Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303736377

Diagnosis of power transformer incipient faults using electronic nose

Conference Paper · May 2016

CITATIONS READS

0 530

2 authors:

Bekir Mumyakmaz Kerim Karabacak


Dumlupinar Üniversitesi Dumlupinar Üniversitesi
18 PUBLICATIONS   79 CITATIONS    6 PUBLICATIONS   44 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Artificial neural networks for controlling wind–PV hybrid power systems View project

All content following this page was uploaded by Bekir Mumyakmaz on 02 June 2016.

The user has requested enhancement of the downloaded file.


Diagnosis of power transformer incipient
faults using electronic nose

Bekir Mumyakmaz1, Kerim Karabacak2

Abstract
Power transformers are very expensive entities of electrical power systems. Since repairing and
maintaining health of a transformer is very difficult and time consuming, their failure will have big
negative impact on the system. Thus, any information regarding incipient faults in the transformer is very
important. In this study, an electronic nose that is implemented to detect incipient faults of power
transformers is presented. The electronic nose has four gas sensors namely hydrogen, carbon monoxide,
carbon dioxide and methane with an addition of humidity and temperature sensor. The results of the study
show that the proposed detection method can effectively diagnose partial discharge and overheated
cellulose type of incipient faults of power transformers.
Keywords: Electronic Nose, Power transformer, Incipient fault diagnosis

1. INTRODUCTION
Power transformers are very expensive components of electrical power systems. Proper operation of them is
crucial to maintain continuity of supply. Therefore, some mechanisms should be developed to predict
transformer failures before they may occur.

Certain gases are released in transformer insulating oil due to changing working conditions over time such as
overloading, overheating, partial discharge, low energy sparking and arching in the insulation system. The
type and amount of the gases generated in the transformer oil provide information about the status of the
transformer. Dissolved gas analysis (DGA) is a reliable method to determine incipient faults in a transformer.
IEC 60599 [1] and IEEE C57.104 [2] are the standard guidelines for DGA diagnosis.

Classical method for DGA is to take some oil sample from the transformer, take it to the laboratory and put it
in the gas chromatography equipment. A modern power transformer may have an online DGA and
monitoring system which is very expensive. Some of them may be capable of extracting full DGA gases
namely hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon monoxide
(CO), carbon dioxide (CO2), nitrogen (N2), and oxygen (O2) [3]. The some others may have only hydrogen
and carbon monoxide monitoring [4].

1
Corresponding author: Dumlupınar University, Department of Electrical and Electronics Engineering, 43100, Kütahya,
Turkey. bekir.mumyakmaz@dpu.edu.tr
2
Dumlupınar University, Kutahya Technical Sciences Vocational School, 43100, Kütahya, Turkey.
kerim.karabacak@dpu.edu.tr

1
ICENS International Conference on Engineering and Natural Science, 24-28 May 2016, Sarajevo, Bosnia
and Herzegovina
In this study, an electronic nose has been implemented to determine the gases that are released from
transformer oil. Various catalytic chemical gas sensors are used in the electronic nose system. The gas
samples, that simulate the gas mixtures that of malfunctioned transformers, have been applied to the
implemented system and the type of fault has been determined by the sensor responses via information
extraction. The chemical sensors that are used in the electronic nose can detect only gases that release in the
case of transformer insulating paper deformations and partial discharge events.

2. E-NOSE AND DIAGNOSIS OF INCIPIENT FAULTS


Basically, an electronic nose (E-Nose) is a device intended to detect odors and it comprises of gas sensor
arrays and pattern recognition systems. The sensors that are commonly used for electronic noses include
metal–oxide–semiconductor, conducting polymer, quartz crystal microbalance (QCM), and surface acoustic
wave (SAW) type of sensors. The data interpretation systems that are used for the analysis of results include
artificial neural network (ANN), fuzzy logic, and other pattern recognition methods.

The metal-oxide type of gas sensors are used in this study due to suitability of their measurement ranges for
incipient faults in transformers. Table 1 lists the sensors that are used in the study. The sensors listed in Table
1 can only be used for transformer insulating paper deformations and partial discharge events. The
determination of other type of incipient faults such as arcing is not possible since the sensors for the other
three gases mentioned above is not available.

Table 1. The sensors that are used in the E-Nose for diagnosis of incipient faults of transformers.

Sensor Gas Measurement Range


TGS2442 CO 30-1000 ppm
TGS3870 CH4+CO 500-12500 (CH4);50-1000 (CO) ppm
FCM6812 H2 0-14000 ppm
TGS821 H2 0- 5000 ppm
TGS2611-E00 CH4 500-10000 ppm
CDM4160-H00 CO2 400-45000 ppm
SHT75 Humidity and Temperature RH %20-80
KE-50 O2 0 - %100

An E-Nose has been designed and implemented using the sensors above to diagnose transformer incipient
faults. The E-Nose is shown in Fig. 1. The sensors are in an aluminum chamber in order to dissipate the heat
from the sensors. The purging gas (dry air) and sample gas representing the fault in the transformer are taken
from tedlar bags and passed through the chamber using a small pump inside the box. The sensor responses are
collected via serial port of a tablet PC as seen in Fig. 1b.

a) From the inside of the box b) From the outside of the box.
Fig. 1- The designed and implemented E-Nose.

2
ICENS International Conference on Engineering and Natural Science, 24-28 May 2016, Sarajevo, Bosnia
and Herzegovina
Each sensor must be characterized and calibrated before using them for the interpretations. Firstly, the output
voltages of the sensors are maximized in the sensing ranges by using adjustable resistances. Secondly, the
sensor responses are checked in terms of repeatability. Then, the sensor responses for changing
concentrations of the sample gases are recorded in order to calibrate the sensors. One sample of calibration
process is shown in Fig. 2.

Fig. 2- The TGS821 response to hydrogen samples in the calibration process.

Only 4 sensors namely TGS821, TGS4160, TGS2442 and TGS2641 are used to determine the fault types
since some of the sensors are redundant. After calibration process of the sensors, the sample gas mixtures that
represent incipient faults in power transformers are applied to the E-Nose. The gas mixtures are prepared
according to the amounts in certain types of faults determined in power transformers mentioned in [5].

3. RESULTS AND DISCUSSION


Figure 3 shows sensor responses for a sample gas mixture after calibration process. At the beginning of the
experiment, dry air is applied to the system in order to purge the sensors of the E-Nose for 10 minutes. The
prepared gas mixture has been applied to the E-Nose for 5 minutes and then again there has been a purging
time. The mixture consists of mainly hydrogen and methane gases with the addition of some carbon
monoxide. These gases are dissolved in the transformer oil when there is a partial discharge or thermal
cellulose type of faults in the transformer. The steady-state responses of the sensors are on the right side of
the Fig. 3.

Fig. 3- Sensor responses for a sample gas mixture after calibration process.

The sensors clearly show the amounts of the gases in the sample bag in ppm (particles per million). The
concentration levels of the gases in the mixture are in the ranges that are seen in partial discharge and
overheated cellulose type of incipient faults in a transformer. The E-Nose does not contain ethane, ethylene,
and acetylene sensor. Thus, it cannot determine thermal faults in oil or electrical arcing faults.

3
ICENS International Conference on Engineering and Natural Science, 24-28 May 2016, Sarajevo, Bosnia
and Herzegovina
The pattern recognition part of the E-Nose is explained in detailed in other studies which employ artificial
neural networks [6-8].

4. CONCLUSION
In this study, an electronic nose that is designed and implemented to detect incipient faults of power
transformers is presented. The calibrated electronic nose has four gas sensors namely hydrogen, carbon
monoxide, carbon dioxide and methane with an addition of humidity and temperature sensor. The results of
the study show that the proposed detection method can effectively diagnose partial discharge and overheated
cellulose type of incipient faults of power transformers.

ACKNOWLEDGMENT
This work was supported by Dumlupınar University (No: 2010-10).

REFERENCES
[1]. Mineral Oil-Impregnated Electrical Equipment in Service-Guide to the Interpretation of Dissolved and Free Gases
Analysis. IEC 60599 - 2007
[2]. IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed Transformers. IEEE Standard C57.104 - 2008
[3]. http://www.energy.siemens.com/us/en/services/power-transmission-distribution/transformer-lifecycle-
management/online-dga-monitoring.htm (Accessed April 2016)
[4]. http://www.mte.ch/categorie_26.html (Accessed April 2016)
[5]. Michel Duval, Alfonso de Pablo, (2001) Interpretation of Gas-In-Oil Analysis Using New IEC Publication 60599 and
IEC TC 10 Databases
[6]. Mumyakmaz B, Özmen A, Ebeoğlu MA, Taşaltın C. “Predicting gas concentrations of ternary gas mixtures for a
predefined 3D sample space.” Sensor Actuat B-Chem 2008; 128: 594-602.
[7]. Mumyakmaz B, Karabacak K, “An E-Nose based indoor air quality monitoring system: Prediction of combustible and
toxic gas concentrations.” Turk J Elec Eng & Comp Sci, 2015, 23(3), 729-740.
[8]. Özmen A, Mumyakmaz B, Ebeoglu M A, Tasaltın C, Gürol I, Öztürk Z Z, Dural D “Quantitative information
extraction from gas sensor data using principal component regression (PCR).” Turk J Elec Eng & Comp Sci, 2016,
24: 946-960.

View publication stats

You might also like