2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 1

Implementation of Computer based Software for

Oil Immersed Power Transformer Condition
Monitoring via Dissolved Gas In-oil Results
Pius Victor 1, Boonruang Marungsri 2*
1
School of Electrical Engineering, Suranaree University of Technology, Thailand
111 University Avenue, Nakhon Ratchasima 30000, Tel. 093-643-8346, Email: piusvictor2013@gmail.com
2*
Corresponding Author Tel. 089717-7065-, Email: bmshvee@sut.ac.th

transformers. Five Condition assessment tools for power

Abstract—The vital role of power transformers in the power transformers through DGA (Key Gas, Dornenburg, Rogers,
system has attracted the consideration of Condition based IEC and Duval) have been reviewed and implemented in
Maintenance for many years. Dissolved gas in-oil analysis (DGA) software bases. Paper also opens the way forward to the access
has gained popularity in condition monitoring and diagnosis
through its diagnosing methods like Key Gas, Dornenburg, and use of database raw data for diagnosis; evaluate method(s)
Rogers, IEC and Duval Triangle. This paper presents the used, data comparison, gases trending, saved work summary
implementation of computer based software with Database for for future evaluation.
transformer condition monitoring with the use of Dissolved Gas An easy graphical user interface was made by using
in-oil Analysis. It involves database for storage of DGA Python programming language. According to [7], it is defined
information in which they can be retrieved by the software for to be a general purpose, open source computer programming
fault diagnosis through the use of five diagnostic methods (Key
Gas method, Dornenburg ratio method, Roger’s ratio method, language, optimized for software quality, developer
IEC ratio method and Duval Triangle method). 50 samples from productivity, program portability and component integration.
other researchers were taken, stored and tested for proving the Also Python is said to support external components coded in
performance of the DGA software. The DGA software showed other languages; a scripting language as it makes it easy to
the accuracy of more than 90 percent in fault interpretation (with utilize and direct other software components.
the use of Duval Triangle). The implementation involved the use As in [8], MySQL was stated as an open source, multi-
of database made with MySQL 5.5 for storage of DGA
information, DGA software made with Python 2.7 and compiled threaded, relational database management system created by
with Eclipse in Ubuntu 14.04 (computer operating system). Michael “Monty” Widenius in 1995, and in 2000, MySQL
was released under a dual license model that permitted the
Index Terms— Condition monitoring, Database, DGA, Power public to use it for free under the GNU Public License (GPL);
transformers. which caused its popularity to be soaring. The success of
MySQL as a leading database was defined due not only to its
INTRODUCTION
price-after all, other cost-free and open source databases are
R EGARDING the importance of power transformer in
the power system and the dynamic nature of faults that
occur in it, condition based maintenance has been much
available-but also its reliability and performances. Most
features contributed to MySQL’s standing as a superb
database system include speed and the storage engine.
considered as the best maintenance practice. This is due to the
In this paper, DGA software features and basics of
fact that most of the faults are accumulated for a certain period
condition monitoring of power transformer are detailed in
of time. Degradation of insulation in oil filled transformer is a
Section 2; revision of Condition assessment tools for DGA,
normal process, it occurs even normal load conditions but
their implementation, usage of software for diagnosis, gas
abnormal conditions accelerates the rate of degradation which
comparison and trending, report and work summary are
affects electrical, chemical and physical properties of
explained in Section III. The results of diagnosis by different
insulating oil [6]. For the case of distribution networks with a
assessment method in comparison with other researchers are
wide range of voltages like in Thailand [2] power transformers
described in Section IV and conclusion is given in Section V.
are available in large number, the use of database can reduce
The diagnosis and evaluation of DGA works are according to
tedious works of inputting DGA information for
IEEE and IEC standards.
interpretation. Some of the DGA software developed are
described in [2], [3], [4], [5] but they do not have database i.e.
they request user input for every diagnostic. All these cannot
make reference with the previous DGA information. This
leads to unnecessary time consumption when user wants to
compare or to know the trend of internal condition in power
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 2

II. FEATURES OF DGA SOFTWARE

A. Graphical User Interface Layout
A simple GUI consists of three parts. (i) User Input Form-
for saving gas details (H2, CO, CO2, N2, CH4, C2H6, C2H4 and
C2H2) and non-gas details (Power-kVA, Voltage-kV, Altitude-
M, Humidity-%, Oil Temperature-C, Location-name,
Sampling date-date) for further evaluation and decision
making about power transformer. (ii) Toolbar with five
diagnosing methods (KGM, Dornenburg, Rogers, IEC,
Duval), gas comparison and gas trending. (iii) Menu bar with
options for customizing software and database details. After
saving gas and non-gas data to the database, user can easily
pick any data by data ID from the DGA software, and
diagnose by using any diagnosing method. All diagnoses give
gas details, execution, analysis and graphical representation
which are easy to learn and understand.
B. Database
A MySQL database was created and connected to the
software. A database has a table in it with eleven columns
named ID (primary key); hydrogen, carbon_monoxide, carbon
dioxide, nitrogen, methane, ethane, ethylene, acetylene,
power, voltage, altitude, oil temperature, humidity, Fig. 1. Algorithm of DGA software
sampling_date, machine_location, machine_number. User has
nothing to do with the database rather than saving and picking D. Condition Classification for DGA Methods
data which is done through the use of DGA software. Five methods were used to interpret DGA information and
analyze the condition of power transformer. This DGA
Software gives user an opportunity to select any data from
C. Development Tools and Algorithm of DGA Software
database and use any DGA method to diagnose it at a time as
DGA Software was developed in Ubuntu 14.04 (operating interpretation varies with the method.
system) and database connected with it. Other details are
shown in Table I below. The software works effectively in this E. Key Gas Method
operating system. Figure 1 shows the flow chart of DGA This method employs principal gases to detect the
software in which user enters DGA information and can presence of faults in the power transformer; these gases are
choose whether to diagnose by any of the five methods, H2, CO, CH4, C2H6, C2H4 and C2H2 whereby the significant
compare or trend. After each operation (diagnosis, comparison and proportion of the gases are called “key gases”. The fault is
or trending) DGA software displays a report which includes identified according to the presence and percentage of each
method execution, graphical representation and analysis. It key gas [9], [10]. Table II below shows the interpretation of
also saves the report in a plain text document which can easily different faults by key gas method.
be viewed by document viewer. Other plain text documents
can be seen through the software while (for further TABLE II COMBUSTIBLE GAS IN KEY GAS ANALYSIS [9]
Gas Normal Abnormal Interpretation
evaluations) others can be viewed by computer document
H2 < 150 ppm > 1,000 ppm Arcing, Corona
viewer (for report purposes). CH4 < 25 ppm > 80 ppm Sparking
C2H6 < 10 ppm > 35 ppm Local Overheating
TABLE I TOOLS USED FOR THE DEVELOPMENT OF DGA-SOFTWARE C2H4 < 20 ppm > 100 ppm Severe Overheating
Tool Type CO < 500 ppm > 1,000 ppm Severe Overheating
CO2 < 1,000 ppm > 15,000 ppm Severe Overheating
Computer HP 550 Duo Core N2 1 – 10% NA NA
Operating System Ubuntu 14.04 O2 0.2 – 3.5% NA > 0.5% Combustibles

Programming Language Python 2.7

Compiler Eclipse F. Dornenburg Ratio Method
Database MySQL 5.5 This method first examines individual gas, and
recommends a test if any gas is beyond the set up limit in parts
per million as shown in Table III. Three faults can be detected
by this method while five gases and four ratios are used for
detecting faults as shown in Table IV. Dornenburg
interpretation is described in Table V.
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 3

H. IEC Ratio Method

TABLE III
Five gases, four gas ratios and seven fault types are
DORNENBURG LIMITS FOR INDIVIDUAL GAS, L1 [10], [11]
Gas Limit, ppm involved in this method as described in Table VIII and Table
H2 100 IX below.
CH4 120
C2H4 50 TABLE VIII IEC GAS RATIOS [11]
C2H2 35
Ratio Gas
C2H6 1
CO 350 Ratio 1 CH4 / H2
Ratio 2 C2H2 / C2H4
TABLE IV
Ratio 5 C2H4 / C2H6
DORNENBURG GAS RATIOS [10], [11], [12]
Ratio Gas
Ratio 1 CH4 / H2 TABLE IX IEC FAULT INTERPRETATION [11]
Ratio 2 C2H2 / C2H4 Case Fault diagnosis C2H2/C2H4 CH4 / H2 C2H4 / C2H6
Ratio 3 C2H2 / CH4
Ratio 4 C2H6 / C2H2 Partial
PD - < 0.1 < 0.2
Discharge

TABLE V Low Energy

D1 >1.0 0.1 to 0.5 >1.0
DORNENBURG FAULT INTERPRETATION [10], [11] Discharges
Fault High Energy
CH4 / H2 C2H2/C2H4 C2H2/ CH4 C2H6/C2H2 D2 0.6 – 2.5 0.1 to 1.0 >2.0
diagnosis Discharge
Thermal
> 1.0 < 0.75 < 0.3 > 0.4 Thermal Fault
Decomp. T1 - >1.0 <1.0
Corona < 0.1 - < 0.3 > 0.4 <300 °C
Arcing 0.15 – 1.0 > 0.75 > 0.3 < 0.4 Thermal
T2 < 0.1 >1.0 1.0 to 4.0
300 - 700 °C

G. Roger’s Ratio Method T3

Thermal
< 0.2 >1.0 >4.0
> 700 °C
In previous, this method was using four gas ratios which
were CH4/H2, C2H6/CH4, C2H4/C2H6 and C2H2/C2H4 for
diagnosis whereby the ratio C2H6/CH4 only indicated a limited I. Duval Triangle Method
temperature range of decomposition, but did not assist in This method was developed by Michel Duval in 1974;
further identification of fault and hence was deleted. The it employs three hydrocarbon gases (CH4, C2H4 and C2H2) in
improved Roger’s method uses the remained three ratios to triangular map and detects seven fault types corresponding to
analyze six conditions of the power transformer [9], [10], [11], the increasing in energy levels of gas formation in
[12]. Two tables are used in improved Roger’s method: one transformers in service [3], [12].
defined the code of the ratio, and the other defined the
diagnosis rule as shown in Table VI and Table VII
respectively [12].

TABLE VI ROGER’S GAS RATIOS [11]

Ratio Gas
Ratio 1 CH4 / H2
Ratio 2 C2H2 / C2H4
Ratio 5 C2H4 / C2H6

TABLE VII ROGER’S FAULT INTERPRETATION [10], [11]

Case Fault diagnosis C2H2/C2H4 CH4 / H2 C2H4 / C2H6
0 Unit Normal < 0.1 >0.1 to <1.0 <1.0
Low energy density
1 < 0.1 < 0.1 <1.0
arcing
Arcing-High
2 0.1 – 3.0 0.1 to 1.0 >3.0
energy discharge
Low temperature-
3 < 0.1 >0.1 to <1.0 1.0 to 3.0
thermal Fig. 2. Duval Triangle
Thermal
4 < 0.1 >1.0 1.0 to 3.0
< 700 °C
Thermal The fault zone is identified by an intersection point between
5 < 0.1 >1.0 >3.0
> 700 °C parallel lines obtained from the percentages of fault gases as
shown in equation (1), (2) and (3).
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 4

%CH 4  100 xCH 4 CH 4  C2 H 2  C2 H 4  (1)

%C2 H 4  100 xC 2 H 4 CH 4  C2 H 2  C2 H 4  (2)

%C2 H 2  100 xC 2 H 2 CH 4  C2 H 2  C2 H 4  (3)

III. IMPLEMENTATION OF DGA COMPUTER SOFTWARE

A MySQL database was created and named as
“dissolved” and connected to the DGA software. It was saved
in a computer server as shown in a list of database in Figure 3. Fig. 5. Details of table in a database

DGA software was connected with the MySQL database by

using mysql-python connector.

Fig. 3. List of Databases in a computer server

A database has two tables but the one named “dga” will be
used as shown in Figure 4. It has eleven columns named ID
(primary key); hydrogen, carbon_monoxide, carbon dioxide, Fig. 6. The interface of DGA software
nitrogen, methane, ethane, ethylene, acetylene, power, voltage,
altitude, oil temperature, humidity, sampling_date, From the interface, user can opt to retrieve single or
machine_location, machine_number as shown in Figure 5. multiple DGA information from the database. Figure 7 shows
User has nothing to do with the database rather than saving single data selected from the database its ID which returns gas
and picking data which is done through the use of DGA and non gas details.
software

Fig. 4. List of tables in a database

The interface of the developed computer program is shown Fig. 7. Single data retrieved from the database
in Figure 6 with five diagnosing methods for interpreting
DGA information (fault diagnosis) and two condition Figure 8 shows multiple data selected from the database by
monitoring tools (comparison – two gases and trending – more their IDs which gives gas and non gas details.
than two gases) in the toolbar. Every DGA information when
is stored in the database, it automatically registered with a The interface of Key Gas diagnostic method of the DGA
unique ID which provides an easiest way to find it by using software is shown in Figure 9.
DGA software for diagnosis.
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 5

Fig. 8. Multiple data retrieved from the database Fig. 11. Diagnosis of Roger’s Ratio Method in DGA software

It includes user request for DGA data, DGA information, The interface of IEC Ratio diagnostic method of the DGA
computations, graphical representation and result analysis. software is shown in Figure 12 including user request for
DGA data, DGA information, computations and analysis.

Fig. 9. Diagnosis of Key Gas Method in DGA software Fig. 12. Diagnosis of IEC Ratio Method in DGA software

The interface of Dornenburg diagnostic method in the DGA The interface of Duval Triangle in the DGA software is
software is shown in Figure 10 including user request for shown in Figure 13 including user request for DGA data,
DGA data, DGA information, computations and result DGA information, computations and analysis.
analysis.

Fig. 10. Diagnosis of Dornenburg Ratio Method in DGA software Fig. 13. Diagnosis of Duval Triangle Method in DGA software
The interface of Roger’s Ratio diagnostic method in the After diagnosis DGA software saves the summary of
DGA software is shown in Figure 11 including user request diagnosis in plain text that can be easily viewed by a computer
for DGA data, DGA information, computations and analysis. viewer as shown in Figure 14.
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 6

100 xNumberof Pr edictions

%Success Pr ed .  (4)
NumberofCases

100 xCorrect Pr edictions

%Consistency  (5)
Numberof Pr edictions

TABLE X
INTERPRETATION SUMMARY OF FIVE METHODS FROM
DGA SOFTWARE
Method Number of % Successful Correct %
Predictions Prediction, P Predictions Consistency,
C
Key Gas 3 6 2 67
Dornen 8 16 6 75
Rogers 38 76 29 76
IEC 18 36 14 77
Fig. 14. Summary of diagnosis viewed by the DGA software Duval 50 100 48 96
The DGA software is able to compare data of the same
transformer sampled in different sampling periods and saved C. Accuracy of the Diagnostic Method
in the database as shown in Figure 15.
Furthermore, the diagnostic method was evaluated by its
accuracy. Reference [13] divided the accuracy into two
categories. The first category involves accuracy based only on
predicted cases (Tp) while the second one involves only the
total number of cases (Tc). Both cases can be calculated by
using equation (6) and (7) respectively.

100 xTR
Accuracyba sedon Pr ediction  (6)
TP

100 xTR
Accuracyba sedonTotalCases  (7)
TC
Fig. 15. Comparison of DGA data made by the DGA software
The overall accuracies between DGA software and from
others researchers can be seen in Table XII for better
IV. RESULTS understanding of DGA performance. The initials in Table XII
can be compared with the previous Table XI. In Table XII, D
A. Software Validation represented results from DGA software while O represented
50 DGA data from Oil Filled Transformers with different those of other researchers.
conditions were taken from other researchers [4], [12] for
testing the DGA software. The DGA data were first stored in TABLE XI
the DGA database and then retrieved and diagnosed by each COMPARISON OF ACCURACIES OF DGA DIAGNOSTIC METHODS
diagnostic method. KGM Dorn Rog IEC Duv
Total Cases, TC 50 50 50 50 50
B. Percentage of Successful Prediction and Consistency No Prediction, TNP 47 42 12 32 0
The number of predictions done by five diagnostic methods Number of Predictions,
3 8 38 18 50
with DGA data in Table X were used to calculate percentages TP
of successful predictions (P) and consistencies (C) as shown in Correct Predictions, TR 2 6 29 14 48
Table X. The formulas are shown in equation (4) and equation
Incorrect Predictions,
(5) as referred [13]. The summary of comparison of 1 2 11 4 2
TW
interpretation results obtained by using five diagnostic
Accuracy (Predicted
methods is also shown in Table X. 67 75 76 77 96
cases), AP
Accuracy (Total
4 12 38 28 96
cases), AT
2016 IEEJ P&ES – IEEE PES Thailand Joint Symposium on Advanced Technology in Power Systems 7

The summary of Table XII was plotted in a chart for easy [2] A. S. Alghamdi, “DGA Interpretation of Oil Filled Transformer
Condition Diagnosis”, Trans. Electrical and Electronic Materials,
interpretation as seen in Figure 14. vol. 13, pp. 229-232, Oct. 2012.
[3] A. Akbari, A. Setayeshmehr, H. Borsi and E. Gockenbach., “Software
TABLE XII Implementation of the Duval Triangle Method”, in 2008 Proc. IEEE Int.
COMPARISON OF ACCURACIES OF DGA AND OTHER Symp. Electrical Insulation, Vancouver, pp: 124-127.
RESEARCHERS [4] O. Gouda, S. Saleh, and S. EL-Hoshy, “Power Transformer Incipient
KGM Dorn Rog IEC Duval Faults Diagnosis Based on Dissolved Gas Analysis”,
D O D O D O D O D O TELEKOMNIKA Indonesian J. Electr. Eng., vol. 17, no. 1, pp. 10-16
January, 2016.
TC 50 50 50 50 50 50 50 50 50 50
[5] S. Singh and M.N Bandyopadhyay, “Duval Triangle: A Noble
TNP 47 46 42 4 12 20 32 6 0 0 Technique for DGA in Power Transformers,” Int. J. Electr. Power
Eng., vol. 4, no. 3, pp: 193 – 197, 2001.
TP 3 4 8 46 38 30 18 44 50 50
[6] R.E. James and Q. Su, Condition Assessment of High Voltage Insulation
TR 2 3 6 20 29 18 14 38 48 44 in Power System Equipment, IET: UK, 2008.
[7] M. Lutz, Powerful Object-Oriented Programming: Programming
TW 1 1 2 14 11 12 4 6 2 6 Python, O’Really Media, Inc: USA, 2011, ch. 1.
[8] R. Dyer, MYSQL in a Nutshell: A Desktop Quick Reference, O’Really
AP 67 75 75 43 76 60 77 86 96 88 Media, Inc: USA, 2005, ch. 1.
[9] H. Sun, Y. Huang, and C. Huang, “A Review of Dissolved Gas Analysis
AT 4 6 12 40 38 36 28 76 96 88 in Power Transformers,” Int. Conf. Advances in Energy Eng, Energy
Procedia vol. 14, pp. 1220 – 1225, 2012.
[10] IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
The Figure 16 is a plot of AP between DGA software and Transformers, IEEE Standard C57.104, 2008.
other researchers systems. [11] R. Kumar, “Different DGA Techniques for Monitoring of
Transformers,” Int. J. Electronics and Electrical Eng, vol. 1, no. 4,
December, pp. 299-303, 2013.
OTHERS 100 [12] S. Ghoneim and S. Ward, “Dissolved Gas Analysis as Diagnostic Tools
for Early Detection of Transformer Faults,” Advances Electrical Eng.
DGA 80 Syst, vol. 1, no. 3, pp. 152-156, 2012.
[13] N. A. Muhamad, B. T. Phung, T. R. Blackburn, and K. X. Lai,
60 “Comparative Study and Analysis of DGA Methods for Transformer
Mineral Oil”, IEEE Power Tech., pp. 45-50, July, 2012.
40
20 Pius Victor was born in Dar es Salaam, Tanzania on 15 th
October, 1986. He obtained his B.Eng. degree in Electrical
0 Engineering from Dar es Salaam Institute of Technology,
KGM DON ROG IEC DUV Tanzania in 2013. He is now a Master’s Degree student in
the School of Electrical Engineering, Suranaree University
of Technology in Thailand. His interest research topics
OTHERS 75 43 60 86 88 include High Voltage Systems Design and Monitoring, Laboratory and
67 75 76 77 96 System Programming.
DGA

Fig. 16. Diagnosis of Duval Triangle Method in DGA software

V. CONCLUSION
The DGA software has successfully developed and database
has connected to it. The accuracy comparison was done for
DGA software and other researchers systems and found to be
higher especially in Duval Triangle. The DGA software has
also shown the great ability of diagnosing large number of
DGA information from distribution network with many
transformers with the presence of database at minimum time.
The DGA data were saved only once and the diagnostic
processes were able to be repeated several times by retrieving
data from the database hence minimizing computational time,
reduces inaccuracy in diagnosis, provides an easy and quick
time-to-time condition monitoring and also it gives a safe
storage of DGA data for future uses. The summary of
diagnosis after interpretation and saved can be used by users
in other times without rerunning again the software.

REFERENCES
[1] P. Pao-la-or, A. Isaramongkolrak and T. Kulworawanichpong, “Finite
Element Analysis of Magnetic Field Distribution for 500kV Power
Transmission Systems,” IEEE Electr. Insulation Magn., vol. 21, pp. 21 –
27, 2005.

View publication stats