1230 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO.

4, AUGUST 2019

Inter-Turn Fault Detection of Dry-Type Transformers

Using Core-Leakage Fluxes
Subash Chandar Athikessavan , Member, IEEE, Elango Jeyasankar, Member, IEEE,
Sai Srinivas Manohar , Member, IEEE, and Sanjib Kumar Panda , Senior Member, IEEE

Abstract—In this paper, a new technique for low-severity inter- Iea ∗ , Ieb ∗ and Iec ∗ Excitation currents in the primary
turn fault detection under online operating conditions of dry-type windings of limbs-a, -b and -c
transformers is proposed. The proposed methodology requires just
respectively.
two inexpensive flux sensors and is based on the fundamental com-
ponent of the core-leakage fluxes at the transformer’s outer limbs. ϕah ∗ and ϕch ∗ Core-fluxes generated by the
Theoretical analysis using the proposed analytical models and limb-a and limb-c winding MMFs
experimental validation on a 10-kVA experimental prototype trans- respectively under healthy operat-
former are carried out to evaluate the performance of the proposed ing conditions of the transformer.
methodology under various fault severities. The real-time fault-
ϕach ∗ and ϕbch ∗ Core-fluxes in limb-c due to limb-a
detection capability of the proposed technique is also evaluated
using a fabricated prototype health monitoring device, which re- and limb-b MMFs respectively un-
veals that as low as 2-turns bolted fault out of 150 turns can be der healthy operating conditions of
detected within 3 fundamental cycles. The methodology is immune the transformer.
to loading and supply voltage unbalance conditions and on par with ϕbah ∗ and ϕcah ∗ Core-fluxes in limb-a due to limb-b
the classical differential negative sequence current based technique
and limb-c MMFs respectively un-
with an additional advantage of 78% reduction in the total number
of sensors required for a three-phase three-winding transformer. der healthy operating conditions of
the transformer.
Index Terms—Power transformers, fault detection, mag- ϕaT h ∗ and ϕcT h ∗ Total core-flux in limb-a and limb-
netic flux leakage, transformer cores, fast Fourier transforms,
industrial power systems. c under healthy operation of the
transformer.
Ra , Rb and Rc Core-reluctances of limbs-a, −b
NOMENCLATURE and -c respectively.
ϕF I (healthy) Proposed inter-turn fault indicator
Ila ∗ , Ilb ∗ and Ilc ∗ Load currents in the primary wind-
under healthy operating conditions
ings of limbs-a, -b and -c respec-
of the transformer.
tively.
 ∗  ∗ Isc ∗ Inter-turn fault current in the trans-
Ila , Ilb and Ilc ∗ Load currents in the secondary
former’s winding.
windings of limbs-a, -b and -c re-
Nna  Total number of shorted turns in the
spectively.
limb-a primary winding.
Na , Nb and Nc Total number of turns in the pri-
RT F Core-reluctance of the limb-a due
mary windings of limb-a, -b and -c
to inter-turn fault.
respectively.
ϕaf −pr i ∗ and ϕcf −pr i ∗ Core-fluxes generated by the limb-
Na , Nb and Nc Total number of turns in the sec-
a and limb-c winding MMFs re-
ondary windings of limbs-a, -b and
spectively under inter-turn fault in
-c respectively.
the transformer’s primary winding.
ϕacf −pr i ∗ and ϕbcf −pr i ∗ Core-fluxes in limb-c due to limb-
Manuscript received April 20, 2018; revised August 17, 2018; accepted Octo- a and limb-b MMFs respectively
ber 17, 2018. Date of publication October 29, 2018; date of current version July under inter-turn fault in the trans-
23, 2019. This work was supported in part by the National Research Foundation former’s primary winding.
Singapore, Sembcorp Industries Ltd, and in part by the National University of
Singapore under the Sembcorp-NUS Corporate Laboratory. Paper no. TPWRD- ϕbaf −pr i ∗ and ϕcaf −pr i ∗ Core-fluxes in limb-a due to limb-
00411-2018. (Corresponding author: Sanjib Kumar Panda.) b and limb-c MMFs respectively
The authors are with the Sembcorp-NUS Corporate Laboratory, Department under inter-turn fault in the trans-
of Electrical and Computer Engineering, Faculty of Engineering, National
University of Singapore Singapore 117581 (e-mail:, [email protected]; former’s primary winding.
[email protected]; [email protected]; [email protected]). ϕaT f −pr i ∗ and ϕcT f −pr i ∗ Total core-flux in limb-a and limb-
Color versions of one or more of the figures in this paper are available online c under inter-turn fault in the trans-
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRD.2018.2878460 former’s primary winding.

0885-8977 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1231

ϕF I (faulty) ∗ Proposed inter-turn fault indicator A worldwide survey on the transformer failures revealed that
under turn fault conditions of the winding fault is the dominant fault which incurs 48% of the to-
transformer. tal transformer failures [1]. The low-severity inter-turn fault in
ϕaf −sec ∗ and ϕcf −sec ∗ Core-flux generated by the limb-a the transformer’s winding evolves into high-severity turn fault
and limb-c winding MMFs respec- which eventually leads to phase-to-ground fault [8], [9]. There-
tively under inter-turn fault in the fore, detecting the inter-turn fault at a low-severity level can
transformer’s secondary winding. prevent the catastrophic ground failure which may lead to sub-
ϕbaf −sec ∗ and ϕcaf −sec ∗ Core-fluxes in limb-a due to limb- stantial core damage and significant downtime for replacement
b and limb-c MMFs respectively of both core and windings [4].
under inter-turn fault in the trans- A comprehensive literature survey on existing inter-turn fault
former’s secondary winding. detection techniques for the transformer has been carried out
ϕacf −sec ∗ and ϕbcf −sec ∗ Core-fluxes in limb-c due to limb- and reported in this section. The Frequency Response Analysis
a and limb-b MMFs respectively (FRA) based winding anomaly detection techniques have been
under inter-turn fault in the trans- reported in [8]–[12]. The FRA is the most sensitive technique
former’s secondary winding. to detect any winding anomalies [8]–[12]. However, external
ϕaT f −sec ∗ and ϕcT f −sec ∗ Total core-flux in limb-a and limb- noises and measurement conditions highly influence the results
c under inter-turn fault in the trans- [10], [12]. Moreover, FRA techniques are equipment specific
former’s secondary winding. and measurement standards are available only for offline oper-
ϕF I l(faulty−pr i) ∗ and Proposed fault indicators under ating condition of the transformer [10], [12]. The magnetizing
ϕF I l(faulty−sec) ∗ inter-turn fault in the transformer’s currents based inter-turn fault detection of as low as 2-turns
primary and secondary windings (0.7%) and 1-turn (0.004%) faults have been reported in [13]
respectively. and [14] respectively. But, the reported online technique is ap-
kh and kf Mean value of the leakage coeffi- plicable only under no-load and light load such as 10% of the
cients at outer limbs-a and -c under full-load condition. The excitation currents and Flux Division
healthy and inter-turn fault condi- Measurement (FDM) based techniques to detect the short-
tion of the transformer respectively. circuited turns have been reported in [9] and [15] respectively.
Vpa ∗ , Vpb ∗ and Vpc ∗ Terminal voltages of the primary However, the techniques are offline, which require the trans-
windings. former to be disconnected from the power system network. The
Ipa ∗ , Ipb ∗ and Ipc ∗ Terminal currents of the primary inter-turn fault detection based on the neutral current of the star-
windings. winding transformer has been presented in [16], which can de-
Vs1a ∗ , Vs1b ∗ and Vs1c ∗ Terminal voltages of the secondary tect as low as 4-turns (3%) fault. But, the technique is applicable
winding-1. only to the transformers with star-winding in its primary-side,
Vs2a ∗ , Vs2b ∗ and Vs2c ∗ Terminal voltages of the secondary which are rarely used for electrical power distribution systems
winding-2. [2]–[6]. The Δ V-I locus-based technique, which can detect
Is1a ∗ , Is1b ∗ and Is1c ∗ Terminal currents of the secondary inter-turn fault from 13-turns (5%) fault has been discussed in
winding-1. [17]. However, 13-turn fault is a high-severity fault which is the
Is2a ∗ , Is2b ∗ and Is2c ∗ Terminal currents of the secondary consequences of low-severity turn faults such as 1 to 3-turns
winding-2. fault. The core- and winding- leakage-fluxes are used to detect
a Phase-displacement operator. inter-turn faults of transformers in [18] and [19] respectively.
ID N S C Differential negative sequence The proposed techniques in [18] and [19] requires six Search
current. Coils (SCs) per phase which result in complex inter-connections.
∗ Fundamental phasor. Furthermore, installation of eighteen SCs on the transformer re-
quires dismantling of primary and secondary windings of the
transformer [18]. Besides, the proposed techniques in [18] and
I. INTRODUCTION [19] are influenced by supply voltage unbalances and loading
HE dry-type transformers are widely used in electrical conditions. In [20], the winding vibration has been analysed us-
T power distribution systems of data centre, power plants,
commercial premises and in various industries such as chem-
ing machine learning techniques to detect inter-turn faults. How-
ever, deployment of vibration sensors on winding is not practi-
ical, paper, pulp, cement etc. [1]–[4]. They are also strongly cally plausible. Furthermore, multiple vibration sensors have to
preferred in transit applications for traction power supply and be mounted at each disc of the winding to detect the inter-turn
in power distribution systems of marine vessels and other off- fault which may occur at any of the discs. The differential cur-
shore floating facilities [5], [6]. Any catastrophic failure of the rents based negative sequence and Extended Park Vector (EPV)
transformer in the aforementioned applications leads to huge techniques for inter-turn fault detection have been reported in
economic loss and for marine floating facilities there are po- [21], [22] and [23] respectively. The differential currents-based
tential Safety Of Life At Seas (SOLAS) issues [7]. Therefore, techniques are sensitive to low-severity inter-turn faults and can
online Health Monitoring (HM) of transformers is of paramount detect inter-turn fault as low as 1% which corresponds to 4-turns
importance to mitigate any unexpected and unplanned outages fault [21] and [22]. The technique requires 3n Current Trans-
of the transformer. formers (CTs), where n is the total number of windings; which

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
1232 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO. 4, AUGUST 2019

transformer are given by,

Nb Ieb Rc Nc Iec Rb
ϕbah = and ϕcah =
RT RT
The total core-flux in the limb-a under healthy operation of
the transformer with (Na = Nb = Nc = N ) is given by (1).
ϕaT h = ϕah − ϕbah − ϕcah
N
= (Iea (Rb + Rc ) − Ieb Rc − Iec Rb ) (1)
RT
Similarly, total core-flux in the limb-c under healthy operation
Fig. 1. Equivalent circuits of the three-phase transformer under healthy
of the transformer is given by (2).
operating condition.
ϕcT h = ϕch − ϕach − ϕbch
N
in turn increases the total cost of the protection system. There- = (Iec (Ra + Rb ) − Iea Rb − Ieb Ra ) (2)
fore, differential current based inter-turn fault detection systems RT
are widely adopted in the industry for expensive transformers The proposed fault indicator under healthy operating condi-
of 10MVA rating and above [24]. tions of the transformer is represented by (3)
From the survey, it is evident that the existing techniques
ϕF I (healthy) = ϕaT h + ϕcT h
have limitations such as, can detect only high-severity faults,
applicable only under light-load/no-load and offline conditions, N
= (Iec Ra + Iea Rc − Ieb (Rc + Ra )) (3)
requires more number of sensors which lead to complex inter- RT
connections and adopted only for expensive assets. In this
Due to structural symmetricity, the reluctances of the trans-
paper, core-leakage flux based online technique to detect the
former’s outer limbs are almost the same (Ra = Rc = ).
low-severity inter-turn fault of the transformers has been pro-
Therefore, from (3) the proposed fault indicator under
posed. The proposed technique is inexpensive, requires just two
healthy operating conditions of the transformer is represented
flux sensors and can detect low-severity faults over various load-
by (4).
ing and supply voltage unbalance conditions. The paper also
presents a comparative study of the proposed and classical dif- N
ϕF I (healthy) = (Iec + Iea − 2Ieb ) (4)
ferential negative sequence current based technique which is RT
widely adopted in the industry.
Theoretically, the core flux at the yoke-center, ϕoT h is zero
[25]. However, due to reluctances asymmetricity, there is a fi-
II. CORE-FLUX BASED ANALYTICAL MODEL OF
nite value [25], which remains the same under inter-turn fault
THE TRANSFORMER
conditions of the transformer. This is because of the three-phase
The magnetic characteristics of the transformers under symmetricity (Iea + Ieb + Iec ≈ 0) which regulates the core
healthy and inter-turn fault conditions are analyzed using the flux at the yoke-center at a constant value. Therefore, the core-
proposed analytical models which are discussed in this section. flux at the yoke-center, ϕoT h is neglected in the proposed fault
indicator.
A. Core-Flux Under Healthy Operation of the Transformer From (4), it is evident that the magnitude of the proposed
Figs. 1(a) and 1(b) represent the electrical and magnetic fault indicator is dependent on number of turns (N ), excitation
equivalent circuits of the transformer under healthy operating currents (Iea , Ieb and Iec ) and core-reluctances (Ra , and Rc )
conditions. The core-fluxes under healthy operating conditions and is independent of the loading conditions.
are due to the excitation currents in the transformer’s primary
windings which are independent of the load in the secondary B. Core-Flux Under Inter-Turn Fault in the Transformer’s
winding. Primary Winding of Limb-a
The fundamental phasor of the core-flux generated by the Figs. 2(a) and 2(b) represent the electrical and magnetic
limb-a winding MMF under healthy operating conditions of the equivalent circuits of the transformer under inter-turn fault in
transformer is given by, the transformer’s primary winding of limb-a. The inter-turn
Na Iea N I (R + Rc ) fault in the transformer’s primary winding emulates the auto-
ϕah =   = a ea b , transformer with its secondary winding shorted as shown in
Rb Rc
Ra + Rb +Rc RT
Fig. 2(a). Therefore, the excitation currents in the primary
winding change to oppose the MMF generated by the shorted-
RT = Ra Rb + Rb Rc + Rc Ra
turn/turns at any of the primary windings. During inter-turn
The fundamental phasors of the fluxes in limb-a due to limb- fault, the voltage across the shorted-turn/turns reduces from the
b and limb-c MMFs under healthy operating conditions of the rated Volts/turn due to the drastic increase of turn-fault current

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1233

Fig. 2. Equivalent circuits of the three-phase transformer under inter-turn fault Fig. 3. Equivalent circuits of the three-phase transformer under inter-turn fault
in the primary winding of limb-a. ( I ∗ = Ie a − Is c ) and ( N a = N a n + N na  ). in the secondary winding of limb-a.

when compared to the terminal currents. Such a high value of Nb = Nc = N ) is given by (5).
current generates MMF to oppose the main MMF generated ϕaT f −pr i = ϕaf −pr i − ϕbaf −pr i − ϕcaf −pr i
by healthy winding as shown in Fig. 2(b). This means, theo-  
retically the flux entering the shorted-turn/turns should be zero N Rb (Iea − Iec ) + Rc (Iea − Ieb ) − NNn a  Isc (Rb + Rc )
as the induced EMF in the shorted-turn due to the excitation =
flux is zero [18], [19], [25]–[27]. Practically, the MMF gener- RT F (Rb + Rc ) + RT
ated by the shorted-turn/turns opposes the excitation MMF and (5)
results in high radial leakage-flux [25]–[27] across the shorted- Similarly, total core-flux in the limb-c under inter-turn fault
turns as shown by dotted lines in Fig. 2(b). Although the radial condition in the transformer’s primary winding is given by (6).
leakage-flux increases nearby the shorted turn, the linkage-flux
remains constant due to the auto-regulation of core-flux by the ϕcT f −pr i = ϕcf −pr i − ϕacf −pr i − ϕbcf −pr i =
 
transformer [26]. The shorted turn/turns increase the reluctances
N Ra (Iec −Ieb)+Rb (Iec −Iea)+RT F (Iec − Ieb)+ NNn a  Isc Rb
faced by the excitation flux which in turn changes the value of
the excitation currents and fluxes at different limbs of the trans- RT F (Rb +Rc )+RT
former. The leakage flux and the additional reluctance due to the (6)
shorted turn/turns depends on the number of shorted-turns, lo- The proposed fault indicator under inter-turn fault in
cation of the shorted-turn/turns in a winding and the distance of any of the transformer’s primary windings of limb-a with
the shorted-turn/turns w.r.t the core limbs for the multi-winding (Ra = Rc = ) is given by (7).
transformers [18], [19], [25]–[27].
The fundamental phasor of the core-flux generated by the ϕF I (faulty) = ϕaT f −pr i + ϕcT f −pr i
limb-a winding MMF under inter-turn fault in the transformer’  
primary winding is given by, N R(Iea −Ieb)+(Iec −Ieb)+RT F (Iec −Ieb)− NNn a  Isc 
=
RT F (Rb +)+RT
Na Iea − Nna  Isc  
ϕaf −pr i =   N  (Iec + Iea − 2Ieb ) + RR T F
(Iec − Ieb ) − NNn a  Isc
(Ra + RT F ) + RRb b+RRc
=
c
RT F (Rb + ) + RT
(Na Iea − Nna  Isc ) (Rb + Rc ) (7)
=
RT F (Rb + Rc ) + RT From (5) and (6), it is evident that the value of the core-fluxes
in limb-a and -c change under inter-turn fault when compared to
The fundamental phasors of the fluxes in the limb-a due to the healthy condition as given by (1) and (2) respectively. This
limb-b and limb-c MMFs under inter-turn fault in the trans- is due to the additional core-reluctance (RT F ) and the MMF

former’ primary winding are given by, (Nna Isc ) at limb-a where the inter-turn fault occurs. The change
in the value of the core-fluxes at limb-a and limb-c causes the
 
Rc value of the proposed fault indicator to change under inter-turn
ϕbaf −pr i = Nb Ieb , and
RT F (Rb + Rc ) + RT fault as given by (7) when compared to the healthy condition as
  given by (4).
Rb
ϕcaf −pr i = Nc Iec C. Core-Flux Under Inter-Turn Fault in the Transformer’s
RT F (Rb + Rc ) + RT
Secondary Winding of Limb-a
The total core-flux in the limb-a under inter-turn fault Figs. 3(a) and 3(b) represent the electrical and magnetic
condition in the transformer’s primary winding with (Na = equivalent circuits of the transformer under inter-turn fault in

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
1234 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO. 4, AUGUST 2019

the transformer’s secondary winding of limb-a. The inter-turn for low-severity faults the change in the value of the leakage
fault in the secondary winding is reflected as the load on the coefficients are obscured by the change in the leakage flux from
primary windings of the transformer. Therefore, the load cur- healthy to turn fault conditions. Therefore, influence of the leak-
rents in the primary windings of the transformer increase. The age coefficients under low-severity turn faults can be neglected.
increase in the primary winding currents generates equal MMFs The values of the proposed fault indicators under healthy and
which in turn opposes the MMF generated by the faulty turns inter-turn fault conditions in the transformer’s windings of limb-
in the transformer’s secondary winding. Therefore, the negative a are given by (11)–(13).
MMF, −Nna  Isc is neglected as shown in Fig. 3(b) when com-
pared to inter-turn fault in the transformer’s primary winding  
N
as shown in Fig. 2(b). Similar to inter-turn fault in the primary ϕF I l(healthy) = kh (Iec + Iea − 2Ieb ) (11)
RT
winding, the fault in the secondary winding also increases the
core-reluctance as shown in Fig. 3(b). Therefore, the proposed ϕF I l(faulty−pr i) =
fault indicator value for inter-turn fault in the transformer’s sec- ⎛  ⎞
Nna 
ondary winding is same as that of (7) with the negative MMF N  (Iec +Iea − 2Ieb )+ RRT F
(Iec − Ieb )− N Isc
kf ⎝ ⎠
removed. RT F (Rb +)+ RT
The total core-flux in the limb-a under inter-turn fault in the
transformer’s secondary winding with (Na = Nb = Nc = N ) (12)
is given by (8). ϕF I l(faulty−sec) =
ϕaT f −sec = ϕaf −sec − ϕbaf −sec − ϕcaf −sec   
N  (Iec + Iea − 2Ieb ) + RR
T F
(Iec − Ieb )
kf (13)
N (Rb (Iea − Iec ) + Rc (Iea − Ieb )) RT F (Rb + ) + RT
= (8)
RT F (Rb + Rc ) + RT
Similarly, the total core-flux in the limb-c under inter-turn The outer limbs-a and -c are symmetrical about the
fault in the transformer’s secondary winding is given by (9). limb-b. Therefore, the characteristics of the proposed fault in-
dicator for inter-turn fault in the outer limbs are similar. Hence,
ϕcT f −sec = ϕcf −sec − ϕacf −sec − ϕbcf −sec the analytical model for inter-turn fault at the outer limb-c is not
N (Ra (Iec − Ieb ) + Rb (Iec − Iea ) + RT F (Iec − Ieb )) discussed. The core-fluxes and the fault indicators for inter-turn
= fault in the primary and secondary windings of the center limb-b
RT F (Rb + Rc ) + RT
(9) are given by (14) to (19).

The proposed fault indicator under inter-turn fault in the trans-

ϕaT f −pr i =
former’s secondary winding of limb-a with (Ra = Rc = ) is
 
given by (10). N
N Rb (Iea − Iec)+Rc (Iea −Ieb)−RT F (Iec −Iea)+ Nn b  Isc R
ϕF I (faulty) = ϕaT f −sec + ϕcT f −sec RT F (Ra +Rc ) + RT
  (14)
N  (Iec + Iea − 2Ieb ) + RR
T F
(Iec − Ieb )
=
RT F (Rb + ) + RT ϕcT f −pr i =
(10)  
N
N Ra (Iec −Ieb)+Rb (Iec −Iea)+RT F (Iec −Iea)+ Nn b  Isc R
From (10), the proposed fault indicator value changes when
there is an inter-turn fault in the transformer’s secondary wind- RT F (Ra + Rc )+RT
ing when compared to the fault indicator under healthy operating (15)
condition as given by (4). ϕaT f −sec =

D. Core-Leakage-Fluxes Under Inter-Turn Fault Conditions N (Rb (Iea − Iec ) + Rc (Iea − Ieb ) + RT F (Iea − Iec ))
of the Transformer RT F (Ra + Rc ) + RT
(16)
The proposed fault indicator is based on change in the value
of the core-fluxes from healthy to turn fault conditions. The ϕcT f −sec =
measurement of the core-fluxes is impractical as the flux flows
N (Ra (Iec − Ieb ) + Rb (Iec − Iea ) + RT F (Iec − Iea ))
inside the core. However, the leakage fluxes at the surface of the
core can be easily measured by using the Hall-effect based mag- RT F (Ra + Rc ) + RT
netic flux sensors. Hence, the proposed fault indicator can be (17)
represented by a leakage coefficient corresponding to the loca-
tion where the magnetic flux sensors are placed. The value of the Substituting, (Ra = Rc = ) in (14) to (17), values of the
leakage coefficient depends on the fault severities. Nonetheless, proposed fault indicator for inter-turn fault in the primary and

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1235

TABLE I
EXPERIMENTAL VALUES USED FOR THE THEORETICAL
ANALYSIS OF THE PROPOSED FAULT INDICATOR

Fig. 4. Phase-shift between the measured leakage-fluxes at various locations

and theoretical analysis of the proposed fault indicators for different fault
severities of an experimental prototype transformer.

at the surface of the core w.r.t the flux inside the core under
inter-turn fault conditions. Nonetheless, both the position of the
flux sensors and the angle of the leakage flux at the surface of
the core have similar influence on the proposed fault indicator.
Therefore, in the theoretical analysis only the influence of the
secondary winding of the limb-b are given by (18) to (19).
sensor positions is considered. Fig. 4(b) presents the polar plot
ϕF I l(faulty−pr i) = kf (ϕaT f −pr i + ϕcT f −pr i ) of the proposed fault indicators for various fault severities at the
   transformer’s primary and secondary windings for two differ-
N 2Nnb  ent locations of the flux sensors under no load condition. From
= kf (Iec + Iea − 2Ieb ) + Isc
2RT F  + RT N Fig. 4(b), the following conclusion can be inferred.
(18) 1) Under healthy operating condition of the transformer, the
phase angles of the proposed fault indicators w.r.t to zero
ϕF I l(faulty−sec) = kf (ϕaT f −sec + ϕcT f −sec )
  crossing of the line-voltage Vab is zero.
N 2) During inter-turn faults in the transformer’s primary and
= kf (Iec + Iea − 2Ieb )
2RT F  + RT secondary windings, the phase angles of the proposed fault
(19) indicators changes w.r.t 0◦ .
3) The change in angle values are better indicators w.r.t to the
The characteristics of the proposed fault indicators for inter- change in magnitude values for various fault severities.
turn fault at the outer and centre limbs of the transformer’s 4) The change in angle of the proposed fault indicators w.r.t
primary and secondary windings must be analyzed to identify 0◦ can be either positive or negative.
the fault detection criteria. Therefore, various parameters used From the theoretical analysis of the proposed fault indicators,
in the proposed fault indicators (11)–(13), (18) and (19) are change in the angle value is a good indicator of inter-turn faults
measured from an experimental prototype transformer for two in the transformer’s primary and secondary windings. Although
different inter-turn fault severities at the primary and secondary the change in angle can be either positive or negative, for a fixed
winding which are presented in Table I. The detailed informa- position of the flux sensors the fault detection criteria is based
tion of the experimental setup is presented in Section III. The on the relative change rather than the actual value.
measured values as given in Table I are used to estimate the
values and analyze the trend of the proposed fault indicators for
various inter-turn fault severities. III. EXPERIMENTAL SETUP
The Hall-effect based flux sensors are placed at the outer limbs Fig. 5 presents the experimental test-rig used for validation
of the experimental prototype transformer at three different loca- of the proposed technique. The specifications of the equipment
tions ((ϕaT f l−1 , ϕcT f l−1 ), (ϕaT f l−2 , ϕcT f l−2 ) and (ϕaT f l−3 , used for experimental validations are given in Table II. The
ϕcT f l−3 )) as shown in Fig. 4(a) to identify the influence of three-winding transformer used for experimental validation has
the flux sensors positions on the measured core-leakage-fluxes. a delta winding at the primary- and secondary-sides with 259-
From Fig. 4(a), the phase-shifts between the measured fluxes turns and a star-winding at the secondary-side with 150-turns
are dependent on the location of the flux sensors. Therefore, the as given in Table II. The external taps are taken from the pri-
influence of the phase-shift due to flux sensor location must be mary and secondary winding at 1%, 2% and 3% of the total
included in the theoretical analysis of the proposed fault indica- number of turns, which corresponds to 3-, 5- and 8-turns for the
tors. Besides the position of the flux sensors, the proposed fault delta winding and 2-, 3- and 5-turns for the star winding. The
indicators are also influenced by the angle of the leakage-fluxes transformer is loaded using the 5 kW three-phase resistive load

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
1236 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO. 4, AUGUST 2019

Fig. 6. Time-domain signals of core-leakage fluxes at the transformer outer

limbs under no-load and full-load conditions.

Fig. 5. Experimental test-rig: 1–10 kVA, three winding transformer, 2–5 kW

resistive load banks, 3-PC, 4-oscilloscope, 5-dSPACE 1103, 6-voltage sensor,
7-4th order Butterworth filter with amplifiers, 8-external fault resistor, 9-top
view of transformer, 10-developed prototype HMD, and 11-current probes. IV. EXPERIMENTAL VALIDATION OF THE PROPOSED
FAULT INDICATORS
TABLE II A programmable power source is used to maintain the line-
SPECIFICATIONS OF THE EQUIPMENT USED FOR EXPERIMENTAL VALIDATION voltages of the experimental transformer at Vab = 413 V, Vbc =
414 V and Vca = 411 V, which corresponds to a Line Volt-
age Unbalance Ratio (LVUR) of 0.7%. In addition to the
programmable AC power source, an auto-transformer is used
to gradually increase the supply voltage of the experimental
transformer which reduces the inrush current. Under no-load
operating condition, the inter-turn faults are introduced at the
transformer’s primary and secondary windings with appropri-
ate external fault resistors as given in Table II. For each fault
severity, the core-leakage fluxes at the outer limbs, ϕaT l and
ϕcT l and the line-voltage, Vab are acquired using dSPACE-
banks on each secondary winding as shown in block 2 of Fig. 5. 1103 controller board. The number of samples corresponding
The bolted turn-faults are introduced by connecting the external to one fundamental cycle ( 20k 50 ) of the core-leakage fluxes,
resistors across the taps as shown in block 8 of Fig. 5. Different ϕaT l and ϕcT l are extracted w.r.t the zero-crossing time in-
values of external resistors are connected across the taps as given stant of the line-voltage, Vab . The proposed fault indicator,
in Table II such that the short-circuit currents are within 3 times ϕF I l = (ϕaT l + ϕcT l ) is calculated, FFT computation is car-
the full-load terminal currents of the experimental prototype ried out and then the fundamental component’s magnitude and
transformer. A prototype Health Monitoring Device (HMD) as phase angle values are extracted which are presented in Fig. 7.
shown in block 10 of Fig. 5 is fabricated to demonstrate the Fig. 7 present the polar plot of the proposed fault indicators
real-time implementation of the proposed methodology which for various inter-turn fault severities in the transformer’s primary
is discussed in Section VI. and secondary windings under no-load operating condition with
For various values of loading and inter-turn fault severities, 0.7% LVUR. The zero degrees in Fig. 7 correspond to the zero-
the terminal currents, voltages and core-leakage fluxes at the crossing time instant of the line-voltage, Vab . From Fig. 7, the
transformer’s outer limbs are captured using dSPACE 1103 con- following conclusions can be inferred.
troller board at a sampling frequency of 20kHz as shown in block 1) Under healthy operating condition of the transformer, the
5 of Fig. 5. The terminal currents and voltages are sensed using phase angles of the proposed fault indicators are constant
clamp on current probes and differential probes as shown in at 60°.
block 11 and 6 of Fig. 5 respectively. The leakage fluxes are 2) During inter-turn faults in the transformer’s primary wind-
captured using Hall-effect based flux sensors whose positions ing, the phase angles of the proposed fault indicators
are represented by dotted circles in block 9 of Fig. 5. Fig. 6 changes widely w.r.t 60°.
presents the time-domain signals of the core-leakage fluxes at 3) During inter-turn faults in the transformer’s secondary
the transformer’s outer limbs under no-load and full-load con- winding, the phase angles of the proposed fault indicators
ditions. The leakage fluxes are measured at the upper surface changes mildly w.r.t 60°.
of the transformer-yoke near the edges of outer limbs as shown 4) The change in angle values of the proposed fault indicators
in block 9 of Fig. 5, which correspond to the core-flux. There- are better indicators than the change in magnitude values
fore, the magnitudes of the measured core-leakage fluxes should for various fault severities as shown by dotted arrows.
show insignificant variations for no-load and full-load operat- 5) The change in angle of the proposed fault indicators w.r.t
ing conditions of the transformer which is evident from Fig. 6. 60° can be either positive or negative.
Hence, the proposed technique is immune to loading conditions 6) The characteristics of the proposed fault indicators are
as it is based on the core-flux. same for both experimental results as shown in Fig. 7 and

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1237

Fig. 7. Polar plot of the proposed fault indicators for various inter-turn fault severities in the transformer’s primary and secondary windings under no-load
operating condition with 0.7% LVUR.

the theoretical analysis presented in Section II-D, thereby The time-domain signals of terminal voltages and currents
validate the proposed fault indicators. for one fundamental cycle are extracted w.r.t the zero crossing
It should be noted that, in real world applications, the time instant of the line voltage, Vab which is then converted to
supply voltage unbalance is inevitable [28] and [29]. The frequency-domain using FFT. The magnitude and phase angle
IEEE Std 1159 and IEC recommend that the maximum volt- of the fundamental phasor is extracted from the FFT spectrum of
age unbalance of electrical supply systems be limited to 2% terminal voltages and currents, which is then used to compute the
[28] and [29]. Therefore, the proposed fault indicators are DNSC as given by (20). Fig. 8 and Fig. 9 present the comparisons
also evaluated for 2% LVUR which is discussed in the next of the proposed and classical DNSC based techniques for inter-
section. Besides, performance comparison of the proposed turn fault detection in the transformer’s primary and secondary
technique and the classical Differential Negative Sequence Cur- windings respectively under various inter-turn fault severities,
rent (DNSC) based technique is also presented in the next loading conditions with 2% LVUR. For each inter-turn fault
section. severity as given in Table II, the transformer is loaded to 0%,
25%, 50%, 75% and 100% (10kVA) of the full-load capacity.
V. COMPARISON OF THE PROPOSED FAULT INDICATOR WITH Therefore, experimental results of the proposed and classical
CLASSICAL DNSC BASED TECHNIQUE DNSC based technique shown in Fig. 8 and Fig. 9 are captured
under various fault severities and loading conditions with a fixed
Traditionally, terminal currents based differential protection LVUR condition.
technique are used to protect the transformer from internal and From Fig. 7, Fig. 8 and Fig. 9, the following conclusions can
external faults [21], [22] and [24]. However, the classical dif- be inferred.
ferential protection systems are insensitive to detect the low- 1) The proposed technique detects the inter-turn faults for
severity turn faults which eventually lead to ground faults and various fault severities in the transformer’s primary wind-
enormous repair cost. The sensitivity of the traditional differ- ing of the outer limb-a under different loading conditions
ential protection systems is increased by Negative Sequence as shown in Fig. 8(a). The various loading conditions for
Current (NSC) based differential protection which is capable each fault severity are represented as a cluster in Fig. 8(a).
of detecting as low as 4-turns (1% of total turns) [21], [22]. The proposed technique is on par with the classical DNSC
Therefore, in this section, the performance of the proposed and based technique as shown in Fig. 8(b).
DNSC based techniques are compared to identify the pros and 2) For inter-turn fault in the transformer’s primary wind-
cons of the proposed technique w.r.t to the widely used tradi- ing of the centre limb-b, the classical DNSC tech-
tional technique. The DNSC are represented by (20) [21], [22] nique shows change in the value of both magnitude
and [24]. and angle as shown in Fig. 8(d). However, the pro-
⎡ ⎡  ⎤
⎡ ⎤ Vs 1 a posed technique shows substantial change in the angle
Is1a
⎢ Ipa ⎢ Vp a ⎥ value in comparison with the magnitude as shown in
 1 a 2 a  ⎢⎢ ⎥ ⎢  V s 1 b  ⎥
ID N S C = 3 ⎢⎣ Ipb ⎦ − ⎢ I ⎥ Fig. 8(c). Nonetheless, both the techniques detect inter-
3 3 ⎢ ⎢ Vp b s1b ⎥
⎣ ⎣  ⎦ turn fault at various fault severities under various loading
Ipc Vs 1 c
Vp c I s1c conditions.
⎡  ⎤⎤ 3) Both the proposed and DNSC techniques are on par in de-
Vs 2 a
Is2a tecting the inter-turn faults in the transformer’s secondary
⎢ Vp a  ⎥⎥
⎢ V ⎥⎥ winding of the outer limb-a under different loading con-
−⎢ s2b
⎢ Vp b I ⎥⎥
s2b ⎥⎥ (20) ditions as shown in Fig. 9(a) and 9(b) respectively.
⎣  ⎦⎦ 4) The sensitivity of the proposed technique for detecting the
Vs 2 c
Vp c Is2c inter-turn fault in the transformer’s secondary winding of

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
1238 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO. 4, AUGUST 2019

Fig. 8. Comparion of the proposed and DNSC based techniques for inter-turn fault detection in the transformer’s primary windings under various inter-turn fault
severities and loading conditions with 2% LVUR.

Fig. 9. Comparion of the proposed and DNSC based techniques for inter-turn fault detection in the transformer’s secondary windings under various inter-turn
fault severities and loading conditions with 2% LVUR.

of the baseline value of the NSC [21], [22] and [24]. Therefore,
the trip signal is generated, only if the magnitude of the DNSC
computed under online operating conditions is greater than 1%
of the base NSC value. Although the proposed technique is
less sensitive than the DNSC based technique in detecting the
inter-turn faults in the transformer’s centre limb windings, the
relative change from healthy to 3-turns fault is more than 10%
as shown in Fig. 10. This means, the relative deviation of the
Fig. 10. Zoomed view of Fig. 9(c). proposed fault indicator is 10 times more than the minimum
threshold of the classical DNSC based technique. Therefore,
the proposed fault indicator is equally good in identifying the
inter-turn faults in the transformer’s winding of the centre and
the centre limb-b is less when compared to the DNSC
outer limbs. The inter-turn fault detection capabilities of the
technique as shown in Fig. 9(c) and 9(d) respectively.
proposed fault indicator under real-time operating conditions
Nonetheless, the deviation in angle value of the proposed
and benefits and limitations of the proposed methodology are
fault indicator is more than 10% for 5- and 8-turns faults as
presented in Section VI and Section VII respectively.
shown in Fig. 10. In both proposed and DNSC techniques,
2-turns fault overlaps with the healthy condition as shown
VI. REAL-TIME IMPLEMENTATION OF THE PROPOSED
in Fig. 9(c) and 9(d) respectively.
5) The supply voltage unbalance conditions do not affect the TECHNIQUE USING THE DEVELOPED HMD
proposed fault indicator as shown in Fig. 7 with 0.7% The real-time evaluation of the proposed fault detection tech-
LVUR and Figs. 8(a), 8(c), 9(a) and 9(c) with 2% LVUR. nique is carried out using the developed prototype HMD as
Holistically, the sensitivities of both the proposed and the shown in block 10 of Fig. 5 and Fig. 11(a). The developed
DNSC based techniques are on par for detection of inter-turn HMD has a power supply unit, a Digital Signal Processor (DSP,
fault in the transformer’s windings of the outer limb. The thresh- TMS320F28335) and signal pre-processing unit as shown by
old limit for the DNSC based techniques varies from 1 to 20% block 1 to block 3 of Fig. 11(a) respectively. The core-leakage

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1239

Fig. 11. Developed prototype HMD and the framework for real-time evaula-
tion of the proposed fault indicator. 1-power supply unit, 2-DSP, and 3-signal
pre-processing unit.

flux signals at the transformer’s outer limbs are measured by the

Hall-effect based flux sensors (SS496A1) which are encircled
in block 9 of Fig. 5, which are then level shifted by the signal
pre-processing unit to make them compatible to the DSP. The
pre-processed signals are then captured by the DSP at a sam-
pling frequency of 10kHz. The FFT computation and the angle
value extraction of the proposed fault indicator, ϕF I l is carried
out in the DSP. Fig. 12. Measured primary voltage, inter-turn fault current, lekage fluxes,
and trip signal under 1% inter-turn fault at the primary winding: 1-Trip sig-
Fig. 11(b) shows the framework implemented in the DSP of nals under intermittent inter-turn faults, 2-Trip signal during fault occurrence,
the developed prototype HMD. The HMD acquires the leakage and 3-Trip signal during fault clearence.
fluxes and a primary winding line-voltage, Vab for two cycles
(40ms). The zero-crossing instant of the line-voltage, Vab is de-
tected in the DSP which is then used as a reference angle value as shown in Fig. 11(b). The additional one cycle is due to the
for the leakage flux signals. The captured leakage fluxes from computational delay by the DSP. The real-time implementation
the zero crossing instant of the line-voltage, Vab upto a cycle by the developed HMD revealed that the proposed technique
(20 ms) are added together to obtain the proposed fault indica- detects inter-turn fault as small as 1% of the total number of
tor, ϕF I l under online operating conditions of the transformer. winding turns within 3 fundamental cycles and provides robust
The FFT computation is then carried out on the proposed fault performance even under intermittent inter-turn fault conditions.
indicator and the angle value is then extracted. If the angle val-
ues obtained under online operating conditions, ϕF I l(faulty) are
VII. BENEFITS AND LIMITATIONS
greater or smaller than the 10% of the value under healthy con-
dition, ϕF I l(healthy) then the trip signal is generated as shown in From Fig. 9 (c) and 9(d), it can be observed that the sensitivity
Fig. 12. of the DNSC based technique is better than the proposed tech-
Fig. 12 presents the measured primary voltage, inter-turn nique to detect inter-turn fault in the transformer’s secondary
fault current, leakage fluxes and trip signal under inter-turn winding of limb-b. This is due to the inevitable requirement
fault across 1% of the total turns in the primary winding of of the Current Transformers (CTs) and Voltage Transformers
the outer limbs. From block 1 of Fig. 12, the inter-turn fault (VTs) to measure the terminal currents and voltages respec-
causes the current in the shorted turn to increase drastically tively. For three-phase, three-winding transformers, the DNSC
which is followed by the generation of the trip signal by the based techniques requires 9-VTs and 9-CTs as given by (20).
developed prototype HMD. The inter-turn fault is introduced However, the VTs can be eliminated in the transformers with-
and cleared at two different time instants with different time du- out OLTC, as the transformation ratio of the terminal currents
rations as shown in block 1 of Fig. 12. The proposed technique remains the same. Nonetheless, VTs eliminates the additional
is able to generate the trip signal reliably for intermittent faults phase angle shift compensation for the secondary and primary
which projects the robust characteristics of the proposed tech- currents [21], [22], [24] and increases the sensitivity of DNSC
nique under intermittent inter-turn faults. The block 2 and block based techniques. The total number of CTs and VTs increases
3 of Fig. 12 present the state of the trip signal generated by the the total cost of the protection system. Moreover, the cost of
developed HMD under fault occurrence and clearance instants the CTs and VTs increases with the increase in the power
respectively. From block 2 and block 3 of Fig. 12, it is evident level. However, the proposed technique is based on the core-
that the time taken by the trip signal to change its state after the leakage flux which is proportional to the core flux. Typically,
occurrence and clearance of the fault is approximately 60 ms (3 the value of core flux of a power transformer varies from 0.2T
cycles). Out of this 3 cycles delay, the digital buffer used to save to 1.8T depending on the iron losses in the magnetic circuit
primary voltage and leakage flux signals incurs 2 cycles (40 ms) [30]. Considering the maximum working core-flux density the

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
1240 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 34, NO. 4, AUGUST 2019

voltage unbalance conditions. The core-leakage flux based pro-

posed methodology is as good as DNSC based technique with
additional advantage of reduced sensor cost, as it requires just
two flux sensors to measure the core-flux for power transform-
ers. The proposed technique with the low-cost health monitor-
ing device, can be easily retrofitted to the dry-type transformers
where the DNSC based techniques are not economical or as an
add-on to the existing DNSC based technique to increase the
confidence level.
Fig. 13. Cost compariosn for sensors of DNSC and proposed techniques.
∗ Bulk prices are taken from Element14.

REFERENCES
sensors cost of the proposed technique is fixed and is insignif-
icant when compared to the costs of the CTs and VTs. Fig. 13 [1] S. Tenbohlen, F. Vahidi, and J. Jagers, “A worldwide transformer relia-
bility survey,” in Proc. VDE High Voltage Technol. ETG-Symp., Berlin,
presents the cost comparison of sensors required by DNSC and Deutschland, 2016, pp. 1–6.
proposed methodology for a 10kVA, three phase three wind- [2] D. D. Shipp, T. J. Dionise, V. Lorch, and B. G. MacFarlane, “Transformer
ing transformer. From Fig. 13, the sensor cost of the proposed failure due to circuit-breaker-induced switching transients,” IEEE Trans.
Ind. Appl., vol. 47, no. 2, pp. 707–718, Mar./Apr. 2011.
methodology is just 7% of the total cost of sensors required [3] B. S. Rajpurohit et al., “A case study of moisture and dust induced failure
by DNSC based techniques and it is nominal for higher power of dry type transformer in power supply distribution,” in Water Ener. Int.,
levels. Altogether, the proposed technique is inexpensive and is vol. 60r, no. 2, pp. 43–47, Jul. 2017.
[4] Y.-L. Tan, “Damage of a distribution transformer due to through-fault cur-
on par with the classical DNSC based technique in detecting the rents: An electrical forensics viewpoint,” IEEE Trans. Ind. Appl., vol. 38,
low-severity inter-turn faults. Furthermore, the proposed online no. 1, pp. 29–34, Jan./Feb. 2002.
fault indicator can be implemented as a standalone fault detec- [5] D. Paul, “Failure analysis of dry-type power transformer,” IEEE Trans.
Ind. Appl., vol. 37, no. 3, pp. 689–695, May/Jun. 2001.
tion system or as an add-on to the existing DNSC based system [6] J. Wimmer, M. R. R. Tanner, T. Nunn, and J. Kern, “Dry-type trans-
to increase the confidence level. formers: Specification, installation, and operational impacts in a marine
The proposed technique is based on the core leakage fluxes environment,” IEEE Ind. Appl. Mag., vol. 19, no. 5, pp. 68–75, Sep./Oct.
2013.
which is proportional to the magnetizing currents. Therefore, [7] T. Hoevenaars, I. C. Evans, and A. Lawson, “New marine harmonic
any anomalies which cause disturbance in the excitation cur- standards,” IEEE Ind. Appl. Mag., vol. 16, no. 1, pp. 16–25, Jan./Feb.
rents will adversely affect the proposed methodology. Some of 2010.
[8] M. Wang, A. J. Vandermaar, and K. D. Srivastava, “Review of condition
the phenomena which leads to disturbances in the excitation cur- assessment of power transformers in service,” IEEE Elect. Insul. Mag.,
rents are Geomagnetically Induced Current (GIC), inrush cur- vol. 18, no. 6, pp. 12–25, Nov./Dec. 2002.
rents, transients such as occurrence of external fault, clearance [9] X. Zhang and E. Gockenbach, “Asset-management of transformers based
on condition monitoring and standard diagnosis [feature article],” IEEE
of fault and bi-directional power flow due to distributed energy Elect. Insul. Mag., vol. 24, no. 4, pp. 26–40, Jul./Aug. 2008.
sources [24], [31], [32]. Both GIC and inrush currents generate [10] W. Herrera Portilla, G. Aponte Mayor, J. Pleite Guerra, and C. Gonzalez-
significant amount of harmonics due positive/negative satura- Garcia, “Detection of transformer faults using frequency-response traces
in the low-frequency bandwidth,” IEEE Trans. Ind. Electron., vol. 61,
tion and part-cycle saturation respectively [24], [31]. Under no. 9, pp. 4971–4978, Sep. 2014.
such scenario, harmonics restraints are used in DNSC to avoid [11] Z. Zhao, C. Yao, C. Li, and S. Islam, “Detection of power transformer
the false trip signal generation [24]. In the proposed methodol- winding deformation using improved FRA based on binary morphology
and extreme point variation,” IEEE Trans. Ind. Electron., vol. 65, no. 4,
ogy, the same concept of harmonic restraint can also be applied pp. 3509–3519, Apr. 2018.
to avoid nuisance tripping due to core saturation. The undesired [12] M. Bagheri, M. S. Naderi, and T. Blackburn, “Advanced transformer
trip signal generation due to momentary transients such as oc- winding deformation diagnosis: Moving from off-line to on-line,”
IEEE Trans. Dielectr. Elect. Insul., vol. 19, no. 6, pp. 1860–1870,
currence of external fault, clearance of fault and bi-directional Dec. 2012.
power flow can be eliminated by coordination with other backup [13] P. A. Venikar, M. S. Ballal, B. S. Umre, and H. M. Suryawanshi, “A novel
and external protection systems such as overcurrent and direc- offline to online approach to detect transformer interturn fault,” IEEE
Trans. Power Del., vol. 31, no. 2, pp. 482–492, Apr. 2016.
tional overcurrent relays [24], [32]. This is achieved by checking [14] N. Asadi and H. M. Kelk, “Modeling, analysis, and detection of internal
the trip status of overcurrent and directional overcurrent relays, winding faults in power transformers,” IEEE Trans. Power Del., vol. 30,
only if the trip status is “False” the proposed technique is used to no. 6, pp. 2419–2426, Dec. 2015.
[15] M. Bagheri, M. S. Naderi, T. Blackburn, and B. T. Phung, “Frequency
identify the internal faults. In general, for reliable inter-turn fault response analysis vs. flux division measurement in detection of trans-
detection using the proposed methodology, coordination from former winding internal short circuit,” in Proc. IEEE Int. Conf. Power
other external protection systems are required and harmonics Syst. Technol., Auckland, 2012, pp. 1–5.
[16] M. S. Ballal, H. M. Suryawanshi, M. K. Mishra, and B. N. Chaudhari,
restraint has to be carried out. “Interturn faults detection of transformers by diagnosis of neutral current,”
IEEE Trans. Power Del., vol. 31, no. 3, pp. 1096–1105, Jun. 2016.
VIII. CONCLUSION [17] A. S. Masoum, N. Hashemnia, A. Abu-Siada, M. A. S. Masoum, and
S. M. Islam, “Online transformer internal fault detection based on instan-
In this paper, an inexpensive low-severity inter-turn fault taneous voltage and current measurements considering impact of harmon-
ics,” IEEE Trans. Power Del., vol. 32, no. 2, pp. 587–598, Apr. 2017.
detection technique for dry-type transformers used in various [18] P. A. Venikar, M. S. Ballal, B. S. Umre, and H. M. Suryawanshi, “Search
industries, transit and marine applications has been proposed coil based online diagnostics of transformer internal faults,” IEEE Trans.
and evaluated under various loading, fault severities and supply Power Del., vol. 32, no. 6, pp. 2520–2529, Dec. 2017.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.
ATHIKESSAVAN et al.: INTER-TURN FAULT DETECTION OF DRY-TYPE TRANSFORMERS USING CORE-LEAKAGE FLUXES 1241

[19] F. Haghjoo, M. Mostafaei, and H. Mohammadi, “A new leakage flux based Elango Jeyasankar (M’17) received the B.E. de-
technique for turn to turn fault protection and faulty region identification gree in electrical and electronics engineering from
in transformers,” IEEE Trans. Power Del., vol. 33, no. 2, pp. 671–679, the Thiagarajar College of Engineering, Madurai, In-
Apr. 2018. dia, in 2013, and the M.S. degree in electrical and
[20] M. Bagheri, A. Zollanvari, and S. Nezhivenko, “Transformer fault con- computer engineering from the National University
dition prognosis using vibration signals over cloud environment,” IEEE of Singapore, Singapore, in 2015. He is currently a
Access, vol. 6, pp. 9862–9874, 2018. Research Associate with the National University of
[21] Z. Gajić, I. Brnčić, B. Hillström, F. Mekić, and I. Ivanković, “Sensitive Singapore. His research interests include motor drives
turn-to-turn fault protection for power transformers,” in Proc. 32nd Annu. for hybrid electric vehicles, and condition monitor-
West. Protective Relay Conf., Spokane, WA, Oct. 2005. ing of permanent magnet synchronous motors, syn-
[22] D. Zacharias and R. Gokaraju, “Prototype of a negative-sequence turn-to- chronous generators, induction motors, and power
turn fault detection scheme for transformers,” IEEE Trans. Power Del., transformers.
vol. 31, no. 1, pp. 122–129, Feb. 2016.
[23] L. M. R. Oliveira and A. J. M. Cardoso, “Comparing power transformer
turn-to-turn faults protection methods: Negative sequence component
versus space-vector algorithms,” IEEE Trans. Ind. Appl, vol. 53, no. 3,
pp. 2817–2825, May/Jun. 2017.
[24] IEEE Guide for Protecting Power Transformers, IEEE Std C37.91-2008
(Revision of IEEE Std C37.91-2000), May 30, 2008.
[25] S. Chandar A. and S. K. Panda, “Detection and isolation of interturn faults
in inductors of LCL filter for marine electric propulsion system,” IEEE Sai Srinivas Manohar (S’16–M’17) received the
Trans. Transp. Electrific., vol. 4, no. 1, pp. 232–248, Mar. 2018. B.E. degree in electrical and electronics engineering
[26] V. Behjat and A. Vahedi, “Numerical modelling of transformers interturn from the Thiagarajar College of Engineering, Madu-
faults and characterising the faulty transformer behaviour under various rai, India, in 2015, and the M.S. degree in electri-
faults and operating conditions,” IET Elect. Power Appl., vol. 5, no. 5, cal and computer engineering specializing in power
pp. 415–431, May 2011. and energy systems from the National University of
[27] L. M. R. Oliveira and A. J. M. Cardoso, “Leakage inductances calculation Singapore, Singapore, in 2016. He is currently a Re-
for power transformers interturn fault studies,” IEEE Trans. Power Del., search Engineer with the Electrical Machines and
vol. 30, no. 3, pp. 1213–1220, Jun. 2015. Drives Laboratory, National University of Singapore.
[28] IEEE Recommended Practice for Monitoring Electric Power Quality, His research interests include condition monitoring,
IEEE Std 1159-2009 (Revision of IEEE Std 1159-1995), Jun. 26, 2009. fault detection techniques for transformers, etc.
[29] A. von Jouanne and B. Banerjee, “Assessment of voltage unbalance,”
IEEE Trans. Power Del., vol. 16, no. 4, pp. 782–790, Oct. 2001.
[30] M. Elleuch and M. Poloujadoff, “Analytical model of iron losses in power
transformers,” IEEE Trans. Magn., vol. 39, no. 2, pp. 973–980, Mar. 2003.
[31] IEEE Guide for Establishing Power Transformer Capability While Under
Geomagnetic Disturbances, IEEE Std C57.163-2015, pp. 1–50, Oct. 26
2015.
[32] H. Gao, J. Li and B. Xu, “Principle and implementation of current differ- Sanjib Kumar Panda (S’86–M’91–SM’01) re-
ential protection in distribution networks with high penetration of DGs,” ceived the M.Tech. degree from the Indian Institute
IEEE Trans. Power Del., vol. 32, no. 1, pp. 565–574, Feb. 2017. of Technology, Banaras Hindu University, Varanasi,
India, in 1987, and the Ph.D. degree from the Univer-
sity of Cambridge, Cambridge, U.K., in 1991, both
Subash Chandar Athikessavan (S’15–M’18) re- in electrical engineering. Since 1992, he has been
ceived the M.Sc. and the Ph.D. degrees in electrical holding a Faculty position with the Department of
and computer engineering from the National Univer- Electrical and Computer Engineering, National Uni-
sity of Singapore (NUS), Singapore. He is currently versity of Singapore, and is currently an Associate
a Research Fellow with Sembcorp-NUS Corporate Professor and the Director of the Power and Energy
Laboratory, NUS. His research focus is on devel- Research Area. He has authored or coauthored more
opment of condition monitoring and fault-detection than 300 peer-reviewed research papers, coauthored one book, and contributed
and diagnosis techniques for electrical machines. His to several book chapters and six patents. He is currently an Associate Editor of
areas of expertise are signal processing techniques, several IEEE Transactions and the Editor of the Journal of Power Electronics,
model-, algorithm-development and validation, and South Korea. He has served in various capacities for the IEEE Power Electron-
prototypes development and demonstration for health ics and Drive Systems Conference and the IEEE International Conference on
monitoring of electrical machines. Sustainable Energy Technologies Conference series.

Authorized licensed use limited to: Cornell University Library. Downloaded on September 01,2020 at 16:56:19 UTC from IEEE Xplore. Restrictions apply.