12th IEEE International Conference on Control & Automation (ICCA)

Kathmandu, Nepal, June 1-3, 2016

Modified Winding Function Approach to Stator Fault Modelling of

Synchronous Generator
Padmanabhan Sampath Kumar1 , Yejin Chen1 , Meng Yeong Lee2 , Sivakumar Nadarajan2 , Lihua Xie1

Abstract— Stator winding insulation failure has been one of low computational complexity and the accuracy required for
the major faults in electrical machines due to the degradation electrical transient simulation studies [7]. The low computa-
of winding insulation materials. The turn-to-turn short circuit tional complexity is due to the approximation of the windings
fault indicates an early stage of insulation failure. This fault
produces distortion of space harmonics in the air gap magnetic being sinusoidally distributed and symmetric. The trade off
field. Capturing these harmonics and reflecting their effect is that they cannot account for higher order harmonics of the
on the measurable quantities from stator would improve the winding and permeance distribution [8]. Moreover, when a
accuracy of the model representing the machine under faulty TTSC fault occurs, the symmetry in the winding distribution
condition. In this paper an approach for simulating a salient would be lost and there will be a distortion in the space
pole synchronous generator with stator inter-turn fault using
a phase domain model is proposed. To account for the space harmonics of the machine [9] [10]. This effect could not be
harmonics in the model, Modified Winding Function Approach captured by the dq0 model effectively.
is utilized to compute the inductances of the machine winding. The abc-model or Phase Domain (PD) model is the
The developed model is compared with the experimental data base form of coupled circuit machine models. There is no
for verifying the accuracy of the developed model. transformation involved and all the variables are presented in
I. INTRODUCTION the natural frame of reference. The PD model does not have a
requirement that the windings be sinusoidally distributed and
Synchronous Generators (SG) are widely used for elec-
be symmetric. The abc model is more complex compared to
trical power generation in marine, aerospace and industrial
dq0 model because the inductance is a function of the rotor
applications. An internal fault in the SG could not only cause
position [7].
damage to the machine but also leads to failure which could
If space harmonics and asymmetries are to be included
cause loss of revenue. Hence, it is important to detect the
in the dq0 based model, the inductance matrices would no
faults occurring in the SG at the earliest stage possible so
longer be constants and would become functions of the rotor
that reliability and availability of the power generators can
angle. The key advantage of using dq0 over abc model is the
be improved [1] [2].
constant inductance matrices [11][12]. On the other hand, a
Stator winding fault is one of the most common faults in
PD model will provide a straight forward way to include
electrical generators [3]. The winding insulation subjected
asymmetries in the machine and model the internal faults
to high electrical and thermal stresses which degrade the
[7].
insulation gradually, leading to inter-turn short circuit within
The accuracy of the coupled circuit based machine models
the windings [4]. This causes the flow of high short circuit
depends on the inductances. The Winding Function Ap-
loop currents and heating the shorted portion of the winding.
proach (WFA) is a method for describing the MMF distribu-
This heating effect would further degrade the insulation of
tion of the winding in the air gap [13]. It has the ability to
the neighbouring windings [5].
account for the various effects such as the space harmonics,
The most widely used techniques for detecting stator
slotting effect, air gap permeance effect and slot skewing.
Turn-to-Turn Short Circuit (TTSC) faults is using the time
The WFA together with air gap permeance function can be
harmonics analysis of the stator terminal voltages and cur-
used to compute inductance of the machine’s windings.
rents. In order to develop condition monitoring systems for
In [8][10][14], the WFA with PD model is utilized to
assessing the faults at their incipient stages, a thorough
model the stator winding phase to ground faults. However,
understanding of the machine under the fault conditions
the inter-turn short circuit within a stator phase winding
is necessary. Developing models to replicate the machine’s
coil is not explored. In [15][16] a hybrid model has been
behaviour during the fault assists to study the characteristic
developed, in which the healthy machine was modelled in
fault signatures, select the crucial measurements and device
dq0 frame and TTSC fault was modelled using WFA. Using
methodologies required for close condition monitoring [6].
this model the space harmonics generated by the TTSC
The synchronous machines are usually simulated using
winding was accounted in the model but the space harmonics
dq0 based models as they are well suited in terms of their
of the healthy machine windings were not considered.
1 Padmanabhan Sampath Kumar, Yejin Chen and Lihua Xie are with In this paper, a generalized procedure for developing a PD
the Rolls-Royce@NTU Corporate Lab, c/o School of Electrical and model of a synchronous generator under stator TTSC fault is
Electronic Engineering, Nanyang Technological University, Singapore presented. The inductances required for the model is derived
padmanab003@e.ntu.edu.sg
2 Meng Yeong Lee and Sivakumar Nadarajan are with the Advanced directly from the winding arrangement and the geometrical
Technology Centre, Rolls-Royce, Singapore Pte., Ltd., Singapore aspects of the machine, hence the model is able account for

978-1-5090-1738-6/16/$31.00 ©2016 IEEE 161

the space harmonics of the machine. This is achieved by
utilizing Winding Function Approach. A fault model with ⎡ ⎤
Lasas Lasbs Lascs Laskd Laskq Lasf d
accurate account for the harmonics aids in the design of ⎢ Lbsas Lbsbs Lbscs Lbskd ⎥
Lbskq Lbsf d
⎢ ⎥
better harmonic based condition monitoring systems and in ⎢ Lcsas Lcsbs Lcscs Lcskd ⎥
Lcskq Lcsf d
L=⎢
⎢Lkqas
⎥
the development of model based fault diagnostic/prognostic
⎢ Lkqbs Lkqcs Lkqkd ⎥
Lkqkq
⎥ Lkqf d
schemes. In the developed model, the progression of the ⎣Lf das Lf dbs Lf dcs Lf dkd ⎦
Lf dkq Lf df d
TTSC fault from an incipient stage to a severe stage is Lkdas Lkdbs Lkdcs Lkdkd Lkdkq Lkdf d
simulated. The model is simulated in MATLAB/SIMULINK (2)
environment and validated using the measurements obtained Lxy indicates the inductance associated with the winding
from experiments. The organization and summary of the x and y. Here x,y {as, bs, cs, kq, f d, kd}
work done in this paper are as follows: The equation (1) represents the healthy synchronous gen-
• Development of a phase domain model of the SG under erator. The winding resistances are obtained from the test rig
stator TTSC fault (Section II) SG datasheet. The computation of the inductance matrix L
• Computation of winding inductance for the SG under is described in Section III.
healthy and TTSC fault using winding function theory B. Synchronous Generator with Stator TTSC faults
(Section III)
The stator of a three-phase synchronous generator with
• The experimental test rig used to emulate the SG TTSC
TTSC fault on phase ”as” is shown in Fig.1. Due to the
faults is presented (Section IV)
TTSC, the phase as is split into two portions; the healthy
• The developed fault model is validated with the exper-
portion as1 and the shorted circuited portion as2. The
imental results (Section V)
percentage of total number of the stator turns shorted is
represented by μ. Thus μNs and (1 − μ)Ns represents
II. M ODELLING OF SG WITH STATOR TTSC FAULTS the number of turns in the shorted and healthy portion
respectively. Here Ns denotes the total number of turns per
The following assumptions were made in developing the stator phase winding. The severity of the short circuit is
synchronous generator model represented using the short circuit fault resistance Rf . The
current flowing through Rf is if s
• Hysteresis and Eddy current losses are not considered The voltage vas across the phase as can now be spit into
• Saturation is not taken into account vas1 and vas2 with the relation vas = vas1 + vas2
• Magnetic circuit is assumed to be linear
• Damper winding is considered to be sinusoidal
• Short circuit fault impedance is resistive
• Stator slot harmonics are negligible

A. Healthy Synchronous Generator

The salient pole synchronous generator is modelled as

a 4-pole, three phase star connected stator winding as, bs
and cs. The dq axis notation is used for indicating the rotor
circuit winding with kq and kd denoting the q-axis and d-
axis damper winding respectively and f d indicating the field
winding along the d-axis [12].
The voltage equations of the machine in phase domain is Fig. 1. Synchronous Generator with stator TTSC in phase as
expressed as follows
The circuit equations of the machine in Fig.1 is repre-
sented in phase domain as follows
v = Ri + pλ (1)
   
 T v = R i + pλ (3)
v = vas vbs vcs vkq vf d vkd   T
 T v = vas1 vas2 vbs vcs vkq vf d vkd
i = −ias −ibs −ics ikq if d ikd 
  i = [−ias −(ias − if s ) −ibs −ics
R = diag rs rs rs rkq rf d rkd T
 T ikq if d ikd ]
λ = λas λbs λcs λkq λf d λkd = Li   
R = diag (1 − μ)rs μrs rs rs rkq rf d rkd
where p is the d/dt operator. The flux linkage is rep-   T
resented by λ with subscript indicating the corresponding λ = λas1 λas2 λbs λcs λkq λf d λkd
  
winding. λ =Li

162
 an additional winding to the normal machine winding. As

expected it can be observed from the matrix R that there
is an electrical coupling only between the phases as and

⎡ ⎤
as2 and from the matrix L it can be observed that there
⎢Las1as1 Las1as2 Las1bs Las1cs Las1kq Las1f d Las1kd ⎥
⎢ ⎥ is a magnetic coupling between faulted winding and all the
⎢ ⎥
⎢Las2as1 Las2as2 Las2bs Las2cs Las2kq Las2f d Las2kd ⎥
⎢ ⎥
⎢ ⎥ other windings of the machine. The key aspect in the model
⎢ Lbsas1 Lbsas2 Lbsbs Lbscs Lbskq Lbsf d Lbskd ⎥
⎢ ⎥
 ⎢ ⎥ formulation would be the computation of the elements in
L =⎢
⎢ Lcsas1 Lcsas2 Lcsbs Lcscs Lcskq Lcsf d Lcskd ⎥⎥ 
⎢ ⎥ the inductance matrix L . The accuracy of the model would
⎢ ⎥
⎢ Lkqas1 Lkqas2 Lkqbs Lkqcs Lkqkq Lkqf d Lkqkd ⎥
⎢ ⎥
⎢ ⎥ depend on these inductances.
⎢ Lf das1 Lf das2 Lf dbs Lf dcs Lf dkq Lf df d ⎥
Lf dkd ⎥
⎢
⎣ ⎦
Lkdas1 Lkdas2 Lkdbs Lkdcs Lkdkq Lkdf d Lkdkd III. SG I NDUCTANCE C ALCULATION
(4)

By adding row 1 and 2 of equation (3) and re-arranging The self and mutual magnetizing inductances of stator
by splitting the terms associated with ias and if s we get, a and rotor field winding circuits required by equation (6)
new set of machine equations [6] are computed using Winding Function Theory. The winding
function of a coil represents the MMF distribution of the
    winding when a unit current is flowing though it. The
v = R i + pλ (5)
WFA method takes into account the spatial distribution of
  T
v = vas vbs vcs vkq vf d vkd vas2 the stator and rotor windings and along with the air gap
  T profile, the inductance of windings can be computed. The key
i = −ias −ibs −ics ikq if d ikd if s
⎡ ⎤ assumptions made in WFA is that the back iron MMF drop is
rs 0 0 0 0 0 μrs negligible and the flux in the air gap is directed radially. The
⎢ 0 rs 0 0 0 0 0 ⎥
⎢ ⎥ WFA theory was first applied to uniform air gap machine
⎢ 0 0 rs 0 0 0 0 ⎥
⎢ ⎥ and later extended to non-uniform air gap machines and is
R =⎢ 0 ⎥


⎢ 0 0 0 rkq 0 0 ⎥ called as Modified Winding Function Approach (M-WFA)

⎢ 0 0 0 0 rf d 0 0 ⎥
⎢ ⎥ [17].
⎣ 0 0 0 0 0 rkd 0 ⎦ From M-WFA the inductance between two windings x and
μrs 0 0 0 0 0 μrs y can be computed using the following formula.
  T
λ = λas λbs λcs λkq λf d λkd λas2
  

2π
λ =L i Lxy (θr ) = μ0 rl Mx (φs , θr )My (φs , θr )g −1 (φs , θr ) dφs
0
Using the relation that the inductance of the shorted turns (8)
and healthy portion should be consistent with the normal

winding [4], we get the inductance matrix L M (φs , θr ) = n(φs , θr ) − M (φs , θr ) (9)
Lasas = Las1as1 + 2Las1as2 + Las2as2
Lasbs = Las1bs + Las2bs 1
M (φs , θr ) =
Lascs = Las1cs + Las2cs 2πg −1 (φs , θr )

2π (10)
Laskq = Las1kq + Las2kq n(φs , θr )g −1 (φs , θr ) dφs
Lasf d = Las1f d + Las2f d 0

Laskd = Las1kd + Las2kd μ0 – permeability of free space

r – mean air gap radius
⎡ ⎤ l – axial length of the rotor
⎢ Lasas Lasbs Lascs Laskq Lasf d Laskd Lasas2 ⎥ Mx – Modified Winding Function for coil x
⎢ ⎥
⎢ ⎥
⎢ Lbsas Lbsbs Lbscs Lbskq Lbsf d Lbskd Lbsas2 ⎥ My – Modified Winding Function for coil y
⎢ ⎥
⎢ ⎥

⎢ Lcsas
⎢
⎢
Lcsbs Lcscs Lcskq Lcsf d Lcskd Lcsas2 ⎥⎥
⎥
g −1 (φs , θr ) – Inverse air-gap function
L =⎢⎢ Lkqas Lkqbs Lkqcs Lkqkq Lkqf d Lkqkd Lkqas2 ⎥⎥ n(φs , θr ) – turns function
⎢ ⎥
⎢ ⎥
⎢ Lf das Lf dbs Lf dcs Lf dkq Lf df d Lf dkd Lf das2 ⎥ φs – angular position along the stator
⎢ ⎥
⎢ ⎥
⎢ Lkdas Lkdbs Lkdcs Lkdkq Lkdf d Lkdkd Lkdas2 ⎥
⎢
⎣
⎥
⎦
θr – angular position of the rotor
Las2as Las2bs Las2cs Las2kq Las2f d Las2kd Las2as2 f (x) – mean of f (x)
(6)

The voltage equation of the faulted portion is represented The inductances are computed from the turns function,
by winding function and the inverse air gap function. The
vas2 = −Rf if s (7) turns functions and winding functions are obtained from the
winding arrangement of the machine. The air gap function is
The equations (5) and (7) provide a straight forward way obtained by measurements carried out in the machine. The
for modelling the stator TTSC faults. From the equations procedure for computing these functions can be found in
it can be observed that the TTSC fault is modelled as [12].

163
 The equation (8) represents inductance calculation for
the continuous case. The discrete version of the inductance
computation equations suitable for computer simulation is
obtained using the formulation used in [18].

Fig. 4. Healthy stator inductance

Fig. 2. Stator turns function

Fig. 5. Healthy stator-field winding mutual inductance

(BLSG). The specifications of the machine is indicated in

Table I . The 3-phase stator has a double layer winding
arranged in 36 slots with each phase consisting of 4 series
connected coils. There are 14 turns/slot/phase with a total of
168 turns per phase in the stator. The main stator armature
of the BLSG had been rewound to accommodate tap outs
Fig. 3. Field Winding turns function and Air gap function for introducing different percentage of TTSC faults.

TABLE I
From the turns function of the stator windings and ro-
SG M ACHINE S PECIFICATIONS
tor field winding the inductances associated with the set
{as, bs, cs, as2, f d} are computed. The computed induc- Quantity Value
tances are shown in the Fig. 4 to 6
Power Rating 14 kVA
The damper windings were assumed to be sinusoidally
Voltage Rating 415 V
distributed and the inductances associated with the set of
No. of poles 4
coils {as, bs, cs, f d} and {kq, kd} were computed from the
Rated Frequency 50 Hz
machine’s dq specifications [12]. For computation of the
Cooling Method Air
inductance associated with the set {as2} and {kq, kd}, the
Insulation Class H
inductance corresponding to as was multiplied by a factor
of μ. For example, Las2kq = μLaskq
The stator and rotor inductances associated with the short The BLSG can be used for emulating TTSC faults with
circuit winding as2 for μ equal to 10% is shown in the μ = 3%, 5%, 10%, 15% and 20%. The tap outs were con-
Fig.6. The computed inductances are stored as a Look-Up nected to a resistor bank via relays. Based on the combination
Table (LUT) for different positions of the rotor angle θr of the relay being turned on, the short circuit resistance (Rf )
of the TTSC fault being emulated can be varied. Inter-turn
IV. E XPERIMENTAL S ETUP short circuit fault occurs among the windings of the same coil
The experimental validation of the developed model as each coil is insulated from the other coils in the slot and
was carried out using a Brushless Synchronous Generator the end winding region [19]. Taking this into consideration,

164
 simulated using the experimental setup and the developed
model. The third harmonic positive sequence voltage has
been observed to increase based on the severity of the
turn to turn short circuit fault [20]. Therefore it was used
as a fault indicator to validate the developed model. Fig.9
displays the third harmonic positive sequence voltage from
the experimental readings and the simulation results. We can
observe that the experimental and simulation results follow a
similar pattern. The deviation of the simulation results from
the experiments can be attributed to the following
• Inherent asymmetries in the test SG
• Assumption of sinusoidal damper windings

Fig. 6. Short circuit inductance • Saturation is not taken into account

• Dynamics of automated voltage regulator and brushless
exciter are not taken into account
the maximum value of μ for the machine under study is 0.25.
In this paper, 10% inter-turn fault (μ=10%) at phase as is
studied.
The experiments were carried out at a lower power rating
to ensure that no permanent damage is incurred to the
machine. The TTSC study was conducted for μ equal to
10% with Rf varying in an increasing steps of severity such
as Rf = ∞, 50Ω, 25Ω, 12.5Ω, 6Ω, 3Ω, 1.5Ω, 0.8Ω. Rf = ∞
represents machine under healthy state. The three phase sta-
tor voltages and currents were logged in PC using dSPACE
data acquisition system. The experimental setup is shown in
the Fig.7

Fig. 8. Spectrum of the stator terminal voltage

Fig. 7. Experimental Setup

V. S IMULATION AND E XPERIMENTAL R ESULTS

The machine equations for TTSC were simulated using
MATLAB/SIMULINK and compared with the data from the
experimental results.
The time harmonics of the stator terminal voltage for the
experimental machine and simulated machine are shown in
Fig. 9. Third harmonic positive sequence stator voltage
Fig. 8. It can be observed that for both the healthy and TTSC
10% case the developed machine model is able to account
for the harmonics as in the test SG machine. VI. C ONCLUSION
The progression of TTSC fault from healthy state (Rf = In this work we have presented a phase domain model
∞Ω) to severe fault (Rf = 0.8Ω) for TTSC μ = 10% was of a salient pole synchronous generator with stator turn

165
to turn short circuit fault. In order to account for space [8] A. B. Dehkordi, P. Neti, A. M. Gole, and T. L. Maguire, “Development
harmonics of the machine in the developed fault model, the and Validation of a Comprehensive Synchronous Machine Model for
a Real-Time Environment,” IEEE Transactions on Energy Conversion,
inductances of the healthy stator and stator with TTSC fault vol. 25, no. 1, pp. 34–48, 2010.
were computed using Modified-Winding Function Approach. [9] A. Megahed and O. Malik, “Synchronous generator internal fault com-
It was found that the developed model was able to come close putation and experimental verification,” IEE Proceedings - Generation,
Transmission and Distribution, vol. 145, no. 5, p. 604, 1998.
in replicating the experiment test rig SG behavior during the [10] S. Hemmati, S. Shokri Kojoori, S. Saied, and T. A. Lipo, “Modelling
healthy state and progression of the TTSC fault. Future work and experimental validation of internal short-circuit fault in salient-
will include incorporating brushless exciter for simulating the pole synchronous machines using numerical gap function including
stator and rotor core saturation,” Electric Power Applications, IET,
model for a Brushless Synchronous Generator and consider vol. 7, no. 5, pp. 391–399, 2013.
the actual damper inductances instead of sinusoidal approx- [11] L. Queval and H. Ohsaki, “Nonlinear abc-Model for Electrical Ma-
imation. The model will also be incorporating saturation chines Using N-D Lookup Tables,” IEEE Transactions on Energy
Conversion, vol. 30, no. 1, pp. 316–322, 2015.
effects to account for machine operation under full load and [12] P. C. Krause, O. Wasynczuk, S. D. Sudhoff, and S. Pekarek, Analysis
the localized saturation effects occurring during the turn to of electric machinery and drive systems. John Wiley & Sons, 2013,
turn short circuit faults. vol. 75.
[13] X. Luo, Y. Liao, H. Toliyat, A. El-Antably, and T. A. Lipo, “Multiple
ACKNOWLEDGMENT coupled circuit modeling of induction machines,” in Industry Applica-
tions Society Annual Meeting, 1993., Conference Record of the 1993
This work was conducted within the Rolls-Royce@NTU IEEE, Oct 1993, pp. 203–210 vol.1.
Corporate Lab with support from the National Re- [14] H. Jiang, R. Aggarwal, G. Weller, S. Ball, and L. Denning, “A new
approach to synchronous generator internal fault simulation using
search Foundation (NRF) Singapore under the Corp combined winding function theory and direct phase quantities,” in
Lab@University Scheme Electrical Machines and Drives, 1999. Ninth International Conference
on (Conf. Publ. No. 468), 1999, pp. 105–111.
R EFERENCES [15] N. Sivakumar, B. Bhangu, S. K. Panda, and A. K. Gupta, “Improved
Brushless Synchronous Generator model to study the turn-to-turn short
[1] I. Tabatabaei, J. Faiz, and H. Lesani, “Modeling and Simulation
circuit fault,” in 2014 IEEE PES General Meeting — Conference &
of a Salient-Pole Using Modified Winding Function Theory,” IEEE
Exposition. IEEE, jul 2014, pp. 1–5.
Transactions on Magnetics, vol. 40, no. 3, pp. 1550–1555, 2004.
[16] S. Nadarajan, S. K. Panda, B. Bhangu, and A. K. Gupta, “Hybrid
[2] N. Sivakumar, B. Bhangu, S. Panda, and A. K. Gupta, “Dynamic
model for wound-rotor synchronous generator to detect and diagnose
model of Brushless Synchronous Generator with turn-to-turn short
turn-to-turn short-circuit fault in stator windings,” IEEE Transactions
circuit fault for condition monitoring,” in 2013 IEEE ECCE Asia
on Industrial Electronics, vol. 62, no. 3, pp. 1888–1900, March 2015.
Downunder. IEEE, jun 2013, pp. 258–263.
[17] N. Al-Nuaim and H. Toliyat, “A novel method for modeling dynamic
[3] “IEEE recommended practice for the design of reliable industrial and
air-gap eccentricity in synchronous machines based on modified
commercial power systems,” IEEE Std 493-2007 (Revision of IEEE
winding function theory,” IEEE Transactions on Energy Conversion,
Std 493-1997), pp. 1–689, June 2007.
vol. 13, no. 2, pp. 156–162, 1998.
[4] S. Hemmati, S. Shokri, and S. A. Saied, “Modeling and simulation
[18] C. Mishra, A. Routray, and S. Mukhopadhyay, “A Computationally
of internal short circuit faults in large hydro generators with wave
Efficient Winding Function Method for Calculation of Inductances in
windings,” in Power Engineering, Energy and Electrical Drives (POW-
an Asymmetric Induction Motor,” Electric Power Components and
ERENG), 2011 International Conference on, May 2011, pp. 1–6.
Systems, vol. 35, no. 1, pp. 43–61, 2007.
[5] B. G. Gu, J. H. Choi, and I. S. Jung, “Development and analysis of
[19] G. Joksimovic and J. Penman, “The detection of inter-turn short
interturn short fault model of PMSMs with series and parallel winding
circuits in the stator windings of operating motors,” IEEE Transactions
connections,” IEEE Transactions on Power Electronics, vol. 29, no. 4,
on Industrial Electronics, vol. 47, no. 5, pp. 1078–1084, 2000.
pp. 2016–2026, 2014.
[20] J. Sottile, F. C. Trutt, and A. W. Leedy, “Condition monitoring of
[6] R. Tallam, T. Habetler, and R. Harley, “Transient model for induction
brushless three-phase synchronous generators with stator winding or
machines with stator winding turn faults,” IEEE Transactions on
rotor circuit deterioration,” IEEE Transactions on Industry Applica-
Industry Applications, vol. 38, no. 3, pp. 632–637, 2002.
tions, vol. 42, no. 5, pp. 1209–1215, 2006.
[7] L. Wang, J. Jatskevich, and H. W. Dommel, “Re-examination of Syn-
chronous Machine Modeling Techniques for Electromagnetic Tran-
sient Simulations,” IEEE Transactions on Power Systems, vol. 22,
no. 3, pp. 1221–1230, 2007.

166