault, 1%
cant be predicted. Hence the assurance to the customers
that their system is 100% efficient cannot be given. In
order to overcome this problem and for system recon-
figuration we just go through a process called restoration.
It is nothing but process of maintaining power balance
after fault. The main aim of system restoration is to re-
store as many as loads possible which are being affected
by faulted nodes by considering available power without
violating constraints of the system. Through this paper
we have implemented a technique for power system res-
toration which uses the BFAM and finally reconfigure
itself using the BPSO. First section of our paper includes
the introduction and basic idea of the BFAM. In this sec-
tion we will have an idea how we achieve our restoration
process using this technique after a fault. The second
section will give a real time analysis of restoration proc-
ess using the BPSO using transmission and distribution
parameters. In this section we take the output solution of
the restoration in BPSO as input and final output is used
to reconfigure the system. In the third section we com-
pared the proposed BPSO using the BFAM. In section
four, an attempt has been made to find a new method for
restoration using BPSO and BFAM considering their
advantages with each other.
2. Br ute For ce Attack Method
Through this method we are going to explain the power
system restoration technique for fault analysis and the
restoration using BFAM. This is a new technique based
Optimal Power System Restoration and Reconfiguration in Distribution Circuit Using BFAM and BPSO 164
on the possible combination in mathematical analysis.
For a three digit number we have possible thousand com-
binations (000999) and there is an only one solution by
taking all the possible combination and arranging them
we can get our required one solution. In the same way, in
our power system if a node has fault then its near-by
agents will make all the possible combination with the
near-by node starting(1
st
) to end(n
th
) and then will go
up to the last load and the best solution path out of these
combination will be used for the restoration. Here we
have compared the BFAM with the restoration using
multi-agent technique [13]. Due to the fault in the load
they are unable to receive the power. Hence our aim is to
restore the faulted load considering the various factors
like time, amount of power available to the feeder and
system continuity. The constraints are
The systems radial structure must be maintained
during the analysis.
Restore maximum possible load within the out-of
service area with the available feeder power.
The plan must not overload any equipment or system
component with its rated value.
The numbers of switching operation should be
minimum in order to save the power losses.
After reconfiguration the system should maintain its
continuity and reliability.
3. Pr oblem For mulation
In this section we will formulate a system so that we can
perform an analysis on this system after a fault. We can
represent any system with the nodes (N) formed by the
network blocks and switches (S) which connects the two
blocks that may be in opened or closed position. Feeder is
one of the parameter of the system which supplies the
power to the system and has the demand power and extra
power. Hence for each load its required power must be
known to us. Thus by calculating the demand of each
node we will calculate the total demand of any feeder.
Now after finding total demand of the feeder we can find
the reserve capacity by subtracting the total demand of
the load with from the total capacity of the system load is
also a one of the important parameter of the system
which consumes the power supplied by the generating
part of the system. In addition, we are going to consider a
three feeder system for simplicity. The real system is
very large and we generalized this method for a large
system also. Beside these parameters there is one more
special connection between the loads of feeder one to the
load of other feeder. This special connection is working
in such a way that, connection is NO (normally open)
condition when there is no fault in the system, but it will
close as soon as there is a fault in the feeder. Now we
analyse our system as a tree which have its sub-tree con-
nected in a proper path. The advantage of tree configura-
tion is that it is efficient in order to maintain distribution
feeder continuity and radiality. Now in figure 1 the status
of a 3 feeder network having 5 loads in each feeder is
shown. In general the feeder will work normally and each
feeder will provide the necessary amount of power to
each load depending upon their requirement and some
amount of power will always be there as in reserve mode.
Now initially the switch D will be in NO (normally open)
condition as there is no fault in the system. As soon as
there is a fault in the system the whole of the system
status of system will change.
4. Fault Analysis
In the power system we cannot give the assurance that
system will work 100% efficiently. Even 0.1% irregular-
ity in the system can lead to a big fault in the system. But
we could restore the system after the fault by alternate
path. If fault occurs in the system the loads which are
getting the power through the faulted load, will also
affects because of the fault. Hence because of the fault
the healthy loads will be affected. So we cant restore the
faulted load and we have to repair it manually, that is the
only option we have for the faulted load. For the affected
load we have a different option, to apply the same
amount of power to this load through another way. With
the help of another feeder, we can restore this load now
just after a fault the faulted load will be disconnected
from the near-by connected load. Now these near-by
loads will make a sub-trees that have to be restored. Now
in this system there are two sub-trees, they are load 3 and
load 46 that is to be restored. Here each of sub-trees is
basically a sub problem that we must solve separately
because they have no direct relation after the fault. The
sub-tree formed is depends on the faulted block. In the
proposed system the feeder 1 and 2 are having extra
1010 unit and feeder 3 is having 5 units and each blocks
in the feeder requires 5 units. Now in the proposed sys-
tem if a fault is present at load 2. The near-by loads of the
faulted load will separate themselves from the faulted
load in order to save the non-faulty loads from the faulty
dead. After a fault the un-served load will make the list of
sub-tree. Now this sub-tree will be equivalent to a
sub-problem and that must be solved separately. In the
above case if fault at load 2, then its near-by agents 1, 3
& 5 will separate themselves from the faulted load 2.
Before fault the power is supplied through load 2 for the
loads 35 & 6. Because of the fault in load 2, loads 35,
6 will become useless and they will be out off from the
network. At this stage our aim is to restore these loads
through another way. Now those blocks which are in off
condition will find the possible path starting from the
initial block (1) to the end block (18) and then these
blocks will check the power availability to the feeder and
then blocks will be restored according to the power avail-
ability of the feeder. If all the feeders dont have addi-
tional power for the blocks then in that condition our
blocks cant be restored. After the separation of near-by
agent from the faulted load the special switch D will
come in to the network. In the above system the switch D
will close for the load 35 & 6. Hence they will get re-
quired amount of power from the near-by feeder and can
restore themselves. Hence load 3 connects itself from
load 14 of feeder 3 and it will be restored. Now feeder 3
is having zero additional units, load 4 and 5 will con-
nected itself from the load 11 and 12 of feeder 2 and it
can be restored. Now feeder 2 is having 0 buffer unit,
finally the load 6 will not be restored as no feeder is hav-
ing additional amount of units.
Now our algorithm maintains a list of solution, which
is initially vacant and to which the solutions are added as
they are found as the maximum number of solution in the
list and it can be changed by the operator. In the list the
best solution will be taken as the final conclusion for our
fault analysis.
4.1 The Steps in the Algor ithm
1) Identify the block in which the fault is occurred.
2) Open all switches connected to that block in order to
isolate it. For each switch that is opened, a sub-tree is
generated, which have to be restored, except for the
switch that connects the faulted block to a block the
feeder is still feeding normally (a block that is before
the fault).
3) Based on step 2, make change in the bus connec-
tivity matrix and from the load status array. Put zero in
bus connection matrix for no connection between node
and put zero for unfed node in load status array. If there
is a connection between two unfed node(load[i,j]=1)
change it by loop Switch(load[i,j]=2).
4) Create a list of the blocks still being fed and con-
taining at this moment only the blocks the substation is
still feeding, if any.
5) Check if there is a loop switch between a fed block
and an unfed one. (Possible connections are represented
as 2 in bus-connectivity matrix).
6) The unfed block which is only connected to faulted
block (node or load) cannot be restored.
7) If there is a loop switch between fed and unfed
block, check the feeder at which the fed block is getting
power.
8) If that feeder has additional power for unfed block,
connect the fed and unfed node and make necessary
changes in the load connectivity matrix and load status
array.
9) Now move to next unfed block and repeat from step
4.
10) After checking the each unfed block, repeat again
from step 4 for n times where n=no of fault node (block)
from starting.
11) Check the load status and which node the value to
load status is zero and that cannot be restored.
5. Fitness Function
To make the stability and reliability of the distribution it
is very important to keep the load balancing via feeder
reconfiguration so that network could be enhanced. The
objective of this optimization problem can be expressed
by the minimization of the load balancing index (LBI) as
in [5]:
LBI=
2
1
| |
( )
N
i
i R i
i
I
L
I

where
N: Total number of branches in the system after restora-
tion,
L
i
: Length of branch i,
I
i
: Complex current flow in branch i,
R
i
I : Current rating (ampacity) of bus i.
The LBI value should be minimum in-order to get the
best solution by adjusting the switch position. According
to the switch operation the complex current is flowing
through sectionalized switches in the simplified model [6].
The reconfiguration problem cannot be solved by the
traditional methods because of the non linearity of the
system after the fault.
6. PSO
The PSO algorithm was first proposed by Eberhart and
Kennedy [7]. In PSO method, each individual is treated
as particle in the H-dimensional search space. Each indi-
vidual in the search space has some velocity which is
dynamically adjusted according to its own and its
neighbours the moving experience.
6.1 BPSO
For a big system having more than one feeder, then
feeder to feeder connection will be established through
the tie and sectionalized switches which will work only in
the fault condition in order to supply the additional power
through the alternate path. Now using the fitness function
of this system our process will find all the possible com-
bination of solutions for the restoration and then recon-
figure itself by using the BPSO method. This method is
proved to be a very efficient way of reconfiguring the
network. By removing the faulted node from the network
and restoring the system by isolating the faulty load.
Various components of the power system have the ability
to withstand just above the rated capacity [4]. Some
nodes are affected by the faulted node, will create an un-
balance in the network which may lead to failure in the
buses. So it needs to be reconfigured as fast as possible in
order to save it from the severe fault. In a binary PSO,
Feeder 1
1

Figur e 1. System wor king in nor mal condition when ther e is no fault in the system

trajectories of V
ih
are changed in the probability that a
coordinate will take on zero or one value or in between
them [8]. The moving velocity is defined in terms of
changes of probability that a bit will be in one state or the
other depending on the trajectory of V
ih
. Thus a particle
moves in a state space restricted to 0 and 1 on each di-
mension, where V
ih
represents the probability of bit X
ih

taking the value 1. In other words, if V
ih
=0.45, then there
is 45% chance that X
ih
will be 1, and 55% chance that X
ih

will be 0. With this definition, p
ih
and x
ih
can take inte-
gers 0 or 1, and v
ih
, since it is probability and it must lie
within the interval [0, 1]. Hence for the proper explana-
tion we can determine the function of this probability as
shown below-
For h=1: H
1 1
( ) 2 2 ( );
ih ih
ih ih gh ih
V w v c rand
p x c rand p x

If ( ( ))
ih
rand S v then x
ih
=1
else x
ih
=0 (3)
end end S(x) =
1
1
x
e

where S(x) sigmoid limiting transformation.
4
3
5
2
6
14
7
8 9
11
10
12
13
15
18
17 16
Feeder 2
Feeder 3

1
4
3
5
2
6
7
8 9
11
10
12
13
15
14
18
17 16
Feeder 2
Feeder 1
Feeder 3
Figur e 2. System after r estor ation with the help of near -by feeder when fault is ther e in block 2

Figur e 3. Analysis thr ough C
While applying the BPSO method there are some con-
straints that we have to follow in order to maintain the
continuity and radiality of the system.
Each feeder should be in active state in the whole
system.
To maintain the radial structure of the system at
least one branch from one feeder to another must be in
open state otherwise the radiality of the system will be
lost. The Potential limit for buses should be between
minimum and maximum value. Beside this the operating
conditions of the tie-switches will be also kept in mind
[9]:
In the system at least one tie switch will be in closed
position and all other in open position. The power flow
should take place in load side end it should not flow in
the feeder side.
6.1.1 Algor ithm [10]
1) Set the population size (n), maximal iteration number
and stop criterion.
2) Randomly select n feasible solutions of x, compute
p for each x, p
g
is minimum in all p
i,
and the initial values
of v
i
is zero.
3) Using (2), calculate v
i
for particle i.
4) Use (7), to update i.
5) Calculate the feeder load balancing index.
6) If the fitness value of particle i is better than the pre-
vious p
i
, then set it to p
i
. If the best p is better than p
g
,
then set it to p
g
.
7) If stop criterion is satisfied, pg is optimal solution,
otherwise, go to step 3.
Now from the BFAM we are restoring our system by
taking the best possible solution. For this system the ini-
tial status and parameter of this system will be known to
us. Now we will form the bus connectivity matrix for our
system and for any fault in the system we will get a dif-
ferent load connectivity matrix after the fault analysis
corresponding to the load status and finally our system
will give required solution for the proper restoration. The
output of the fault analysis which will be stored in a
variable will be the input of BPSO. Now corresponding
to this BPSO we will get a final best solution which will
be used for reconfiguration of the system. In this solution
we keep the nearby switch of faulted load have to always
at off state by proper logical AND operation. After this
we will get our final solution and implement the same
procedure for the reconfiguration.
7. Result Analysis
In the Figure 1 given in the last suppose fault is there in
the load 2 than results for the restoration in the c pro-
gramming will be-Enter the no of fault load=2
Demand node 1 is separated from 2
Demand node status is: - 1 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1
Demand node 3 is connected to demand node 14, De-
mand node 3 is restored.
Extra power Feeder1=10, Feeder2=10, Feeder3=0,
Demand node 4 is connected to demand node 11, De-
mand node 4 is restored. Extra power Feeder1=10
Feeder2=5, Feeder3=0 Demand node 5 is connected to
demand node 4, Demand node 5 is restored. Extra power
Feeder1=10, Feeder2=0, Feeder3=0
Demand node 6 cannot be restored
Demand node 2 cannot be restored (Figure 3)
8. Conclusions
Through this paper an attempt has been made to restore
the faulted area in the power system and finally reconfig-
ured the system in the stable state. The above process can
be implemented in a real time network and can be used
for the restoration through BFAM and reconfiguration
through BPSO. Also we perform the fault analysis in a
real time network and simulated the results in the
MATLAB and C programming. We have implemented
the BFAM to get the priority based solution. We find that
the results found from the analysis are matching with the
actual data. Hence the above method can be used to per-
form the analysis of fault and to restore, reconfigure the
system in a real network. The restored network is shown
in the Figure 2 and the prospective analysis is given in
the result analysis (7) and the proof in C is shown in the
Figure 3. But the conventional methods are limited only
for less number of loads.
REFERENCES
[1] J. A. Momoh, and J. L. Feng, A multi-agent-based resto-
ration approach for NAVY ship power system, IEEE
Center for Energy Systems and Control (CESaC) De-
partment of ECE, Howard University, Washington DC,
2009
[2] T. Sakaguchi and K. Matsumoto, Development of a
knowledge based system for power system restoration,
IEEE Trans. Power App. Syst., Vol. PAS102, pp. 320
329, Feb. 1983
[3] K. Manjunath and M. R. Mohan, A new hybrid multi-o-
bjective quick service restoration technique for electric
power distribution systems, Science Direct, Electrical
Power and Energy Systems, Vol. 29, pp. 5164, 2007.
[4] M. P. Papadopoulos, G. J. Peponis, N. G. Boulaxis, and N.
X. Drossoss, Heuristic methods for the optimisation of
MV distribution networks operation and planning, Elec-
tricity Distribution Part I. Contributions. 14th Interna-
tional Conference and Exhibition on (IEE Conf Publ. No.
438), Vol. 6, pp. 9/1 9/5, 25 June, 1997.
[5] J. Liu, P. X. Bi, Y. Q. Zhang, X. M. Wu, Power flow
analysis on simplified feeder modelling.
[6] J. Kennedy and R. Eberhart, Paticle swarm optimizaion,,
Proceedings, IEEE International Conference on Neural
Networks, Vol. 4, pp. 19421948, 27 Nov.1 Dec. 1995.
Optimal Power System Restoration and Reconfiguration in Distribution Circuit Using BFAM and BPSO
169
[7] J. Kennedy and R. C. Eberhart, A discrete binary version
of the particle swarm algorithm, Systems, Man, Cyber-
natics, IEEE International Conference on Computational
Cybernetics and Simulation, Vol. 5, pp. 41044108, 12
15 Oct. 1997.
[8] A. Augugliaro, L. Dusonchet, M. G. Ippolito, and E. R.
Sanseverino, Minimum losses reconfiguration of MV
distribution networks through local control of tie-
switches, IEEE Transactions on Power Delivery, Vol. 18,
No. 3, pp. 762771, July 2003.
[9] X. L. Jin, J. G. Zhao, Y. Sun, K. J. Li, and B. Q. Zhang,
Distribution network reconfiguration for load balancing
using binary particle swarm optimization, International
Conference on Power System Technology-POWER-
CON04, Singapore, 2124 Nov., 2004.
[10] T. Nagata, H. Sasaki, and R. Yokoyama, Power system
restoration by joint usage of expert system and mathe-
matical programming approach, IEEE Trans. Power
Syst., Vol. 10, pp. 14731479, Aug. 1995.

Islanding Detection Method for Multi-Inver ter
Distr ibuted Gener ation
Alben CARDENAS, Kodjo AGBOSSOU, Mamadou Lamine DOUMBIA

Institut de recherche sur l'hydrogne (IRH), Dpartement de Gnie lectrique et Gnie Informatique, Universit du Qubec
Trois-Rivires, Trois-Rivires (Qubec), Canada.
Email: {Alben.cardenasgonzalez, Kodjo.agbossou}@uqtr.ca.

Received May 21
st
rd
rd
, 2009.
ABSTRACT
Islanding detection is an essential function for safety and reliability in grid-connected distributed generation (DG) sys-
tems. Several methods for islanding detection are proposed, but most of them may fail under multi-source configura-
tions, or they may produce important power quality degradation which gets worse with increasing DG penetration. This
paper presents an active islanding detection algorithm for Voltage Source Inverter (VSI) based multi-source DG sys-
tems. The proposed method is based on the Voltage Positive Feedback (VPF) theory to generate a limited active power
perturbation. Theoretical analyses were performed and simulations by MATLAB /Simulink /SimPowerSystems were
used to evaluate the algorithms performance and its advantages concerning the time response and the effects on power
quality, which turned out to be negligible. The algorithm performance was tested under critical conditions: load with
unity power factor, load with high quality factor, and load matching DERs powers.
Keywords: Distributed Generation (DG), Interconnected Power Systems, Islanding Detection, Power Generation,
Voltage Positive Feedback.
1. Intr oduction
The Distributed Energy Resources (DER) including Dis-
tributed Generation (DG) and Distributed Storage (DS)
are, as renewable energy resources, very important to
improve power distribution reliability and capability.
Their penetration is increasing nowadays and their utili-
zation shows potential for rural utility solutions [1]. The
Hydrogen Research Institute (HRI) has designed and
developed a renewable energy (RE) system which in-
cludes Photovoltaic (PV) arrays, Fuel Cells (FC) and
Wind Turbine Generators (WTG) with an energy storage
capability using electrolytic hydrogen [2]. This RE sys-
tem operates presently in stand-alone mode. It can be
adapted for rural dispersed generation solutions and in-
terconnected with the electric utility grid by using in-
verter based interfaces (DC/AC static converter). Figure
1(a) is a simplified diagram of the basic RE unit as im-
plemented at HRI. Figure 1(b) shows the possible multi-
source DER system presently under construction.
An important technical issue with utility interfaced
DER systems is unintentional islanding operation. The
islanding condition occurs when the utility is discon-
nected and the DG continues to supply power to the local
load. This condition is not desirable because it can gener-
ate voltage and frequency instability and power quality
degradation; and it constitutes a great risk for mainte-
nance personnel. In view of the importance of human and
equipment protection, unintentional islanding for DG
operation is not tolerated [3]. For these reasons the detec-
tion of unintentional islanding operation is required as
rapidly as possible to allow the timely disconnection of
the DG units. According to the IEEE 1547-2003 standard
[4], the DG disconnection is required within two seconds
after the utility disconnection. Consequently, for safety
DER integration, Anti-Islanding (AI) protection is a re-
quirement.
Remote and local techniques are used for islanding de-
tection. Remote techniques such as Supervisory Control
and Data Acquisition (SCADA), Trip (disconnect) Signal
and Power Line Carrier Communication (PLCC) systems
are centralized methods implemented on the utility side.
They offer high performance and applicability on multi-
source topologies. However, those centralized methods
are expensive to implant [5]. On the other hand, local
techniques include passive and active methods which are
implemented on the DG side. Local passive methods
have a large Non Detection Zone (NDZ), and hence are
not useful for high DG penetration. A solution for the
NDZ reduction is the utilisation of local active
anti-islanding methods.
Islanding Detection Method for Multi-Inverter Distributed Generation 171

Figur e 1. Simplified diagr am of stand-alone RE system implemented at HRI and the possible multi-sour ce DER system

Those active methods are currently based on the injec-
tion of voltage, frequency or output power perturbations,
and the subsequent monitoring for the detection of
changes in electric parameters to confirm islanding con-
dition. Those methods can detect the islanding condition,
but one of their problems is that they can fail when mul-
tiple sources are connected at PCC, because the effect
produced by one source may be interfered by another one
if synchronization between the multiple converters is not
possible. Another drawback of active methods is that they
can cause power quality disturbances as Total Voltage
Harmonic Distortion (TVHD) increase and voltage and
frequency fluctuations or instability. These problems
become bigger if the introduced perturbation is increased
to make possible the islanding detection [6,7], especially
in systems with high penetration.
Use of the Correlation Function combined with active
methods is proposed in [8,9] and [10] for multi-source
topologies. In [8], the correlation function is combined
with an active method that introduces a constant alternat-
ing perturbation of reactive power (5% and 10%), the
anti-islanding algorithm is implanted in only one (master
unit) of multiple DGs, and the others units use a passive
anti-islanding scheme. The detection time depends on the
output power of the master unit and on the reactive power
perturbation level. In [9] and [10], the correlation func-
tion is combined with an active algorithm that introduces
a user defined or random (M-sequence) perturbation of
the output voltage (fixed to 2V for 120V/60Hz system).
The correlation function may change with the number of
connected DGs, and consequently a threshold adjustment
is necessary if the number of units change.
In this article we propose an active islanding detection
method based on Voltage Positive Feedback (VPF) and
passive method Under/Over Voltage Protection and Un-
der/Over Frequency Protection (U/OVP-U/OFP). The
proposed method can be used on multi-source configura-
tions, and allows both unity power factor and power fac-
tor improvement operation modes. This method intro-
duces a limited active power perturbation proportional to
measured variations of PCC voltage (V
PCC
). Simulations
using MATLAB/Simulink and SimPowerSystems
are carried out to validate the algorithm under several
operating conditions.
2. Power Contr ol Scheme
The system we consider is illustrated in Figure 1, where
several DG units are interconnected with the utility at
PCC. Each unit has an IGBT voltage source inverter (VSI)
and its active and reactive power control using a current
control scheme [11] as shown by Figure 2.
In this power control scheme, the output current fun-
damental magnitude (I
INV
(1)) and phase angle (|
I
) are
calculated respectively using (1) and (2).
) cos(
) 1 (
*
I PCC
INV
INV
V
P
I
|
= (1)
|
|
.
|
\
|
=

*
*
1
tan
INV
INV
I
P
Q
| (2)
where, P
INV
*
and Q
INV
*
are respectively the reactive and
active power external set points for the DG unit.
The power angle |
I
represents the phase angle between
the inverter output fundamental current and the funda-
mental voltage measured at PCC.
The resultant set-point current (3) is used to generate the
switching signals for the IGBT bridge inverter, using
Hysteresis Current Control (HCC) or Sinusoidal Pulse
Width Modulation (SPWM) techniques.
) 2 ( ) 1 (
*
u | t + + =
I INV INV
ft Sin I I (3)
where u and f are the phase angle and frequency of the
voltage measured at PCC, and t is time in seconds.
Considering that the proposed algorithm (see Section 3)

Figur e 2. Power contr ol scheme for single gr id connected DG unit

Figur e 3. Voltage positive feedback with d-q cur r ent contr ol scheme

introduces an active power perturbation that is added to
the external set point, we are not limited to the Figure 2
power control scheme, and it may be changed to another
one such as the d-q transformation based power control
scheme [3]. Notice that the d-q control scheme is con-
venient when decoupled active and reactive power con-
trol is required principally in three phase systems.
3. Islanding Detection Algor ithm
This section describes the voltage positive feedback princi-
ple and the proposed active islanding detection algorithm.
3.1 Voltage Positive Feedback Islanding
Detection Methods
Positive feedback with d-q current control based family
of islanding detection methods is presented in [12] and
[13]. These methods consider the relation between the
active (P) and reactive (Q) powers with the voltage mag-
nitude (V) and frequency (f) as shown in (4) and (5), and
the effects of current magnitude and angle deviation on
the output active and reactive powers.
R
V
P
2
= (4)
( )
L
C V Q
=
e
e
1
2
(5)
where, e=2tf, and R, C and L are the resistance, capaci-
tance and inductance of the resonant load.
This family of islanding detection methods includes fre-
quency and magnitude of voltage positive feedback based
schemes. The positive feedback is used to generate a low
frequency perturbation signal (i
d
or i
q
) that is added to
the i
d
*
and/or i
q
*
set points.
Figure 3 shows the principle of Voltage Positive
Feedback (VPF) with d-q current control scheme. The
d-axis component of V
PCC
(V
d
) is monitored and filtered
using a band pass filter (BPF) to obtain the voltage varia-
tion V
d
, this voltage variation is amplified with a preset
gain G (A/V) and used as d-axis current perturbation
(i
d
). The d-axis current perturbation signal affects di-
rectly the inverter output power and consequently the
V
PCC
magnitude and frequency in islanded mode. A satu-
ration block is used to limit the output current perturba-
tion. As a result, on islanding condition a rising deviation
of frequency (df) or magnitude (dV) of V
PCC
is observed,
and this deviation can trip U/OVP or U/OFP for DG
safety disconnection.
An important characteristic of the VPF based methods
is the low power quality degradation in contrast with
other active methods that use distorted signals injection,
as proposed in [14] and [15].
On the other hand, the time necessary to generate the
trip signal using the VPF based method is determined by
the load quality factor q
F
(6) and the feedback preset gain
G. One simple way to improve the response speed is to
increase G, but this solution increases the risk of voltage
or frequency instability, especially in multi-source to-
pologies.
L
C
R
P
C
Q
L
Q
F
q =
=

Figur e 4. Pr oposed voltage positive feedback scheme

voltage measured at PCC (V
RMS
) as the feedback variable
to generate a limited active power perturbation. The ba-
sics of the proposed scheme are presented in Figure 4.
The V
RMS
(after the LPF filter) is compared with a refer-
ence voltage V
REF,
and

the difference V is used to cal-
culate the active power perturbation P.
The reference voltage V
REF[k+1]
is set initially equal to
the nominal RMS voltage (V
NOM
), and is subsequently
updated only on System Stable Condition (SSC) using
the historic RMS average voltage V
AV
(7). Otherwise, the
new voltage reference (V
REF[k+1]
) is set equal to the old
reference value (V
REF[k]
) according to (8). The SSC is
defined as the condition where both the power and the
voltage perturbations (P and V) are stable.
(6)
The Sandia Voltage Shift (SVS) method [6] uses the
utility voltage to calculate the output current amplitude;
in this method the average voltage of the utility is com-
pared with the actual voltage in each electric cycle (or-
half cycle) to calculate the current perturbation that is
amplified by a preset gain.
In both methods, SVS as well as VPF with d-q transforma-
tion, the output voltage at the islanding condition is forced to
the trip points of the U/OV protection by an important output
current reduction or increase, and it is finally the U/OVP that
shuts down the power converter. This important perturbation
of the output current before the disconnection may affect the
load, and is not appropriate if stand-alone operation of the
system is desired after the safety disconnection.
3.2 Pr oposed Voltage Positive Feedback Scheme
We propose to use the VPF concept, taking the RMS
| |
| |
m
V
V
m
i
i k RMS
k AV
=
=

1
(7)
where, m is the number of samples considered for the
average calculation.
(8)
The active power perturbation P is calculated using
the maximal allowed power perturbation P
MAX
, the
minimal power perturbation P
MIN
, a gain factor G, and
the difference between V
REF
and V
RMS
, according to (9,10)
and (11).
(9)
| | | |
| |
=
=
+
otherwise V
SSC V
t V
V
k REF
k AV
NOM
k REF
,
,
0 ,
1

| |
A > A A
A < A < A A
A < A A
= A
+
MAX C MAX
MAX C MIN C
MIN C MIN
k
P P P
P P P P
P P P
P
, ) sgn(
,
, ) sgn(
1
c
c

c = A G P P
C
*
(10)
( )
REF
REF RMS
V
V V
= c (11)
3.3 Par ameter Selection
The parameters P
MIN
, P
MAX
and G are selected to pro-
duce a low active power variation in the interconnected
mode and a low voltage variation in the islanding detec-
tio
V).
nnected mode, the active power deviation is
reduced to minimal P
MIN
on voltage stability condition.
The expected evolution of voltages (V
REF
and V
RMS
) and
the active power perturbation for the islande

n period. Considering an ideal utility source, the volt-
age error in the interconnected operation mode may be
close to zero, but in practice a minimal error is always
resent, and this error affects the real output power. We set
the P
MIN
near the mean active power perturbation cal-
culated using the typical utility voltage variations (
MIN
).
Based on the measured voltage of the utility, we take a
MIN
=0.167% (0.2V at 120V) as the minimal voltage
error. This
MIN
allows us to set a minimal active power
perturbation P
MIN
=0.5% using a gain of G=3. At the
islanding condition, if the system operates at unity power
factor, this P
MIN
introduces a voltage variation of
0.25% (0.3V at 120
To limit the effects on the output voltage in the detec-
tion period, we set the P
MAX
=2.5% to produce a maxi-
mal voltage error = 1.24% (1.5V at 120V). This set-
ting permits the islanding detection without output volt-
age degradation if the load and DG powers are close or
matched.
3.4 Expected Oper ation of the Pr oposed
Algor ithm
In grid co
d mode is

Figur e 5. Expected effect of the pr oposed VPF scheme un-
der islanding condition

Figur e 6. Expected effect of the pr oposed VPF scheme un-
der voltage r eduction and nor mal voltage var iation
shown by Figure 5. Before the utility disconnection (t<t
0
),
the normal voltage variation can generate a minimal ac-
tive power deviation without important effect on V
vo
S
<
produces a
negative increasing of P until its saturation at t=t
1
. The
SSC condition is reached with V
RMS
stab
Consequently, the V
REF
is updated and th
oltage variation.

3.5 Islanding Confir mation
If the average of the magnitude of the active power per-
tu
AV
ed to two thresholds values P
AC
(Active
Counter) and P
RC
(Reset Counter) to activate or
a time counter (TC). If the TC count is larger than a p
limit of time T
MAX
, the islanding condition can be confirmed.
RMS
tage. After the utility disconnection (t>t
0
), if V
RM

l
V
REF
, the negative P
MIN
and the VPF effect produce a
progressive V
RMS
reduction and negative increase of P,
until the P saturation at t=t
1
.
The SSC condition is reached at t=t
2
with voltage sta-
bilization, and V
REF
is updated to V
AV
. Consequently, a
reduction and a subsequently positive increase of P are
expected to produce a voltage level increase (from t
2
to
t
4
). The power perturbation is saturated at t=t
3
, and a new
SSC is reached at t=t
4
. A cyclic power perturbation and a
voltage level oscillation can be observed and used to con-
firm the islanding condition.
The expected trajectories for a voltage variation are
presented in Figure 6. In this case, if an important voltage
variation is occurred at t=t
0
, the VPF effect
ilization at t=t
2
.
e active power
perturbation is reduced to P
MIN
. Subsequently the
minimal active power perturbation may be observed as an
effect of the normal v
rbation (P
AV
) is calculated and observed during a de-
tection period T
DET
, then under islanding condition the
expected profile of this variable is shown in Figure 7, and
we can use this new variable to establish the islanding
condition. For islanding condition confirmation, the P
is compar
to reset
reset
175

Figur e 7. Expected magnitude aver age of the a

ctive power per tur bation on islanding condition

osed anti-islanding pr otection scheme
3.6 Classic VPF Methods and Pr oposed Method
Compar ison
Figur e 8. Simplified diagr am of the pr op

In the example of Figure 7, the utility is disconnected
at t=t
0,
the activation of TC is produced at t=t
1
, and the
islanding condition is confirmed at t=t
2
.
We set the reset and the activation thresholds of time
counter (TC) as P
RC
=30% and P
AC
=90% of the P
MAX
,
and T
MAX
=9 electric cycles (150ms).
If the DG and load powers dont match, the classical
solutions as U/OVP and U/OFP can be employed to con-
firm rapidly the islanding condition. The combination of
the proposed algorithm with the classic passive method
(U/OVP and U/OFP) produces a fast islanding detection
in all output power conditions. A simplified diagram o
the proposed anti-islanding protection scheme is pre-
between the proposed The most important difference
method and the known voltage positive feedback meth-
ods, such as SVS and others VPF methods, is that the
proposed method does not produce the U/OVP-U/OFP
trip if it is not really necessary. That is essentially be-
cause in the proposed method the detection is not based
on the forced deviation of voltage or frequency beyond
the trip points. That way, the ride-trough operation of the
DER system is possible if the load and DG powers are
close or matched at the instant of the utility disconnec-
tion.
f
sented in Figure 8.
It is known that the classical VPF methods have low
impact on power quality; however, as mentioned in [16],
the voltage instability risk is high when the classic VPF
methods are employed with a strong feedback gain and
when most of the local load is supplied by the DG. In
contrast the proposed method uses a limited active power
perturbation that does not produce voltage or frequency
instability risk.
On the other hand, considering that the proposed
method uses an active power perturbation that is added to
the external set point, the proposed method may be easily
implemented in current controlled or in phase angle
(power angle) controlled inverters.
4. Simulation Results
or preliminar
orithm, we consider two scenarios: unity power factor
operation and power factor improvement operation. The
power converters of the DG units are connected at PCC
with a 120V/60Hz grid source. We consider two different
disconnection modes, as shown by Figure 9, where the
main circuit breaker (CB1) disconnects the utility, and
the secondary circuit breaker (CB2a or CB2b) is the con-
trolled switch operated by the anti-islanding (AI) control.
In the first case, with the islanding confirmation, the DG
units are shutting down and are disconnected by breaker
CB2a. In the second case, we consider the possibility of
automatic operational mode change from the grid-con-
nected mode to the stand-alone mode when the DG sys-
tem can supply the totality of the load power: in this case
CB2b is opened and CB2a remains closed to allow the
supply of power to the load with safety disconnection
from the utility. Otherwise, the DG units are turned off
and disconnected after the islanding detection as in the
first case.
The simulations are carried out using Matlab/Simu-
link
TM
for the islanding detection algorithm and the
power control schem
PowerSystems
TM
for
entation.
e most difficult
si
4.
ntly a voltage and an active power
F
g
y validation of the islanding detection al-
e implementation, and using Sim-
the power devices models imple-
m
The load RLC elements are calculated to obtain quality
factor values between 0.5 and 2.5. The load and DG sys-
tem power match is considered as th
tuation.
The anti-islanding parameters setup is presented in the
Appendix.
1 Results for Islanding Detection in DG Unity
Power Factor Oper ation
Figure 10 presents the active power and voltage deviation;
the grid, load and DG output currents; and Total Voltage
Harmonic Distortion (TVHD) of V
PCC
, for a system with
three grid-connected DG units, with load at unity power
factor and quality factor q
F
=2.5. The grid is disconnected
at t=0 and subseque

Figur e 9. DG unit disconnection modes

perturbations are observed.
We can also observe that the disco signal is gen-
erated before 14 electric cycles (233ms), and the algo-
le before and
after the islanding condition.
Similar traces are obtained for systems with one and
six DG units. The time needed for the islanding confir-
mation (T
IC
) using the proposed algorithm in the systems
working at unity power factor is presented in Table 2.
In this case we consider also that the DG and load
power are matched. We can notice that this time is al-
ways lower than 2 seconds as recommended by the IEEE
1547-2003 standard. The simulation results show that the
islanding event is confirmed faster for the systems with a
high number of DG units.
As expected, the performance of the proposed method
is increased if an active power mismatch is considered
between the DG and the load. Table 3 shows the simula-
tion results for the time needed for the islanding confir-
mation (T
IC
) in a system with 3 DG units operating at
unity power factor. In this case, we consider different
conditions of active power mismatch (12) between the
DG and the load.
nnect
rithm effect on voltage TVHD is negligib
( )
LOAD
DG LOAD
P
P P
P

= o (12)
rtant variation in the
voltage measured at PCC is observed with the islanding
condition. In this case the O/UVP acts and disconnects
4.2 Results for Islanding Detection in DG Power
F
In all simulated cases, as shown by Table 4, the
If oP is greater than 0.2 an impo
the DG units within 2 electric cycles.
Factor Impr ovement Oper ation
For power factor improvement operation mode, the load
RLC elements are calculated to produce an inductive
power factor PF=0.95 and a quality factor q
F
=2.5. The
DG system is configured to supply the active and reactive
load powers.
The performance of the islanding detection method was
evaluated for systems with 3 and 9 DG units operating in
power factor improvement mode with different load
power factors, and considering a high load quality factor
(q =2.5).
Table 1. Compar ison of classic VPF methods and pr oposed method
Method
Characteristic
Classic VPF
methods
Proposed
Forced trip of U/OVPU/OFP. Yes Not
Negative impact on power quality.
Risk of voltage instability in multi-inverter configura
systems with high penetration.
Low Low
tions and/or
High Low
islanding condition was confirmed within 13 electric
cycles (216ms).
The simulation traces for a system with six DG units
operating in the power factor impro
Table 4. Time necessar y for the islanding confir mation us-
ing the pr DG units
r ating in the pow
T
IC (Cycles)
vement mode are
mulation result shows that
fast detection of the island-
T
IC (Cycles)

presented in Figure 11. This si
the proposed method offers
ing condition in the systems working in the power factor
improvement mode; in this case the disconnection signal
is generated within 11 electric cycles (183ms).
The performance of the proposed method was also
evaluated in the power factor improvement mode, con-
sidering different load quality factor. In this case, we
consider a system with 3 DG units operating in power
improvement mode, with load power factor between 0.85
and 0.98 (0.85sPFs0.98), and load quality factor of 1.0,
1.5 and 2.5 (qF=1.0, qF=1.5, qF=2.5). The simulation
results show that the time needed for the islanding con-
firmation is similar for different conditions of the load
quality factor and the power factor. As shown in Table 4,
the islanding confirmation in the simulated cases is
reached within 12.5 electric cycles (208ms)
4.3 Results for Automatic Change of DG Operating
Mode fr om Gr id-Connected to Stand-Alone
Mode after the Islanding Detection
We tested the performance of the proposed islanding
detection method and observed the effect of the algo-
rithm on the power quality when the stand-alone opera-
tion of DG is allowed after the islanding detection. In

ing the pr oposed method in the systems oper ating at unity
power factor

q
F
1 DG unit 3 DG units 6 DG units
2.5 15.20 13.7 10.8
1.0 13.80 12.8 11.6
0.5 13.30 11.8 11.5

ing the pr oposed method in the systems with 3 DG units
when the DG and load power s dont match
T
IC (Cycles)

oP 0 0.02 0.05 0.10 0.15 0.20
PF=1.0 13.7 11.5 11.5 11.5 11.5 2.0
oposed method in systems with 3 and 9
er factor impr ovement mode ope

q
F

DG
units
PF=0.98 PF=0.95 PF=0.90 PF=0.85
9 12.5 12.5 13.0 12.5
2.5
3 12.5 12.5 12.5 12.0
1.5
3 12.5 12.0 12.5 12.5
1.0
3 12.0 12.0 12.5 12.5

this case the AI protection opens the breaker CB2b to
disconnect the DG system from the utility at the island-
ing confirmation.
Figure 12 shows the V
PCC
voltage and DG
1
output
current (I
INV
) traces before and after the islanding condi-
tion, the disconnection signal, the TCHD of the I and
th
INV
de within a few cycles (10.75 cycles)
after grid disconnection. The Total Current Harmonic
Distortion (TCHD) and TVHD are always m
less than 2%. According the 1547-IEEE standard [4], the
m
tection algorithm
is is generated
racteristics are imposed by the DG system
e TVHD of the V
PCC
for a system composed by four
DG units. In this case, the DG system and load powers
were matched, and the load elements were calculated for
unity power factor (PF=1.0) and high quality factor
(q
F
=2.5).
The simulation results show that the system can
change its operational mode from grid connected mode
to Stand-alone mo
aintained at
aximal allowed TCHD of DG is 5% and the TVHD
measured at PCC must be less than 2.5%. As in the pre-
vious cases, the grid is disconnected at t=0. After the grid
disconnection (t>0), the islanding de
produces the output power perturbation to detect the
isla nding condition nding condition. In this case, the isla
confirmed and the disconnect signal
within 10.75 electric cycles. Then, at t=10.75 cycles,
CB2 is opened and the system changes its operation
mode from the grid-connected mode (on islanding condi-
tion) to the Stand-Alone mode. In the stand-alone mode,
the voltage cha
output power. In the simulated case, the V
PCC
voltage
reaches its nominal level in the stand-alone mode be-
cause the DG and load powers are matched. In a practical
situation, and with a variable load power, the voltage and
frequency control is possible by controlling the active

Figur e 10. Simulation r esults for islanding detection on a thr e
tor mode. Load PF=1.0 and q
F
=2.5

e gr id connected DG units system oper ating in unity power fac-

with six gr id- connected DG units, oper ating in power factor F . Sim tion lts et f ystem
im ent mode
igur e 11 ula r esu for islanding d ection or a s
pr ovem

Figur e 12. Simulation r esults for islanding detection and oper ational mode change fr om gr id connected to stand-alone mode
for four gr id connected DG units system oper ating in unity power factor mode

and reactive output powers using the Voltage-Power
Droop/Frequency-Reactive power Boost scheme (VPD/
FQB) as mentioned in [3] and as proposed in [17].
5. Conclusions
This paper presents an active islanding detection algo-
rithm for multi-source DG systems. The proposed algo-
rithm is based on the voltage positive feedback. Theo-
retical analyses are provided and simulation results show
that the proposed islanding detection algorithm offers
fast anti-islanding protection with negligible impact on
power quality. This method may be useful for systems
with single or multiple grid connected DG units. The
synchronization between the different DG units is theo-
retically not required due to the fact that all units use a
common variable to generate their local anti-islanding
protection.
Considering its fast response and the negligible effect
on power quality, this anti-islanding scheme could be
used to allow the operation of the grid-connected distrib-
uted generation systems with safety disconnection. In
contrast with commonly used methods, that force the
possible the stand-alone operation of the DG system after
the islanding is confirmed, and that without interruption
of the load power.
The proposed method is easy and inexpensive to im-
plant. The detection algorithm may be added to the pro
gram of the power control unit to use the same digital
processor.
Current and future works include the implementation
of the algorithm using FPGA for the experimental vali-
dation under real and critical scenarios using multi-in-
verter configurations.
6. Acknowledgments
This work was supported by the LTE Hydro-Qubec and
the Natural Sciences and Engineering Research Council
of Canada.
REFERENCES
[1] B. Kroposki, R. Lasseter, T. Ise, S. Morozumi, S. Pa-
pathanassiou, and N. Hatziargyriou, Making microgrids
work, IEEE Power and Energy Magazine, Vol. 6, No. 3,
pp. 4053, 2008.
voltage or frequency to the U/OVP or U/OFP trip lim-
its,using the proposed islanding detection method makes
[2] K. Agbossou, M. Kolhe, J. Hamelin, and T. K. Bose,
Performance of a stand-alone renewable energy system
based on energy storage as hydrogen, IEEE Transactions
on Energy Conversion, Vol. 19, No. 3, pp. 633640,
2004.

[3] F. Katiraei, R. Iravani, N. Hatziargyriou, and A. Dimeas,
Microgrids management: Controls and operation aspects
of microgrids, IEEE Power and Energy Magazine, Vol.
6, No. 3, pp. 5465, 2008.
[4] IEEE standard for interconnecting distributed resources
with electric power systems, IEEE standards, in Stan-
dards Coordinating Committee 21 on Fuel Cells, Photo-
voltaics, Dispersed Generation, and Energy Storage, July
28, 2003.
[5] P. Mahat, C. Zhe, and B. Bak-Jensen, Review of island-
ing detection methods for distributed generation, Third
International Conference on Electric Utility Deregulation
and Restructuring and Power Technologies-DRPT, Vol. 1,
pp. 27432748, 2008.
[6] W. Bower and M. E. Ropp, Evaluation of islanding de-
tection methods for utility-interactive inverters in photo-
voltaic systems, Sandia Report SAND20023591, 2002.
[7] R. Bhandari, S. Gonzalez, and M. E. Ropp, Investigation
of two anti-islanding methods in the multi-inverter case,
IEEE Power and Energy Society General Meeting-
Conversion and Delivery of Electrical Energy in the 21st
Century, Vol. 1, pp. 17, 2008.
[8] C. Jeraputra, E. C. Aelwizal, P. N. Enjeti, and S. Choi,
An improved anti-islanding algorithm for utility inter-
Solar
, No. 13, pp. 915, 1996.
[10] M. L. Doumbia, K. Agbossou, and D. Tran-Khanh-Viet,
Correlation technique investigation for islanding detec-
tion of inverter based distributed generation, IEEE
Power Electronics Specialists Conference-PESC2008,
Vol. 1, pp. 45564561, 2008.
[11] Z. Chen and E. Spooner, Voltage source inverters for
high-power, variable-voltage DC power sources, IEE
Proceedings Generation, Transmission and Distribution,
Vol. 148, No. 5, pp. 439447, 2001.
[12] Z. Ye, L. Li, L. Garces, C. Wang, R. Zhang, M. Dame, R.
Walling, and N. Miller, A new family of active
anti-islanding schemes based on DQ implementation for
grid-connected inverters, IEEE 35th Annual Power Elec-
tronics Specialists Conference2004PESC04, Vol. 1,
pp. 235241, 2004.
[13] Z. Ye, R. Walling, L. Garces, R. Zhou, L. Li, and T.
Wang, Study and development of anti-islanding control
for grid-connected inverters, General Electric Global
Research Center Niskayuna, New York, NREL/SR-560-
36243, 2004.
[14] M. E. Ropp, M. Begovic, and A. Rohatgi, Analysis and
performance assessment of the active frequency drift
method of islanding prevention, IEEE Transactions on
Energy Conversion, Vol. 14, No. 3, pp. 810816, 1999.
[15] S. J. Huang and F. S. Pai, Design and operation of
grid-connected photovoltaic system with power-factor
control and active islanding detection, IEEE Proceedings
on Generation, Transmission and Distribution, Vol. 148,
No. 2, pp. 243250, 2001.
scheme on inverter-based distributed generator

g Detection Algor ithm
, G = 3.0
T = 9 electric cycles
connection of multiple distributed fuel cell powered gen-
eration, Twentieth Annual IEEE Applied Power Elec-
tronics Conference and Exposition-APEC-2005, Vol. 1,
[16] X. Wang and W. Freitas, Influence of voltage positive
feedback
pp. 103108, 2005.
[9] O. Tsukamoto and K. Yamagishi, Detection of islanding
of multiple dispersed photovoltaic power systems,
Energy, Vol. 58
stability, IEEE Transactions on Power Delivery, Vol. 24,
No. 2, pp. 972973, 2009.
[17] C. K. Sao and P. W. Lehn, Control and power manage-
ment of converter fed microgrids, IEEE Transactions on
Power Systems, Vol. 23, No. 3, pp. 10881098, 2008.

Appendix
A. Gr id and DG Units Char acter istics
Grid voltage: 120V-60Hz
Grid line inductance: Ls= 0.05mH
DG type: IGBT-VSI, 1kW, 120V-60Hz
DG Output filter: L=5mH, C=0.1F

B. Setup of Islandin
P
MAX
= 2.5%, P
MIN
= 0.5%
P = 0.75% P = 2.25%
RC , AC , MAX
181
Effect of War m Ionized Plasma Medium on
Radiation Pr oper ties of Mismatched Micr ostr ip
Ter mination
Ayman Al-SAWALHA

Physics Department, College of Science, King Faisal University, Saudi Arabia.
Email: drswalha@hotmail.com

Received April 14
th
th
nd
, 2009.
ABSTRACT
Study of radiation properties of mismatched microstrip termination is carried out in warm ionized electron plasma me-
dium. Linearised hydrodynamic theory coupled with vector wave technique is used to investigate radiation patterns and
radiated power of a microstrip mismatched termination in one component electron plasma media for different plasma
to source frequencies .These properties are used to compare the performance of an open circuit discontinuity with those
of matched microstrip termination. Matched terminations are found more suitable for applications at higher frequen-
cies than an open circuit discontinuity in the plasma medium though in free space they are less suitable.
Keywords: Mismatched Termination, Open Circuit Discontinuity, Matched Microstrip Termination, Ionized Plasma
1. Intr oduction
Microstrip antenna has proven to be an effective, light
weight and quite inexpensive radiator for aerospace vehi-
cles [1]. Many workers have reported radiation properties
of different microstrip radiators in free space [24]. Me-
tallic and dielectric losses, breakdown effect at higher
power level and radiations from discontinuities, however,
limits the application of a microstrip line structure over
the ground plane [5]. Radiations from microstrip discon-
tinuity become quite dominant at microwave frequencies
and hence become major limiting factor. Hence, it is al-
ways necessary to control them somehow. Radiations
from an open circuit microstrip discontinuities are inves-
tigated recently [6] in an ionized plasma medium of infi-
nite thickness. Radiations from a mismatched termination
in an ionized plasma medium are investigated in this
communication. Using these relations, performance of
two other types of discontinuities viz. a matched micro-
strip termination and an open circuit microstrip termina-
tion are investigated in free space as well as in ionized
hot plasma medium.
In strip line structure, field does not lie uniformly in
between the conducting patch and the ground plane but
some fringe fields leak in the air near the edges of the
strip. Dielectric polarization beneath the strip takes place
which gives rise to the polarization current in addition to
the already existing strip current [5]. Considering both
these currents together, radiation patterns and radiated
power by a mismatched termination are obtained for dif-
ferent plasma to source frequency (
P
) e e . With the
presence of an actual ionized plasma medium, effective
permittivity of structure changes marginally in compari-
son to the effective permittivity in free space and hence
resonance frequency also changes marginally [7]. It is
found that percent deviation in frequency is about
0.003% in the presence of an ionized plasma media of
infinite thickness which dose not affects predicted results
seriously.
2. Radiation Field Expr essions
The geometry and coordinate system of a mismatched
microstrip termination is shown in Figure 1.
A strip line above the ground plane is located along the
z-axis of the coordinate system. The thickness of the di-
electric substrate is considered to be , width of strip
,relative permeability and permittivity
" " h
" " w 1
r
= and
1
r
c > respectively.
It is considered here that the termination mismatches
the strip. Far fields are obtained by considering a mag-
netic current density alone with a perfect electric current
condition.
For strip line configuration, the integration to find elec-
tric field using vector potential is carried out over a cross
Effect of Warm Ionized Plasma Medium on Radiation Properties of Mismatched Microstrip Termination 182
section which gives a combination of currents formed by
polarization current ( along x-axis) and a strip current
(along z-axis).
Let the strip current be
( )
jk z jk z
s o
I I e Re
'
= +
'
(1)
where
2
eff
o
j
k
R R e

t
c
' =
=

o
is the wavelength in free space,
eff
c is the effective
dielectric constant of substrate material, R is the re-
flection coefficient at the end of the line and its limiting
value lies in between 0 and 1, and is the phase angle
between the incident wave and the reflected wave.
Assuming current amplitude is to be constant a cross
the strip line, the surface current density in z-direction for
end fed line will become
j t
s s z
J I e i
e
= (2)
The polarization current density in x-direction will be
1
2
j t s
p x
I
J h e
z
e
c
c
-
-
| | c
=
|
c
\ .
i (3)
where
x
i and
z
i are unit vectors in the x and z direc-
tion respectively and c
-
is the actual dielectric constant
of the substrate material.
Considering the presence of both these currents to-
gether and following the method of [8], the expressions
for the radiation patterns are obtained in electromagnetic
mode as well as in electroacoustic mode.
These expressions are:
2.1 In Electr omagnetic Mode
( )
2 2
2 2
2 2
2 2
60
( cos ) ( ) cos
( cos )
cos
( ) cos ( cos )
( cos )
e
j t r o o
e
j hw I
E e
r
A A
A
A A
R
A
e |
u
|
c
c c u c c u
c u
c c u c c u
c u
-
- -
- -
=
(
+
(
(
X
(

(
+
(

(4)

( )
60
cos cos
sin
( cos ) ( cos )
e
j t r o O
e
j hw I
E e
r
A A
R
A A
e |
|
c
c u c c c u

c u c u
-
- -
=
(
+
X
(
+ (

(5)

Figur e 1. The geometr y and coor dinate system of a mis-
matched micr ostr ip ter mination

where:
( sin sin
2
sin( sin sin )
2
( sin sin )
2
e
e
w
j
e
e
w
e
w
|
) u
|
u
|
u
X =
2
2
2
1
e
o
p
A
A
t
|
e
e
=
=

e
|
is the propagation constant in electromagnetic mode.
o
|
is the propagation constant in free space (radian/m)
p
e
is the angular plasma frequency ( )
1
sec
e
is the angular source frequency ( )
1
sec
2.2 In Electr oacoustic Mode

( )
2
2 2
( )
2 ( )
(cos sin cos )
( cos )
(sin cos cos )
( cos )
p
p p o
p
p
p
p
j t r
p
p
p
wh I
E
r
k
k
e
k
R
k
e |
| e
t e e e c
| u u c
| u
c | u u
| u
-
-
-
=
( '
(
'
(
X
(
'
(
(
' +

(6)
where
( sin sin )
2
sin( sin sin )
2
( sin sin )
2
p
p
w
j
p
p
p o
o
w
e
w
c
A
|
u
|
u
|
u
| |
0
X =
| |
=
|
\ .

-14
-12
-10
-8
-6
-4
-2
0
0
30
60
90
120
150
180
210
240 300
330
-14
-12
-10
-8
-6
-4
-2
0

P
o
w
e
r

i
n

d
B

A=1.0
A=0.2

270
Figur e 2. E
u
r adiation patter ns for 1.0 A = and 0.2 A = , 0.7 , R t = =
-16
-14
-12
-10
-8
-6
-4
-2
0
0
30
60
90
120
150
180
210
240
270
300
330
-16
-14
-12
-10
-8
-6
-4
-2
0

P
o
w
e
r

i
n

d
B
A=
A=

Figur e 3. E
r adiation patter ns for 1.0 A = and 0.2 A = , 0.7 , R t = =

A : is the propagation constant in plasma mode.
These value of E
u
, E
and
p
E are computed and
plotted in Figures (24) for two different values of
plasma to source frequency. Computations are carried out
for , , op-
eration frequency = 1.2GHz ,
0.158 h c = m 0.471 , 2.31, 3.0 w cm c c
-
= =
E
u
Patterns in free space are almost uni-
form. Radiation intensity in the end fire direction
is slightly more than the radiation intensity in
( 1.0 A = )
)
)
( 0
o
u =
the broadside direction . On increasing plasma
to source frequency
( 90
o
u =
( 0.2) A = , radiation pattern modifies
drastically and direction of maximum intensity shifts
=
0.7 R = and t + = .
50 52 54 56 58 60
0
10
20
30
40
E
p

i
n

d
B
Thet a i n degr ee

Figur e 4.
p
E r adiation patter n in plasma for 0.2 A = , 0.7 , R t = =
Table 1. 3dB beamwidth in u and planes for 1.0 A = and 0.2 A =
E
u
E

Free Space
( 1.0 A = )
Plasma
( 0.2) A =
Free Space
( 1.0) A =
Plasma
( 0.2 A = )
3dB
Front
direction
3dB
Back
direction
3dB
Front
direction
3dB
Back
direction
3dB
Front
direction
3dB
Back
direction
3dB
Front
direction
3dB
Back
direction
108
o
116
o
0
o
120
o
104
o
106
o
82
o
114
o

from to direction and minimum
appears at ( . Almost similar behaviors appear for
( 0
o
u =
u =
) ) ( 180
o
u =
) 0
o
p
E patterns. These patterns in the free space ( 1.0 A ) =
are almost symmetric in all the four quadrants, but for
higher plasma to source frequency value ( , the
3dB beam width in direction is much smaller
than in direction.
0.2) A =
( 0 )
o
u =
) ( 1 u = 80
o
p
E radiation patterns indicate that only one lobe ap
pears in the to range. One substituting 50
o
60
o
1.0 R =
and t + = all the expressions obtained for mismatched
microstrip termination reduce immediately to those of
open circuit microstrip discontinuity [6]. Similarly on
substituting 0 R = and t + = , expressions for micro-
strip matched termination can be obtained [8].
3. Radiated Power
The power radiated by the microstrip mismatched ter-
mination through upper half space is obtained by using
Poynting vector. For different values of plasma to source
frequencies, expressions for radiated power are obtained
by using the method of [8].These expressions are:
3.1 For Electr omagnetic Mode
2
2 2 2
2
2 2
2 2 2
1 2
0
2
15
cos cos sin
o o
e
e
h w AI
P
S S d
t
t
t
|
tc
d u u
-
=
(
+ X

} }

(7)
where
2 2
2 2
1
2 2
2 2
( cos ) ( ) cos
( cos )
( ) cos ( cos )
( cos )
A A
A
S
A A
R
A
c c u c c u
c u
c c u c c u
c u
- -
- -
(
+
(
(
=
(

(
+
(

And
2
cos cos
( cos ) ( cos
A A
S R
A A )
c u c c c u
c u c
- -
u
(
+
=
(
+ (

and the radiation resistance in electromagnetic mode
e
R can be defined as

2
2
e
e
o
P
R
I
= (8)
3.2 For Electr oacoustic Mode
2 4 2 / 2
2 2 2 2
/ 2 0
2
(cos sin cos ) /
[
( cos ) 8 ( )
/ (sin cos cos )
] sin
( cos )
p p o p
p
p o p
p
p
p
mh I k
P
k e n
k
R X d d
k
t t
t
| e | u u
| u t e e e
c | u u
u u
| u
c
- +
-
'
=
'
'
' +
} }

(9)
and the radiation resistance in plasma mode
p
R can be
defined as
2
2
..................................(10)
p
p
o
P
R
I
=
These values of and
e
R
p
R
0 0
are computed and plotte
for different values of plasma to source frequencies in
Figure 5.In the electromagnetic mode, power radiated in
free space ( ) is maximum but decreases on in-
creasing plasma to source frequency. On the other hand,
plasma mode field patterns give a quasi periodic pattern.
Initially, radiated power in plasma mode is less than the
radiated power in electromagnetic mode but it over takes
in between the range
1. A = 0
.25 A to =
.
During the voyage in free space, aerospace vehicle in-
teracts with plasma media and radiates electroacoustic
waves in addition to the usual electromagnetic waves.
Presence of the electroacoustic wave is mainly responsi-
ble for such variation in this plasma media.
It is evident from the expressions of radiated power
that they are a function of reflection coefficient. The
variations of radiated power in free space or ( 1.0 A = )
(
p
0)
e
e
=
and in plasma media
(0 1)
p
e
e
<
)
<
are pre-
sented for both electromagnetic mode and longitudinal
plasma mode in Figure 6.
In free space ,radiated power in electro-
magnetic mode is maximum for
( 1.0 A =
0 R = (matched ter-
mination).As mismatch increases, radiated power de-
creases and becomes minimum at 0.8 R = .Thereafter it
becomes almost uniform up to 1.0 R = (open circuit mi-
crostrip termination).On increasing the plasma to source
frequency , the radiated power by a matched
termination becomes quite small in comparison to its free
space value . On increasing
( A =
A
) 0.2
( ) 1.0 = R value, radi-
ated power in electromagnetic mode increases continu-
ously but it always remains less than the free space value
even for an open circuit termination( 1.0 R = ).
In longitudinal plasma mode, radiated power at
for a matched termination ( 0.2 A = ) ( 0 R = ) is low
but increases continuously up to 1.0 R = (open circuit
termination). The total power radiated in free
space
(
e p
P P + )
( 0)
p
P = is low for an open circuit termination.
4. Conclusions
Effect of the presence of plasma medium on the different
discontinuities is observed here by considering different
plasma to source frequency
( )
p
e
e
values.
It can be concluded from the present study that for op-
eration in free space an open circuit discontinuity better
than a matched termination operating at very high fre-
quencies. In the plasma media, matched termination is
1.0
10
-1
10
0
10
1
10
2
0.8 0.6 0.4 0.2 0.0
10
10
0
10
1
10
2
-1

R
e
A
R
e
R
p

R
p

Figur e 5. Radiation r esistance in EM mode and in plasma
mode for differ ent A and 0.7 , R t = =

0.0 0.2 0.4 0.6 0.8 1.0
0.01
0.1
1
10
0.01
0.1
1
10
R
e
(
o
h
m
)
R
R
e
(A=0.2)
R
e
(A=1.0)
R
p
(A=0.2)
R
p
(A=0.95)
R
p
(
o
h
m
)

Figur e 6.
e
R and
p
R for differ ent values of R and
t =
Effect of Warm Ionized Plasma Medium on Radiation Properties of Mismatched Microstrip Termination
186
better than an open circuit discontinuity because total
radiated power is less for matched termination than an
open ended termination. More radiations from any dis-
continuity cause less utility of that structure. Hence, a
matched termination is suitable with an antenna in
plasma media though open circuit discontinuity suits
more in free space. A theoretical effort is made here in
this communication which requires experimental verifi-
cation, though simulation of plasma media in laboratory
is very difficult.
REFERENCES
[1] R. E. Post and D. T. Stephenson, IEEE Transactions, AP
P29,129 P133, 1981.
[2] J. R. James and G. J. Wilson, Microwave option and
acoustics, 1,165 P174, 1974.
[3] K. R. Carver and J. W. Mink, IEEE Transactions, AP29,
1024, 1981.
[4] A. G. Derneryd, IEEE Transactions, AP24,846851,
1976.
[5] M. D. Abouzahra and L. Lewin, IEEE Transactions, MTT
P27,722 P723, 1979.
[6] D. Bhatnagar, Journal of Inst. Electronic and Telecom.
Engrs. 38, 13 P16, 1992.
[7] M. V. Schneider, Bell System Tech. J. 48, 1421 P1444.
1969.
[8] A. M. Salem, D. Bhatnagar, J.of Plasma Physics, Vol.56,
Pt.1, 25 P34, 1996.

187
Fault Diagnosis Based on ANN for Tur n-to-Tur n
Shor t Cir cuit of Synchr onous Gener ator Rotor
Windings
H. Z. MA, L. PU

Electrical Engineering Department, Hohai University, Nanjing, China.
Email: hhumhz@163.com, pu.lan@hotmail.com

Received March 15
th
, 2009; revised June 8
th
th
, 2009.
ABSTRACT
Rotor winding turn-to-turn short circuit is a common electrical fault in steam turbines. When turn-to-turn short circuit
fault happens to rotor winding of the generator, the generator terminal parameters will change. According to these
parameters, the conditions of the rotor winding can be reflected. However, it is hard to express the relations between
fault information and generator terminal parameters in accurate mathematical formula. The satisfactory results in fault
diagnosis can be obtained by the application of neural network. In general, the information about the severity level of
the generator faults can be acquired directly when the faulty samples are found in the training samples of neural net-
work. However, the faulty samples are difficult to acquire in practice. In this paper, the relations among active power,
reactive power and excitation current are discovered by analyzing the generator mmf with terminal voltage constant.
Depending on these relations, a novel diagnosis method of generator rotor winding turn-to-turn short circuit fault is
proposed by using ANN method to obtain the fault samples directly, without destructive tests.
Keywords: Generator, Rotor Winding, Turn-to-turn Short Circuit, ANN, Diagnosis
1. Intr oduction
The statistical material form China Electric Power Re-
search Institute indicated that the rotor winding turn-t-
o-turn short circuit is a common electrical fault in a gen-
erator [1,2]. However, minor turn-to-turn short circuit
will not affect the normal operation of generator unit, so
it is often ignored. But if this fault develops, something
serious will appear, such as remarkable increasing of ro-
tor current, higher temperature of winding, deceasing of
reactive power, distortion of voltage, vibration of gen-
erator unit and many other mechanical faults. Therefore,
estimation of the early signs of failure severity and its
develop trends can be made based on the identification of
the faults early signals, and this task has gradually be-
come important in condition-based maintenance of gen-
erators [2,3].
At present, there have been many scholars studying in
the monitoring of rotor winding turn-to-turn short circuit
worldwide [25]. Albright proposed differential search-c-
oil test method: Its diagnosing effect is good for a gen-
erator under on-load and three phase short circuit, but
one-time location is difficult to make under load and it is
not sensitive to minor turn-to-turn short circuit. Russian
scholar B. T. Carsman proposed to detect turn-to-turn
short circuit based on the circulating current in stator par-
allel branch, but this method depends on the structure of
stator winding. Travelling-wave method based on online
diagnosing technique for rotor winding turn-to-turn short
circuit is immature. The alternating impedance method
and loss method are often-adopted in experiment, but the
method can not give an accurate conclusion in monitor-
ing miner rotor winding turn-to-turn short circuit all the
time. Further more, it is difficult to realize with the affec-
tion of the factors like slot wedge etc. [68].
This paper analyses the fault mechanism and mmf
(magnetomotive force) for generator rotor winding
turn-to-turn short circuit. It discovers that when the ma-
chine terminal voltage is in the condition of constant,
there exist certain relations among active power, reactive
power and field current. Thus it finds a kind of experi-
ment on the electrical engineering which does not need to
carry on a destructiveness experiment, but can obtain
fault sample. And then it makes use of artificial neural
network to carry out fault diagnosis for generator rotor
winding turn-to-turn short circuit.
*
This work project was supported by National Natural Science Foundation
of China (No. 50477010), China Education Ministrys Foundation Program
for Excellent Young Teachers in Universities and Importance Natural Sci-
ence Foundation of Hohai University.
Fault Diagnosis Based on ANN for Turn-to-Turn Short Circuit of Synchronous Generator Rotor Windings 188
2. Causes of Rotor Winding Tur n-to-Tur n
Shor t Cir cuit
The causes of rotor winding turn-to-turn short circuit
mainly include manufacture and operation. For example,
the mounting of rotor end winding is not firm; spacer
block becomes loose; the trimming of leading wire sol-
dered joint is not good; there are remaining metal scales
inside rotor protective ring; the dynamic forces such as
centrifugal force cause displacement deformation of
high-speed rotating rotor winding; the choking of rotor
winding causes local overheating, which makes turn-to
-turn insulation to burn down.
Otherwise, while the generator is operating or con-
verting from the static state to the dynamic state, due to
the abrasion of the turn-to-turn insulation or the relative
dislocation caused by the relative motion between the
turns in rotor, the turns may contact each other. When
this fault has developed to a certain extent, the turn-to-
turn shot circuit will happen. As a result of the emergence
of this, the Effective magnetic field of generator will de-
crease, and the generator reactive power will be affected.
Those lead to imbalance in the magnetic circuit which
causes vibration, and then monopole potential and
monopole current will be produced to magnetize the
generator shaft seriously. In addition, partial over heating
in fault point may be extended to grounding fault in rotor
windings.
3. Mmf Based Analysis of Electr omagnetic
Characteristics of Generator Rotor Winding
Turn-to-Turn Short Circuit
3.1 Analysis of Mmf under Rotor Winding
Tur n-to-Tur n Shor t Cir cuit
Spatial distribution of mmf in rotor windings is shown in
Figure 1 Just like Figure 1(b), while the generator units
are operating normally, the spatial distribution of Mmf is
trapezoidal-like, ignoring the minor intermittent of mmf
which is caused by the grooves. The mmf will loss par-
tially, while short-circuit happens in rotor windings. This
kind of loss will result in partial loss of shorted magnetic
pole, so that the average and amplify of shorted magnetic
pole will decrease as shown in Figure 1(c). Therefore, the
spatial distribution of mmf in the turn-to-turn short circuit
can be considered as that in demagnetization. So the
equivalent effect of short-circuit can be considered as a
mmf with the opposite direction adding on the main mmf
of short circuit.
The mmf of rotor winding under normal condition is
represented by
0
F , the mmf caused by short-circuited
turns is represented by F A , after short circuit, the rotor
mmf is
0
'
F F = F A

Figur e 1. Spatial distr ibution of mmf in r otor windings

a
F
1 f
F
f
F
o
F
'
o
F
'
a
F
a
F
o
E
a
r I
o
x I j
|
0
E
Figur e 2. Magnetic-electr onic potential vector of gener ator s

consider ing satur ation

The mmf-emf vector diagram of non-salient poles gen-
erator considering saturation is shown in Figure 2.
The air-gap mmf fundamental component F
o
is esta-
blished by exciting mmf fundamental component
1 f
F

and armature reaction mmf fundamental component
a
F
,
i.e.
1 f a
F F F
o
= +

(1)
Reduced to the exciting mmf wave, it can be got:
'
f
'
a
F F F
o
= +

(2)
where,
f f f
F w I = , is the turn number of rotor winding;
is exciting current; the phase angle of
f
w
f
I
, smaller than the former value. 1

1.35
w
a
w k I
F
p
=
Fault Diagnosis Based on ANN for Turn-to-turn Short Circuit of Synchronous Generator Rotor Windings
189
in time-space vector diagram is the same as I , is the
1
w
series number of stator winding phase, is the coeffi-
cient of stator winding.
w
k
Suppose terminal voltageU, active output and reac-
tive output are invariants, then stator current
P
Q I and
power factor angle become invariable. And x
o
has
little relation to the saturation level. (r ) a I E U jxo
o

= + +
.
Then the angle between E
o
'
and is unchangeable. So
that the angle between
I
F
o
and
'
a
F
will not change. Be-

cause of this,
f
F
is unchangeable. If the working condi-

tion of a generator before and after rotor winding fault
remains unchanged, in order to satisfy the air-gap com-
posite flux condition,
f
I will increase, but does not
change, from this it can be see: the relation betweenU,
, , not only manifest the state of the rotor
winding, but also reflects the effective turn number of the
rotor winding.
f
w I
f
P Q
f
I
3.2 Impact of Exciting Cur r ent on Rotor
Winding Turn-to-Turn Short Circuit Diagnosis
The relation betweenU , P , Q ,
f
I can manifest the state
of the rotor winding. The exciting current of a gen-
erator under normal operation can be calculated by the
mathematical Equation, and then compared with meas-
ured exciting current , the existence of rotor winding
turn-to-turn short circuit can be determined, furthermore
the calculation of relative deviation
0 f
i
fc
i
0
0
% o
fc f
f
i i
i
=
can be used to estimate the severity of the fault. 100%
4. ANN Diagnosis Method for Gener ator
Rotor Winding Tur n-to-Tur n
Shor t Cir cuit
Recently it is mainly according to the measurement of the
generator terminal parameters which namely generator
active power , reactive power , the generator ter-
minal voltage U, current , field voltage and other
generator parameters, uses a formula calculation to ac-
quire the field current
P Q
I
0 f
I which operates under nor-
mally flows, and then compare the measurement of actual
electric current
'
f
I with to diagnose rotor Winding
Turn-to-turn Short Circuit fault of generators.
0 f
I
This kind of method needs to consider the influence of
magnetic field saturation, and it needs accurate mathe-
matics model and the parameter of generator in the
meantime. The parameter of generator will also have va-
riety change along with the operating way and the variety
change of operating conditions. The accuracy of online
recognition is not very high, therefore there exists a cer-
tain error margin.
The artificial neural network (ANN) does not need ac-
curate mathematics model and the detailed parameters of
generator, and it has no interference to the operation of
generator in the meantime. It only needs to measure the
generator terminal parameters accurately, and depends on
a great deal of training samples. Through sufficient net-
work trains, diagnosis can be directly carried out of the
faults operated under dissimilar ways. By having faulty
sample, we can not only diagnose the faults, but also es-
timate the seriousness of short-circuit.
The terminal voltage U of generator is generally a
rated voltage, which could be supposed to be constant.
According to the analysis of the basis generator magnetic
field, certain , Q will correspond to certain P
f
F , namely
certain of . So the relation of , and
f f
w I P Q
f
I can
reflect the turn-to-turn short circuit fault, with the gen-
erator parameter , Q , P
f
I as the ANN importation, and
circles of turn-to-turn short circuit have the percent of
total number of full circles o % as the output.
The key of the fault diagnosis which carries on with
ANN to is to obtain the train samples. The selection of
normal samples can take in various samples in normally
operation in the P-Q diagram of generator, but in the ac-
tual power station for guaranteeing the ergodicity of
samples, we could detect the generator parameters with
long hours at normal operation conditions.
In order to estimate the severity of generator faults and
the number of short-circuit turns, the faulty samples of
generators are needed. Relatively speaking, to acquire the
samples of the generators under normal operation condi-
tion is still easy, but to obtain generator fault samples is
usually very difficult. The general method is to do a de-
structive experiment in dynamic simulation laboratory,
short connecting several turns of the rotor windings of
generator factitiously. This kind of method can barely be
carried out on the engineering.
This paper uses the method of balancing the mmf to
obtain fault sample of the generator under turn-to-turn
short circuit of rotor winding. Suppose turn-to-turn short
circuit fault occurs to generator at the rated condition,
before and after the short-circuit, , ,U are constant.
By analysing the magnetic field we can know the mag-
netic field will maintain constant, and suppose the
short-circuit turns of the total rotor winding number is
P Q
f f
w I
o %, after the fault, the field current is:
1
1 %
f fN
I I
o
=
(3)
where, the rated value of the field current.
fN
I
Changing the number of short-circuit turns, we will get a
series of faulty samples. And also we can acquire faulty
samples under different operation conditions.
5. Simulation
To verify the credibility of this method, we can adopt
cultural heritage [9] generator parameters, see Table 1;
and then take its normal operating samples. The rated
operate conditions of the fault sample can be seen in
Table 2. In Table 2, % is the short-turns as percentage
of the total rotor winding turns number. We can carry out
network training, and then carry out fault diagnosis; the
fault samples can be seen in Table 3. From Table 3,
%(actual) is the factitiousness number of short-turns as
percentage of the total number of rotor winding turns in
dynamic simulation laboratory. a%(emulation) is the
number of short-turns simulated by computer as percent-
age of the total number of rotor winding turns.
The back propagation artificial neural network(BP) is
used here, and the active function adopts S type function.
The BP network adopts 3 layers. As it is shown in Figure
3, the first layer has three importation nodes, hide layer
has four nodes, and exportation layer has one node. The
generator parameters contain active power, reactive
power, exciting current, and all these parameters are nor-
malized values. Table 3 shows the results simulated by
MATLAB. According to this, the actual results con-
firmed the BP network diagnoses.
In a word, this method has realized the direct acquisi-
tion of faults seriousness and the solution of fault sam-
ples acquisition which are difficult to get.
In particular process, the current and the terminal
voltage of stator windings are measured by CT (Cur-
rent Transformer) and PT(Potential Transformer), and the
exciting current
I
U
f
I is measured by current converter.
Here, select the CT and PT used to measure system pa-
rameters as the measure equipment. The generator capac-
ity and voltage level decide the selection of them. And
the hall current converter produced by Swiss company
LEM is used to measure in rotor winding. The pa-
rameters andU are used to calculate both the active
power and reactive power Q .
f
I
I
P
After training, this method is applied to all kinds of
load. By measuring andU of normal generator in all
running states, we can obtain the corresponding
andQ by calculating. The training of ANN needs ,
and to obtain the relationship among the parame-
ters( , and
I
P
Q
P
f
I
P Q
f
I ) in all the running states.
5. Conclusions
While ANN is applied to the fault diagnosis of generator,
the most difficult part is the acquirement of the

Figur e 3. Schematic diagr am of ANN

Table 1. Par ameter s of synchr onous machine
Model number MJF-30-6
Rated Voltage/V 400
Rated Cur r ent/A 43.3
Power Factor 0.8
Rotor Cur r ent/A 2
f w /tur n 100

Table 2. Diagnostic patter ns in neur al tr aining
P(*)
Active
Power
Q(*)
Reactive
Power
f
I
(*)
Exciting
Cur r ent
%
Shor t-tur n Number as
Per centage of Total
Winding tur ns Number
in Rotor
1 1 1.053 5
1 1 1.111 10
1 1 1.25 20
1 1 1.429 30

Table 3. The compar ison between the output of the neur al
networ k and the actual output
P(*)
Active
Power
Q(*)
Reactive
Power
f
I
(*)
Exciting
Current
o%(actual)
Short-turn Number as
Percentage of Total
Winding Turns Number
in Rotor
o%(emulation)
Short-turn Num-
ber as Percentage
of Total Wind-
ing Turns Num-
ber in Rotor
0.42555 0.51906 0.963 1.21 2.56
0.4295 0.5102 0.9771 3.91 4.86
0.4289 0.5054 0.99832 6.07 7.00
0.4325 0.5001 1.02026 10.06 9.48
0.4309 0.4899 1.06012 12.93 13.32
0.4168 0.4678 1.0765 14.83 15.50

fault samples among training samples. This paper analy-
ses generator turn-to-turn short circuit of rotor winding
fault and according to the certain operation of generator,
namely active power , reactive power Q , terminal
voltageU keep constant, and the field current in
reases, but the mmf
P
f
I
f f f
F w I = maintains constant.
Faulty samples are obtained through direct calculation,
turn-to-turn short circuit fault of rotor winding is diag-
nosed by making use of artificial neural network, and we
could obtain seriousness level of fault information di-
rectly. That method can avoid damage experimenting and
it is convenient in engineering.
The shortcoming of this method is that it could not lo-
cate the faulty position. After diagnosing and confirming
the existence of the fault, if we assist it with other ways
such as traveling wave method, location of fault can be
carried out.
REFERENCES
[1] W. J. Wang and L. Gui, Present situation and improve-
ment of relay protection for large sized generator trans-
former set of 600~1000MW, Huadian Technology, Vol.
30, No. 1, pp. 58, 2008.
[2] W. Q. Li, Prevention and fault diagnosis of turbogen-
erator, Beijing: Chinese electric power publisher, 2002.
[3] S. Wan, H. M. LI and Y. G. LI, Analysis of generator
vibration characteristic on rotor winding inter-turn short
circuit fault, Proceedings of the CSEE, Vol. 25, No. 10,
pp. 122126, 2005.
[4] H. W. Fang, C. L. Xia and J. Xiu, Analysis of generator
electro-magnetic torque on armature winding inter-turn
short circuit fault, Proceedings of the CSEE, Vol. 27, No.
15, pp. 8387, 2007.
[5] G. G. Mao, Reasons of failures occurred in large capac-
ity turbogenerators in China, Power System Technology,
Vol. 24, No. 11, pp. 17, 2000.
[6] R. J. Streifel, R. J. Marks II, and M. A. EI-Sharkawi, De-
tection of shorted-turns in the field winding of turbine
-generator rotors using novelty detectors development and
field test, IEEE Trans on Energy Conversion, Vol. 11,
No. 2, pp. 312317, 1996.
[7] A. S. Kulkarni, M. A. El-Sharkawi, R. J. Marks II, G.
Andexler, X. Jian, and I. Kerszenbaum, Development of
a technique for on-line detection of shorts in field wind-
ings of turbine-generator rotors: Circuit design and test-
ing, IEEE Trans on Energy Conversion, Vol. 15, No. 1,
pp. 813, 2000.
[8] S. E. Guttormsson, R. J. Marks II, and M. A. EI-Sharkawi,
Elliptical novelty grouping for on-line short-turn detec-
tion of excited running rotors, IEEE Transactions on En-
ergy Conversion, Vol. 14, No. 1, pp. 1622, 1999.
[9] D. W. Auckl, I. E. D. Pickup, and R. Shuttleworth,
Novel approach to alternator field winding inter-turn
fault detection, IEE Proc., Gener., Transm., Distrib., Vol.
142, No. 2, pp. 97102, 1995.

An Appr oach to Har monic State Estimation of
Power System
Niancheng ZHOU
1
, Li LIN
1
, Jizhong ZHU
1,2

1
State Key Laboratory of Power Transmission Equipment & System Security and New Technology,

Chongqing University, Chongqing,
China;
2
AREVA T&D Inc., 10865 Willows Rd. NE, Redmond, USA
Email: jizhong.zhu@areva-td.com

Received March 17
th
, 2009; revised June 15
th
nd
, 2009.
ABSTRACT
This paper presents the harmonic state estimation (HSE) based on the Total Least Squares (TLS) through comprehen-
sively considering the harmonic network parameter error and measurement system error. The proposed approach is
tested on the IEEE 14bus harmonic testing system. The satisfied results are obtained.
Keywords: State Estimation, Harmonic Analysis, Measurement Errors, Total Least Squares, Probability, Uncertainty
1. Intr oduction
The harmonic state estimation (HSE) in power system
estimates the harmonic state in whole network according
to the measured values of limited points

[13]. The prob-
lem of harmonic state estimation was first proposed by
Heydt in 1989 [1], where HSE is handled as an inverse
problem of harmonic power flow, and the adopted algo-
rithm for estimating the harmonic state and identifying
the harmonic source is least square estimation. Since it is
difficult to obtain accurately all system parameters for a
non-fundamental frequency harmonic network, the pa-
rameter error always exists. However, the existing HSE
methods only consider the harmonic measurement error,
and ignore the effect of parameter error to harmonic state
estimation, which is addressed in this paper.
2. HSE Based on TLS
Since the parameter error always exists in non-funda-
mental frequency harmonic network, the mathematical
model of HSE should consider not only the impact of
measurement error, but also the impact of parameters
error. Therefore, the HSE mathematical model can be
expressed as:
| |
true true
= + = Hx H x z + z v c = (1)
where, H is the parameter matrix (mn) containing the
relationship between the measured values and the state
variables. x is the undetermined state vector (n1). H
true

is the parameter matrix containing the truth values. v is
the parameter error matrix (mn). z
true
is the measure-
ment vector (m1) related to the truth values, which are
unknown variables. is the measurement error vector
(m1). z is the measurement vector (m1).
The harmonic state estimation Equation (1) is an over-
determined linear system. In this paper, the TLS method
is proposed to solve such HSE problem. The estimation
method of TLS is trying to estimate the noise matrix v
and the noise vector to meet the exact solution of linear
system. Select v and to make a minimum of
2
2
F
1 1 1
[ ]
m m n
i
i i j
v c
= = =
= +
2
ij
v (2)
where, is the element of matrix v, and
i
is the i-th
element of .
ij
v
Let | | H H = z , Equation (1) can be written as fol-
lows:
TLS
- 1
( [ v]) H
x
c
(
0 =
(

(3)
where, 0 is an m-dimension vector in which all the ele-
ments are 0.
In general, because of the existence of noise, the aug-
mented matrix H is full rank. If m>n+1, the rank of ma-
trix H is (n+1). Using the singular value decomposi-
tion (SVD) method, the matrix can be expanded as fol-
lows:
1
1
[ , ]
n
T
i i i
i
H z H u w o
+
=
= =

(4)
AnApproachtoHarmonicStateEstimationofPowerSystem
193
where,
i
is the singular value of matrix H , arranging in
the decreasing order of values, u
i
and w
i
are left singu-
lar vector containing m elements and right singular vector
containing (n+1) elements, respectively.
Lets define rank approximate value as the minimum
of the sum of various perturbations in matrix H . It is
given as follows [4]:
1
n
T
i i i
i
H u w o
=
=

(5)
Moreover, error matrix
[ v] c
is given as follows
1 1 1
[ ]
T
n n n
u w c v o
+ + +
= (6)
From Equations (3) and (5), we can get
1
TLS 1,1
- 1
n
n
w
+
+
(
=
(

w
x
(7)
According to [4], when the noise sequence meets the
central limit theorem conditions of independent identical
distribution, the standard TLS estimation is unbiased.
For the h-th harmonic, the harmonic power flow Equa-
tion can be expressed as follows:
( ) ( ) ( ) h h h =

Y V I (8)
In this paper, the node harmonic voltage vector
is used as measuring point, and the node harmonic injec-
tion current is used as state variables. If different
measurements and state variables are selected, the similar
state estimation expression can still be obtained through
the matrix transformation. If parameters error and meas-
urement error are considered simultaneously, that is,
harmonic impedance matrix error and harmonic node
voltage measurement error are considered together, the
relationship between measurements and state variables
can be obtained as follows:
( ) v h
( ) I h
true true
( ) ( ) ( ( ) ( )) ( ) V h h Z h v h I h + = +

c (9)
According to Equations (37), the estimated value of
the node harmonic injection current, ( )
TLS
I h
, can be com-
puted. Thus, the harmonic state estimation is solved.
3. Simulation Results and Analysis
The proposed TLS based HSE approach is examined us-
ing the IEEE14 node harmonic test system, which in-
cludes two harmonic sources, one is high voltage direct
current (HVDC) transmission system's terminal located at
node 3, and another is static var compensator (SVC) lo-
cated at node 8. The node harmonic voltages located at
buses 212 are selected as the measurement data.
The following several combination modes are used for
simulation analysis, respectively:
1) Case 1: Both measurement error and parameters er-
ror obey normal distribution with zero mean and different
variances;
2) Case 2: Measurement error obeys uniform distribu-
tion of zero mean, parameters error obeys normal distri-
bution of zero mean;
3) Case 3: Measurement error obeys normal distribu-
tion of zero mean; parameters error obeys uniform dis-
tribution of zero mean.
Assume that the errors are independent. The sampling
frequency is selected as 10000. In order to do comparison
easily, the three-phase harmonic simulation is computed
to obtain the true value, which is used to compare with
the estimated value of harmonic state estimation.
During the simulation, superimpose a random error at
each node harmonic voltage and each element of har-
monic impedance matrix. The errors are produced by the
normrnd function in Matlab. The variance is selected as
1% of the absolute value of each element in the matrix.
Due to the limitation of the space, this paper only gives
the analysis of case 2. In this case, it is needed to deter-
mine the upper and lower limit of the uniform distribu-
tion of the measurement error U[a,b]. Considering that
the measurement error is the random signal, which has
the similar definition with the noise in signal processing
[5], we determine the values of a and b through the use of
the signal-to-noise ratio (SNR) in signal processing.
101 ( / )
S U
SNR g P P = (10)
where, P
S
is the maximum true value in various meas-
urements. P
U
is the variance of measurement error, which
is the uniform distribution.
In this paper, the signal-to-noise ratio SNR is selected
as 20, together with the error is zero mean (that is, a =
b). Then, the upper and lower limits of the uniform dis-
tribution can be obtained through the following Equation:
/10 2
/ 10 ( ) / 12
SNR
U S
P P b a = = (11)
Thus, the random sequence of measurement error can
be obtained according to the upper and lower limit of
uniform distribution. Then, the measurement value in-
cluding uniform distribution error can be obtained by
superimposing it to the true value.
The results are shown in Figures 14. Figure 1 shows
the error cumulative probability density curve of the cal-
culation results. Node harmonic injection current and
harmonic voltage estimation are shown in Figures 2 and 3.
The probability density curves of harmonic injection cur-

Figur e 1. Har monic cur r ent injections estimation r elative
er r or at bus 3
AnApproachtoHarmonicStateEstimationofPowerSystem
194

Figur e 2. Har monic cur r ent injections

Figur e 3. Har monic bus voltages
rent estimation values are shown in Figure 4. The results
show that the probability density curves of all harmonic
injection current estimation values obey normal distribu-
tion approximately. For the low-order harmonic, its con-
fidence interval of probability density curve is relatively
wide.
4. Conclusions
This paper has studied the harmonic state estimation

Figur e 4. The pr obability density cur ve of the har monic
cur r ent injections estimated value at bus 3

(HSE) through considering comprehensively the meas
urement error and parameters error. The proposed
method is the Total Least Squares (TLS) and statistical
approach. The details of HSE model based on TLS are
presented, and error analysis for the HSE is conducted by
using the probability density function and cumulative
probability density function. The proposed method and
algorithm are tested on IEEE 14-node harmonic network.
The simulation results show the effectiveness and cor-
rectness of the paper.
REFERENCES
[1] G. T. Heydt, Identification of harmonic sources by a
state estimation technique, IEEE Trans., Power Delivery,
Vol. 4, No. 1, pp. 569576, 1989.
[2] A. P. S. Meliopoulos, F. Zhang, and S. Zelingher, Power
system harmonic state estimation, IEEE Trans., Power
Delivery, Vol. 9, No. 3, pp. 17011709, 1994.
[3] Z. P. Du, J. Arrillaga, N. Watson, and S. Chen, Identifi-
cation of harmonic sources of power systems using state
estimation, IEE Proceedings Generation, Transmission
and Distribution, Vol. 146, No. 1, pp. 712, 1999.
[4] C. Fu, The probability distribution of measurement er-
rors, Journal of Chengdu Institute of Radio Engineering,
Vol. 3, No. 3, pp. 5665, 1981.
[5] G. S. Hu, Digital Signal Processing, Beijing: Qing Hua
University Publishing House, 2004.

9771942073000 03

MPPT

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

MPPT

Caricato da

Copyright:

Formati disponibili

J.

Electromagnetic Analysis & Applications, 2009

in one second we set

table C0Discharge=Table [{t,N[ /.t1./6]}, {t,0,1,

s34=Plot [{1- /.t1./6,Q [t]/. solQUniformRota-

2.2 In Electr oacoustic Mode

r adiation patter ns for 1.0 A = and 0.2 A = , 0.7 , R t = =

Figur e 2. Magnetic-electr onic potential vector of gener ator s

, smaller than the former value. 1

will not change. Be-

is unchangeable. If the working condi-

Potrebbero piacerti anche