...

Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in Electrical Transformers

by user

on
2

views

Report

Comments

Transcript

Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
Conventional Ratio and Artificial
Intelligence (AI) Diagnostic methods for
DGA in
Electrical Transformers
Amin Samy, Sayed A. Ward, Mahmud N. Ali
Benha University
[email protected]
Abstract: Transformers are always under the impact of
electrical, mechanical, thermal and environmental stresses that
degrade their insulation quality. To avoid the power failure,
periodical monitoring of the conditions of transformers is
necessary. Results of early detection of fault can provide large
savings in operation and maintenance costs and preventing any
premature breakdown/failure. In this paper the DGA
(dissolved gas analysis) is studied and the different diagnosis
methods are discussed. The conventional ratio methods and the
artificial intelligence (AI) Diagnostic methods for DGA in
electrical transformers are presented. A comparison between
the two different diagnoses methods for a certain transformers
already existed in the Egyptian electrical network is presented.
These results are also compared to the results received from
the central laboratories of the ministry of electricity in Egypt.
The simulated results show that the ratio of matched results
from the artificial intelligence diagnostic methods (using
neural network) was higher than the ratio of matched results
from the conventional ratio methods.
Keywords: dissolved gas analysis, DGA, conventional ratio
methods, artificial neural network.
I. INTRODUCTION
DGA is an effective test for the early detection of
incipient faults. DGA is considered the most important oil
test for insulating liquids in electrical apparatus. More
importantly, an oil sample can be taken at anytime from
most equipment without having to take it out of service
[2].Degradations produce fault-related gases such as
hydrogen (H2), methane (CH4), ethylene (C2H4), acetylene
(C2H2), ethane (C2H6), carbon dioxide (CO2), and carbon
monoxide (CO).The gas concentrations, generation rates,
specific gas ratios, and the total combustible gas are
important parameters for interpreting the result of DGA [1].
There are several conventional ways for the diagnosis
of the transformer fault using the DGA method which
include the key gas analysis, Rogers Ratio method, IEC gas
ratio code(IEC-60599), DORNENBURG Ratio method and
Duval triangle method [2].
These conventional diagnostics methods don’t always
yield an analysis, miss too many incipient faults and may
lead to the “no decision” problem. On the other hand the
artificial intelligence diagnostic method is more useful tool
for transformer diagnosis and maintenance planning [2].
This paper presents the comparison between the results
from using conventional diagnostic methods (ratio method)
and the results from artificial intelligence diagnostic method
(Neural Network) for studying actual cases of transformers
in service which exist in the electrical networks according to
the results of central laboratories of the ministry of
electricity in Egypt.
This paper presents the following sections: in section 2,
brief descriptions of Conventional Diagnostic methods for
DGA are presented. Section 3 presents a brief description of
artificial intelligent Diagnostic methods for DGA. Section 4
presents case studies of (66/11 kV) power transformers. In
section 5 the case studies of (220/66 kV and 500/220 kV)
power transformers are presents. Section 6 presents the
conclusion.
2096
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
II. Conventional diagnostic methods
There are many conventional diagnostic methods for DGA.
These methods comprise key gas method[2], Duval triangle
method [3][4] and conventional ratio methods, which
include (DORENENBURG method [2], Rogers method [5]
and IEC 599 method [6], The conventional ratio methods
use the ratios of dissolved gas concentrations as the basis of
fault diagnosis. Historically five ratios, presented in Table 1,
have been used [2].
Rogers’s method depends on four ratios R1, R2, R4, and
R5. But it may give no conclusion in some cases. This is the
“no decision” problem [8].This method was further
modified into an IEC gas ratio code (IEC599).The refined
Rogers method used a table which defined the code for the
ratios, as shown in table 3.
Table 3- The codes used in the Rogers ratio method
R1
R4
R5
R2
Diagnosis
0
0
0
0
Normal deterioration
5
0
0
0
Partial discharge
1 or 2
0
0
0
Slight overheating – below 150⁰
C
1 or 2
1
0
0
Slight overheating –150 – 200 ⁰
C
0
1
0
0
Slight overheating – 200 – 300 ⁰
C
0
0
1
0
General conductor overheating
1
0
1
0
Winding circulating currents
1
0
2
0
Core and tank circulating
currents, overheated joints
0
0
0
1
Flashover without power follow
through
0
0
1 or 2
1 or 2
Arc with power follow through
2.1 DORNENBURG method
0
0
2
2
Continuous sparking to floating
potential
In 1970 DORNENBURG was able to differentiate
between thermal and electrical faults using four ratios and
six gases. The six gases are H2, CH4, CO, C2H2, C2H4 and
C2H6 and the four ratios are R1, R2, R3, and R4. This
method has many validation tests before reaching the final
decision and it often fails to do so [8].
The most important validation test is the L1 ‘normal
test’, which sets up a critical level for each gas. In order to
apply the method, at least one gas for each of the ratios must
exceed the corresponding L1 normal value [8].
5
0
0
1 or 2
Partial discharge with tracking
(note CO)
Table 1 - Ratio definition of the conventional ratio methods
Ratio
Abbreviation
CH4/
C2H2/C
C2H2/
C2H6/C2
C2H4/
H2
2H4
CH4
H2
C2H6
R1
R2
R3
R4
R5
All ratio methods are applied only if at least one of the
gases is at a concentration and a rate of gas increase above
the following typical values [7] which are presented in
Table 2.
Table 2 - Gas concentration in PPM (Part per Million)
H2
150
CH4
110
C2H6
90
C2H4
280
C2H2
CO
50
900
2.2 Rogers ratio method
CO2
13000
2.3 IEC Gas Ratio Code (IEC599)
This technique was standardized by IEC in 1978 [6].
The DGA results which cannot be matched by the existing
codes lead to unsuccessful diagnosis. In multiple fault
conditions, gases from different faults are mixed up
resulting in confusing ratios between different gas
components. This could only be dealt with by the aid of
2097
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
more sophisticated analysis methods such as the intelligence
diagnosis method.
is chosen as an example. The PPM of C2H2 in all cases
equal zero.
III. Artificial Intelligent Diagnostic methods
Table 4- Diagnostic data history for (66/11 Kv) transformer
In the past decade, there has been extensive research on
the use of artificial intelligence techniques to assist the
DGA. These investigations include the expert system
approach, fuzzy system approach and the artificial neural
network (ANN) approach [9].
The basic idea of neural network based fault diagnosis is
nonlinear mapping [8]. It is assumed that the relationships
between the input vector X and the output vector Y are
predefined by the physical nature of the problem, and these
relationships can be represented by a limited number of
input-output pairs (data samples).
The application of neural network in fault diagnosis has
two phases. Phase 1 is the training process, during which the
data samples are provided to the network, the memorial
coefficients of the network are iteratively adjusted to
“memorize” the input-output relationships. Phase 2 is a
testing process, during which the input vector x is fed into
the network, and the memorized coefficients of the network
are recalled to “discover” the possible output [10].
To presents the differences between using the
conventional methods and the artificial intelligent methods
in diagnostic the results of DGA, some case studies of
different real power transformer (66/11 Kv), (220/66 Kv)
and (500/220 Kv) are presented in the next sections.
IV. Case studies of (66/11 Kv) power transformers
In this section, firstly, DORNENBURG method, as
conventional ratio method, and programming IEC599
method using C++ language are compared. Secondly, these
results are compared with the formal results from Central
laboratories of the Ministry of Electricity in Egypt. Finally,
all these results are compared with artificial neural network
(ANN) diagnostic results.
Date of
sample
Some of the dissolved gases (ppm)
H2
CH4
C2H4
C2H6
Conclusion
Thermal fault of
high temp. >
700°C
Hot spots T<
200°C
Thermal fault
150-300°C
Thermal fault
<150°C
2003
4
7
4
1
2004
44
7
5
2
2005
2
12
5
9
2006
3
3
8
4
2008
9
6
12
5
Normal
2009
20
3
14
4
Normal
2010
36
24
4
5
Normal
2012
61
1
2
5
Normal
4.1.1 Using conventional methods
Applying the conventional methods (DORNENBURG
and IEC 599 method), it is found that the Programming IEC
599 method using C++ language matched the
Recommendation of Central laboratories of the Ministry of
Electricity in Egypt as presents in table 5.
Table 5- Diagnostic using conventional methods for (66/11 Kv)
transformer
Test name
Result
DORNENBURG method
(as conventional ratio method)
Not pass the criteria of a fault
exists
So, there is no fault.
4.1 (66/11Kv) transformers
In the first case study, the conventional methods and the
artificial intelligent method are applied to the (66/11 Kv)
transformer which exist in the EGYPT ministry of
electricity. Table 4 presents an example of the data history
of (66/11 Kv) transformers. For applying the conventional
methods and the artificial intelligent method, case 5 on 2008
2098
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
Programming IEC 599 method
using C++ language
H2=9
CH4=6
C2H6=5
C2H4=12
C2H2=0
CO=90
CO2=2796
Normal DGA/healthy equipment
Press any key to continue...
Central laboratories of the Ministry
of Electricity in Egypt
Recommendation
-
Ratio of the
average matched
results
methods
Diagnostic methods
50 %
87.5 %
V. Case studies of (220/66 Kv and 500/220 Kv) power
transformers
As presented in the previous section, other case studies,
which are (220/66 Kv and 500/220 Kv) transformers, are
presented in the next sections.
Normal
5.1 (220/66) Kv transformer
4.1.2 Using artificial intelligent method
Neural network training requires input and output
patterns. Input patterns are obtained from samples of DGA.
For each input pattern there exists an output pattern which
describes the fault type for a given diagnosis criterion.
The input pattern is a vector of the three ratios R1, R2
and R5. The output pattern is a vector of six elements
according to the transformer conditions. For example,
transformer normal condition is represented by a unit value
in the first element and the remaining zero: [1,0,0,0,0,0],
And an overheating >700C is represented as
follows:[0,0,0,0,0,1] as presented in table 6. The conclusion
for this case study of the (66/11 Kv) transformers are
presented in table 7.
Table 8 presents an example of the data history of
(220/66 Kv) transformers. For applying the conventional
methods and the artificial intelligent method, case 1 on 2002
is chosen as an example. The PPM of C2H2 in all cases
equal zero.
Table 8- Diagnostic data history for (220/66 Kv) transformer
Some of the dissolved gases(ppm)
Date of
sample
H2
CH4
C2H4
C2H6
2002
29
10
2
4
2003
19
46
5
20
2004
7
15
3
3
2005
10
3
2
5
2/2006
16
24
3
10
8/2006
33
52
6
14
2/2007
9
26
2
5
Normal
8/2007
12
12
3
4
Normal
2008
14
17
2
5
2009
3
11
3
11
2010
7
5
4
13
1/2011
44
20
2
10
Conclusion
Table 6- Training patterns for ANN method for (66/11 Kv) transformer
Date of
sample
R2=
C2H2/
C2H4
R1=
CH4
/H2
R5= C2H4
/C2H6
Conclusion
2003
2004
2005
2006
2008
2009
2010
2012
0
0
0
0
0
0
0
0
1.75
0.16
6
1
0.666
0.15
0.666
0.016
4
2.5
0.555
2
2.4
3.5
0.8
0.4
0,0,0,0,0,1
0,1,0,0,0,0
0,0,0,1,0,0
0,0,1,0,0,0
1,0,0,0,0,0
1,0,0,0,0,0
1,0,0,0,0,0
1,0,0,0,0,0
Table 7 – The conclusion for (66/11 Kv) transformer
Item
Using Conventional
Ratio Diagnostic
Using Artificial
Intelligence (AI)
Normal
Thermal fault
150-300°C
Thermal fault
150-300°C
Normal
Thermal fault
150-300°C
Thermal fault
150-300°C
Fault involving
celluluse
Thermal fault
300-700°C
Thermal fault
300-700°C
Thermal fault <
300°C
2099
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
7/2011
22
16
6
6
Normal
2012
16
30
2
8
Thermal fault <
300°C
5.1.1
Using conventional methods
Using conventional methods (DORNENBURG and
IEC 599 method), it is found that the Programming IEC
599 method using C++ language matched the
Recommendation of Central laboratories of the Ministry
of Electricity in Egypt as presents in table 9.
2002
2003
2004
2005
2/2006
8/2006
2/2007
8/2007
2008
2009
2010
1/2011
7/2011
2012
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.345
2.42
2.143
0.3
1.5
1.5758
2.89
1
1.214
3.66
0.714
0.45
0.727
1.875
0.5
0.25
1
0.4
0.3
0.43
0.4
0.75
0.4
0.273
0.31
0.2
1
0.25
1,0,0,0,0,0
0,0,0,1,0,0
0,0,0,1,0,0
1,0,0,0,0,0
0,0,0,1,0,0
0,0,0,1,0,0
1,0,0,0,0,0
1,0,0,0,0,0
0,0,1,0,0,0
0,0,0,0,1,0
0,0,0,0,1,0
0,0,0,1,0,0
1,0,0,0,0,0
0,0,0,1,0,0
Table 11 – The conclusion for (220/66 Kv) transformer
Table 9- Diagnostic using conventional methods for (220/66 Kv)
transformer
Test name
Result
DORNENBURG method
(as conventional ratio method)
- Not pass the criteria of a fault
exists
- So, there is no fault.
Programming IEC 599 method
using C++ language
H2=29
CH4=10
C2H6=4
C2H4=2
C2H2=0
CO=156
CO2=1719
Normal DGA/healthy
equipment
Using Conventional
Ratio Diagnostic
methods
Item
Ratio of the
average matched
results
Using Artificial
Intelligence (AI)
Diagnostic methods
36 %
92.86 %
5.2 (500/220Kv) transformer
Table 12 presents an example of the data history of
(500/220 Kv) transformers. For applying the conventional
methods and the artificial intelligent method, case 3 on
2007 is chosen as an example. The PPM of C2H2 in all
cases equal zero.
Table 12- Diagnostic data history for (500/220 Kv) transformer
Press any key to continue...
Central laboratories of the Ministry
of Electricity in Egypt
Recommendation
5.1.2
-
Normal
Using artificial intelligent methods
The input and output patterns are presented in table 10,
And the conclusion for this case study of the (220/66 Kv)
transformers are presented in table 11.
Table 10- Input and output patterns of ANN methodfor (220/66 Kv)
transformer
Date of
sample
R2=
C2H2/
C2H4
R1=
CH4
/H2
R5=
C2H4
/C2H6
Conclusion
Date of
sample
H2
Some of the dissolved gases
CH4
C2H4
C2H6
6/2006
3
20
3
12
12/2006
10
9
26
20
6/2007
8
13
2
1
12/2007
10
6
1
1
3/2008
9
206
5
25
4/2008
6
4
2
1
6/2008
12
3
1
1
10/2008
22
4
8
8
5.2.1
Conclusion
Thermal fault
150-300°C
Thermal fault
<150°C
Normal
Thermal fault
300-700°C
Thermal fault
<300°C
Thermal fault
300-700°C
Thermal fault
300-700°C
Thermal fault
>700°C
Using conventional methods
2100
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
Using conventional methods (DORNENBURG and IEC
599 method), it founds that the Programming IEC 599
method using C++ language matched the Recommendation
of Central laboratories of the Ministry of Electricity in
Egypt as presents in table 13.
Table 13- Diagnostic using conventional methods for (500/220 Kv)
transformer
Table 15 – The conclusion for (500/220 Kv) transformer
Item
Using Conventional
Ratio Diagnostic
methods
Using Artificial
Intelligence (AI)
Diagnostic methods
Ratio of the
average matched
results
37.5 %
87.5 %
V. Conclusion
Test name
Result
DORNENBURG method
(as conventional ratio method)
Programming IEC 599 method using
C++ language
-Not pass the criteria of a
fault exists
- So, there is no fault.
H2=8
CH4=13
C2H6=1
C2H4=2
C2H2=0
CO=184
CO2=2592
Normal
equipment
DGA/healthy
Press any key to continue...
Central laboratories of the Ministry of
Electricity in Egypt
Recommendation
5.2.2
-
Normal
REFERENCES
Using artificial intelligent methods
The input and output patterns are presented in table 14,
And the conclusion for this case study of the (500/220 Kv)
transformers are presented in table 15.
Table 14- Input and output patterns for ANN methodfor (500/220 Kv)
transformer
Date of
sample
6/2006
12/2006
6/2007
12/2007
3/2008
4/2008
6/2008
10/2008
R2=
C2H2/
C2H4
0
0
0
0
0
0
0
0
R1=
CH4
/H2
6.66
0.9
1.625
0.6
23.22
0.666
0.25
0.182
In this paper the conventional ratio methods and the
artificial intelligence (AI) Diagnostic methods for DGA
in electrical transformers are presented. A comparison
between the two different diagnoses methods for certain
transformers, already existed in the Egyptian electrical
network, is presented. For different case studies, the
DGA results show that the conventional ratio methods
are unable to diagnose multiple faults. Some DGA
results fall outside the ratio codes and the “no decision
problem” occurs. In addition, sufficient gases are
needed (gases need to exceed a minimum level). On the
other hand, The DGA results using the ANN method
provides promising results compared to the
conventional methods.
R5=
C2H4
/C2H6
0.25
1.3
2
1
0.2
2
1
1
Conclusion
0,0,0,1,0,0
0,0,1,0,0,0
1,0,0,0,0,0
0,0,0,0,1,0
0,0,0,1,0,0
0,0,0,0,1,0
0,0,0,0,1,0
0,0,0,0,0,1
[1]. Domerburg E., Strittmatter W.,, "Monitoring oil cooling
transformers by gas analysis”, Brown Boveri Review, vol 61.
238-247, 1974
[2]. ANSI/IEEE Std C57.104-1991, IEEE Guide for The
Interpretation of Gases Generated in Oil-Immersed
Transformers, IEEE Power Engineering Society, 1992
[3]. M. Duval and A. depablo, “Interpretation of Gas-in-Oil
analyses using new IEC Publication 60599 and IEC TC10
Databases”,
IEEE
Electrical
Insulation
Magazine,Vol.17,No.2,P.31, (2001).
[4]. M. Duval, “A Review of faults Detectable by Gas-in-Oil
Analysis in Transformers”, IEEE Electrical Insulation
Magazine, Vol.18, No.3, P.8 (2002).
[5]. R. R. Rogers, ”IEEE and IEC Codes to interpret incipient faults
in transformers, using gas in oil analysis”, IEEE Trans. Elec.
Insul. Vol.13, no.5, 349-354, 1978.
[6]. R. E. Brown, “budget constrained planning to optimize power
system
reliability
”,IEEE
Transactions
on
power
systems,Vol.15,No.2,May 2000
2101
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
International Electrical Engineering Journal (IEEJ)
Vol. 6 (2015) No.12, pp. 2096-2102
ISSN 2078-2365
http://www.ieejournal.com/
[7]. N.N.: IEC 60599: Mineral Oil-Impregnated equipment in
service-Guide to the interpretation of dissolved and free gases
analysis.1997.
[8]. Zhenyuan Wang, ,“Artificial Intelligence Applications in the
Diagnosis of Power Transformer Incipient Faults“,
,Chapter2,p1-12, 2000
[9]. Sherif S.M. Ghoneim,Sayed A. Ward,‘‘ Dissolved gas analysis
as a diagnostic tools for early detection of transformer faults”,
Advances in Electrical Engineering Systems (AEES), Vol. 1,
No. 3,ISSN,2167–633X,PP. 152-156, United States, 2012.
[10]. Zhang, Y, Ding, X. and Liu, Y,, "An artificial Neural network
Approach to Transformer fault diagnosis", IEEE Transactions
on power Delivery, Vol. 11, No. 4, 64-66, 1996
2102
Samy et. al.,
Conventional Ratio and Artificial Intelligence (AI) Diagnostic methods for DGA in
Electrical Transformers
Fly UP