1836 IEEE Transactions on Power Delivery, Vol. 11, No.

4, October 1996

An Artificial Neural Ne to Transformer Fault ~ i ~ g ~ o ~ i ~

Y. Zhang, X. Ding, Y. Liu, P. J. Griffin

The Bradley Department of Electrical Engineering Doble Engineering Company
Virginia Tech, Blacksburg, VA 24061-0111,USA Watertown, MA 02172,USA

Abstract: while the fuzzy-set represents the decision rules by using

This paper presents an artificial neural network (ANN) vague quantities. Artificial neural network method
approach to diagnose and detect faults in oil-filled power (ANN)[2][3] has also been used for this purpose since the
transformers based on dissolved gas-in-oil analysis. A two- hidden relationships between the fault types and dissolved
step ANN method is used to detect faults with or without gases can be recognized by ANN through training process.
cellulose involved. Good diagnosis accuracy is obtained with
the proposed approach. A two-step ANN approach is presented in this paper. The
accuracy of the ANN is carefully verified. With two ANNs,
Keywords: Artificial neural network, Dissolved gas analysis high diagnosis accuracy is obtained.
(DGA), Fault detection and diagnosis, transformers.
2. DISSOLVED GAS ANALYSIS
1. INTRODUCTION
Different patterns of gases are generated due to different
As a major apparatus in a power system, the power intensities of energy dissipated by various faults. Totally or
transformer is vital to system operation. Techniques for partially dissolved into the oil, the gases present in an oil
diagnosis and incipient-fault detection are valuable. sample make it possible to determine the nature of fault by the
gas types and their amount. Dissolved gas analysis (DGA)
A transformer is subject to electrical and thermal stresses. method is a widely used technique for transformer fault
These two stresses could break down the insulating materials diagnosis. One of the DGA methods was derived from the
and release gaseous decomposition products. Overheating, Halstead’s discovery. Halstead made a theoretical-
corona and arcing are three primary causes of fault related thermodynamic assessment of the formation of the simple
gases. Principally, the fault related gases commonly used are decomposition gaseous hydrocarbons[4]. It’s stated that the
hydrogen (H,), carbon monoxide (CO), carbon dioxide (CO,), rate of evolution of any particular gaseous hydrocarbon varies
methane (CH,), acetylene (C,H,), ethane (C,H6), and ethylene with temperature, and that at a particular temperature there
(C,H,). The analysis of dissolved gases is a powerful tool to would be a maximum rate of evolution of that gas. Study of
diagnose developing faults in power transformers. Many the Halstead thermodynamic equilibrium suggested that with
diagnostic criteria have been developed for the interpretation increasing temperature, the hydrocarbon gas with maximum
of the dissolved gases. These methods would find the rate of evolution would in turn be methane, ethane, ethylene,
relationship between the gases and the fault conditions, some and acetylene. Halstead’s work proved the existence of
of which are obvious and some of which may not be apparent relationship between fault temperature and the composition of
(hidden relationships). However, much of the diagnostics dissolved gases. Efforts have been made to create simplified
relies on experts to interpret the results correctly. New diagnosis criteria such as the key gas method and the ratio
computer-aided techniques can consistently diagnose method which in essence are based on this variations in
incipient-fault conditions for the novice and in some cases gassing characteristics with temperature to which the materials
may provide further insight to the expert. Expert system and are subjected.
fuzzy-set approach have been developed to reveal some of the
hidden relationships in transformer fault diagnosis [13. Expert a) Key gas method
system derives the decision rules from the previous experience
Characteristic “Key gases” have been used to identify
particular fault types[5]. The suggested relationship between
96 WM 225-3 PWRD A paper recommended and approved by the IEEE key gases and fault types is summarized as:
Transformers Committee of the IEEE Power Engineering Society for
presentation at the 1996 IEEUPES Winter Meeting, January 21-25, 1996,
Baltimore, MD. Manuscript submitted August 1, 1995; made available for H,----Corona
printing January 2, 1996. 0, & N,----Non-fault related gases
CO & CO,----Cellulose insulation breakdown
CH, & C&----Low temperature oil breakdown
C,H,----High temperature oil breakdown
C,H,----Arcing

0885-8977/96/$05.00 0 1996 IEEE

 1837
Excluding 0, and N2, there are seven fault related gases. The other than the inputs and outputs are called hidden layers) and
fault condition is indicated by the excessive generation of nodes (computation units which take inputs from a lower
these gases. Since this method does not give the numerical layer, perform a specific mathematical function, and produce a
correlation, the diagnosis depends greatly on experience and output) according to training performance and prediction
therefore this technique is simple yet labor intensive. accuracy. Figure 1 illustrates an overall ANN design process
with step by step adjustments to achieve desired structure and
feature space. The corresponding connection weights are also
b) Ratio Methods determined in the process. In this paper, back propagation
learning algorithm, alternatively known as the generalized
Rogers, Dornenberg and IEC are the most commonly used delta rule, is used for ANN training[8]. The back propagation
ratio methods. They employ the relationships between gas learning algorithm consists of repeatedly passing the training
contents. The key gas ppm values are used in these methods to set through the neural network until its weights minimize the
generate the ratios between them. The ranges of the ratio are output errors over the entire set.
assigned to different codes which determine the fault types
[6]. Coding is based on experience and is always under Once the process is done, all weights will be retained and the
modification. Ratio methods are limited in discerning ANN is now "trained" and ready for use. New samples will be
problems when more than one type of fault occurs fed into the trained ANN and predicted values can be readily
simultaneously. In addition, for some cases there is no obtained as shown in Figure 1.
diagnosis for a code as there are more possible combinations
of the code than there are for the number of diagnosis.

The fuzzy-set approach and expert system have been used to

incoqporate various rules. A knowledge base or a fuzzy
membership function is selected based on the past experience.
The fault diagnosis is a weighted conclusion drawn from a
number of data pertinent to the equipment. Its reliability change n o d e s 7 7

increases with the amount 'of information available from

previous tests and the degree of experience of the laboratory
performing the analysis. Therefore, the knowledge base compare iterations
required could be large and complex. does it have less iterations 7
J, yes
I working output I
no
3. ARTIFICIAL NEURAL NETWORK

Very complex systems can be characterized with very little

explicit knowledge using ANNs.The relationship between gas
composition and incipient-fault condition is learned by the
ANN from actual experience (through training samples).
Obvious and not so obvious (hidden) relationships are 'MSE -Mean Sauare Ermr
detected by the ANN and used to develop its basis for
interpretation of dissolved gas-in-oil data. Through training
process, ANN can reveal complex mechanism that may be Fig. 1 ANN Design and Training Flow Chart
unknown to experts. In contrast, expert and fuzzy-set
systems can only use explicit knowledge to establish If a mapping function (relationship) exists between the input
knowledge base and fuzzy membership function selection. and output, finding weights and biases for the neural network
Theoretically, a neural network could represent any is equivalent to finding the input/output relationship that
observable phenomenon. describes the mapping function. Thus, neural network is
especially appropriate and powerful when used to find
An ANN design includes selection of input, output, network relationships that are difficult to describe explicitly, since the
topology (structure, or arrangement of nodes), and weighted weights and biases can represent a given function.
connections of the nodes. Input feature (information) selection
constitutes an essential first step. The feature space needs to
be chosen very carefully to ensure that the input features will 4. TWO-STEP APPROACH TO FAULT ]DIAGNOSIS
correctly reflect the characteristics of the problem. The
procedure is problem dependent, as described in detail in [7]. Since there are already various diagnostic criteria in use, the
significance of using ANN is to achieve better diagnosis
Another major task of the ANN design is to choose network performance. It is also useful to distinguish the fault in
topology. This can be done experimentally through a repeated cellulose from that in oil.
process to optimize the number of hidden. layers (See Fig. 2 To increase the accuracy of the ANN, it is necessary to
and 3, a horizontal row of nodes is called a layer. The layers choose the proper input features, output nodes and ANN
1838
topology. Overheating, corona and arcing are three major fault and testing results of these four combinations are listed in
types. Since each of them could involve cellulose breakdown, Tables (1) through (4).
there could be seven patterns to be identi,fied, including the
normal condition.
Table( 1) Input: H,,CH,,C,H,,C,H,,C,H,,CO;
Often there are not a large number of data available for the Output: Normal, Overheating, Corona and Arcing.
ANN training and performance verification. Thus, Ten-fold
cross-validation is used to train and test the ANN, which is
typically used in ANN accuracy verification. This is done as
follows. Data are equally divided into ten groups. Alternately,
data of each group are tested by the ANN trained with the
other nine groups of data. This approach effectively increases Ten -Fold Cross-Validation: 90%
the number of data for ANN testing.
* (Here 6-50-4 indicates 6 input nodes, 50 hidden
nodes, and 4 output nodes)
4.1 ANN for Major Fault Type Diagnosis

Theoretically the above diagnosis and detection problem can

Table(2) Input: H,,CH,,C,H,,C,H,,C,H,,CO;
be realized by one ANN using all 7 fault related gases as
Output: Overheating, Corona and Arcing.
inputs. It was found in a previous study that the ANN with
CO, as one of the input feature has worse performance in
major fault type (arcing, corona, overheating) diagnosis than
that without CO,[22]. CO and CO, are needed as the input
features since they are the primary gases formed from the
degradation of cellulose. For this reason, two ANNs are used
to separate the cellulose condition detection from major fault
type diagnosis. One ANN is trained only to classify the major
fault types while another focuses on determining if the
cellulose is involved. Table(3) Input: H,,CH,,C,H,,C,H,,C,H,; ‘

Output: Normal, Overheating, Corona, Arcing.

In[2], an ANN with six input features and four output nodes
was found to be a better choice. There, H,, CH,, C,H,, C,H,, Topology Iteration
C,H, and CO were chosen as the input features, while the four 5-30-20-20-4 898 0.008682
output nodes were normal, overheating, corona and arcing 5-40-20-4 663 0.009675
I 1 I
(three major fault types). Since reducing the irrelevant input
features and output nodes may affect the ANN performance,
the ANN obtained in[2] is refined in the present study to
obtain the optimal design for the major fault diagnosis. The
correlation between CO content and major fault types are
checked since it is a cellulosic breakdown related gas. It is Table(4) Input: H,,CH,,C,H,,C,H,,C,H,;
also interesting to see if the “normal” output node is Output: Overheating, Corona and Arcing.
necessary, since no fault means normal condition. Details are
given in this section.

In our study, five key gases H,, H,, C,H,, C,H6, C,H, are
chosen as input features. Both the cases with and without
“CO” as input features have been studies. The cases with and
without “normal” output node are also considered. In total
there are four combinations to be tested and compared. The
ANN is trained using the conventional back-propagation Comparing the accuracy, iteration number and network
method. Since the ANN topology has great impact on complexity, it is shown that the two-hidden layer ANN
classification accuracy and convergence time during training, topology is the optimal for all four cases. All ten-fold cross-
several network topologies are compared. Usually, one validations are referred to the two hidden layer ANNs. As can
hidden-layer ANN is enough for most nonlinear mapping. be seen, an input feature space without “CO increases the
However, ANNs with 1,2 and 3 hidden layers are trained and accuracy. Keeping “normal” as an output node also increases
compared. According to the number of training iterations the accuracy. The second ANN in Table(3) is the optimal one
and training error, the optimal is determined. The ten-fold (among the 12) that has 95% chance of success in classifying
cross-validation of this optimal ANN is computed to verify the three major fault types. This chosen ANN topology is
the accuracy. Forty sample sets from different transformers shown in Figure 2.
are used for training and testing of the ANNs. The training
 1839
Normal Overheating Corona Arcing (outputs) Cellulose Involved

H,
I\ I\ I\ I\ I\
CH, C,H, C,H, C,H,
(inputs)
Hi CH, C,H, C;H, CiH, CO Cb,

Fig. 2 ANN for Major Fault Type Diagnosis Fig. 3 ANN for Cellulose Condition Detection

4.2 A.” for Cellulose Condition Detection The CO and CO, are generated from the breakdown of
cellulosic materials. The ratio of CO, to CO can be influenced
Another ANN is constructed with one output to test only the to some extent by leaks during sampling and testing as CO, is
cellulose condition. Output “1” means cellulose is involved in present in detectable quantities in the atmosphere. Under some
the faults while “0” represents no cellulose involved. conditions carbon oxides can be generate from the oil or
Although only CO and CO, contain the information of mineral oil containing paints. Therefore, the presence of other
cellulose condition, their ratios to the hydrocarbon gases are degradation products from the cellulosic materials such as
helpful information in cellulose condition detection. Together furanic compounds has been used as a confirmatory tool [9].
with CO and CO,, five hydrocarbon key gases are also used
as input features. As there is only one output, training focuses
on determining cellulose condition. Three topologies are 5. EXAMPLES OF DIAGNOSIS RESULTS
tested and compared. Twenty two sample sets are used in
training and testing. Two ANNs were constructed according to the above
evaluation. The weight matrices and biases were stored as

,
According to the results in Table(S), the two hidden layer files. Testing data were inputted from a data file, then the two
ANN is the optimal one and about 90.9% accuracy is obtained ANNs were used to detect the faults and ceAlulose condition.
for this ANN topology as shown in Figure 3. Ten new sample sets from different transformers were tested.
Nine of the cases were transformers where the results of the
Since the CO, to CO ratio is an indicator of the cellulose physical investigation were published. For case 6, the data is
condition, this ratio is used directly as one of the inputs in an average value for 200 transformers. The testing data and
addition to the actual ppm values of CO and CO, in the ANN the results are listed in Table(6) and Table(7) respectively.
training. It is found that the rate of convergence does
improve, while the overall diagnosis accuracy remains the The results between the actual inspection of the transformer
same for the samples tested. and ANN diagnosis match very well. There are two cases (2
and 4) where the ANN for major fault type diagnosis and the
Compared with one step ANN approaches in [2][3], the actual inspection disagree. These two cases where there is
training for the above two ANNs is fairly easy and the training arcing but the ANN indicates there is arcing and overheating.
convergence is reached without many iterations. It shows the The ANN can be trained with refined input data to distinguish
potential of this approach as large amounts of data becomes between these two as the arcing degradation products are the
available. same as the thermal degradation products but in different
relative composition. There are cases where there is
TabYe(5) Input: H,,CH,,C,H,,C,H,,C,H,; overheating which causes damage to the insulation which then
r 3 . Fault
1 : in;, ;;:;lel ~
results in electrical discharges. It is also quite possible to have
Training Training Error a condition which involves simultaneous overheating andl
Topology Iteration electrical discharges such as from an unintentional core.
7-30-20-20-1 0.008876 ground. However this does not appear to be true in these two
cases.
-
7-40-20- 1 0.008458
-
7-50-1 0.009502
Ten -Fold Cross-Validation: 90.9%
1840
Table(6) Testing Data(ppm)

* ND -- Not detected

Table(7) Testing Results (0 - 1 range represents the different degree and certainty of a particular diagnosis)

*(p)-------- Paper or other cellulosic materials involved

In case 3 there is a large quantity of C,H, 16,000 ppm. A than the fault. This is one of the difficulties when there is no
Quick evaluation may suggest that there is arcing in the oil as previous data to show which gases are most actively being
many people are sensitive to the appearance of C,H,. generated. This is similar in case 8 where it is difficult to
However, the relative composition of the gases indicates that determine without any specific previous data whether the
the problem is thermal in nature in agreement with the ANN carbon oxides generated is from the incipient-fault location or
output and as confirmed by the internal investigation. In this from normal aging of the cellulosic materials. In case 8 the
case the amounts of carbon oxides is not high, however the investigation revealed loose connections resulting in hot metal
ratio of CO, to CO is quite high indicating the possibility of and therefore it seems likely that the majority of the carbon
the involvement of cellulosic materials in the arc as confirmed oxides was not from the incipient-fault areas but rather from
by inspection of the transformer. the aging of the winding insulation.

For case 7, it is difficult to justiry the ANN evaluation of the

overheating of paper even though some was found on
inspection of the transformer. It would appear that the ANN is 6. DISCUSSIONS
picking up on the low CO, to CO ratio which is not always a
good indicator of overheating of cellulosic materials as The results of this study show that the two-ANN approach is
suggested earlier in this paper. It may be that some minimum promising for transformer fault diagnosis even with limited
quantity limits for these gases may be required. This can be sample data. Theoretically, the ANN can be trained to
implemented easily in the ANN computer code. represent any observable phenomenon if there are sufficient
data available. The more complex a relationship is, the more
A number of the cases for which the diagnosis indicate training data are needed. The transformer fault diagnosis
involvement of cellulosic materials requires careful review. In could be very complicated. For example, it is desirable to
case 2 there is an indication of overheating of paper both distinguish between faults of oil and cellulosic materials
during inspection of the transformer and by ANN. There is an (paper), different temperatures (low, medium, and high) for
appreciable amount of CO, present. However, this is not an overheating in oil, or low energy and high energy sustained
unusual amount and a significant portion is likely from the arcing. To deal with such a complicated diagnosis problem,
general aging of the transformer cellulosic insulation rather the available input data may not always be enough. It can be
 1841
a very effective way to construct different ANNs for different [7] X. Ding, Y. Liu, “ Predicting Power Transformer Resonant
pattern recognition to obtain the highest diagnosis accuracy Frequencies Using ANN”, NAPS, 1993.
for each pattern.
[8] Y. H. Pao, “Adaptive Pattern Recognition and Neural
The accuracy of the diagnosis can be improved with the Network”, Addison-Wesley publish company, Inc., New
increase of training data. Another very important part of York, NY, 1989.
dissolved gas-in-oil analysis is evaluation of trends -- the
increasing rates of gas generation. Once this information is [9] P. J. Griffin, and L. R. Lewand, “A Practical Guide for
available, the diagnosis can be made more reliable. Due to the Evaluating the Condition of Cellulosic Insulation in
limited trend data available, this work will be done at a later Transformers,” Doble Client Conference Proceedings, 1995,
stage. Sec 5 - 6.1.

BIOGRAPHIES
7. CONCLUSION
Yuwen Zhang is a Ph.D. student of Electrical Engineering at
A two-step neural-network classifier has been developed and Virginia Polytechnic Institute and State [Jniversity. He
carefully tested for transformer fault diagnosis using dissolved received both his B.S. and M.S. degrees from Shanghai
gas-in-oil analysis. Several feature types have been evaluated Jiaotong University, China, in 1985, 1988 respectively. His
and several neural-net topologies have been considered. current research interests include electrical device fault
Patterns with different identification sensitivity are detected detection and diagnosis.
by different ANNs to obtain the highest accuracy for each
pattern. The two-step approach makes ANN easier to train Xiaoaun Ding is an Associate Professor of Electrical
and more accurate in detecting faults. Engineering at Ho Hai University. He received his B.S. degree
in Electrical Engineering from South East University,
Nanjing, China in 1982, and M.S. degree in 1989 from Ho
ACI(N0WLEDGMENTS Hai University, Nanjing, China. He joined the faculty of the
Department of Electrical Engineering at Ho Hai University in
This work is partially support by a Grant from The National March 1982. His research interests include artificial neural
Science Foundation. network application to electric devices fault diagnosis,
transient stability and reliability in power systems.

REFERENCE Dr Yilu Liu is an Assistant Professor of Electrical Engineering

at Virginia Polytechnic Institute and State University. She
[l] C. E. Lin, J. M. Ling, and C. L. Huang, “An Expert received her BS degree from Xian Jiaotong University, China,
System for Transformer Fault Diagnosis Using, Dissolved in 1982, and MS, PhD degrees from The Ohio State
Gas Analysis Analysis,” IEEE Trans. on Power Delivery, University in 1985, 1989. Her current research interests we
Vo1.8, NO.l, Jan. 1993. power system transients, power quality, power system
equipment modeling and diagnoses. Dr. Liu is the recipient of
[2] X. Ding, E. Yao, Y. Liu, P. .I.Griffin, “ANN based the 1993 National Science Foundation Young Investigator
transformer fault diagnosis using gas-in-oil analysis”, 57th Award and the 1994 Presidential Faculty Fellow Award.
American Power Conference, April 1995, Chicago IL.
Paul J. Griffin received his B.S. degree at American
[3] S. K. Bhattacharyya, R. E. Smith, and T. A. Haskew, “A International College, and his M.S. degree at the University of
neural network approach to transformer fault diagnosis using Rhode Island. Mr. Griffin has been with Doble Engineering
dissolved gas analysis data”, NAPS, 1993. Company for 16 years and is currently the Manager of the
Materials Laboratory for electrical insulation. Mr. Griffin is a
[4] W. D. Halstead, “A thermodynamic Assessment of the member of ASTM Committee D-27; Subcommittee Chairman
Formation of Gaseous Hydrocarbons in Faulty Transformers”. Analytical Test; Section Chairman, Gases in Oil; U S .
J. Inst. Petroleum 59 (Sept.1973) 569 pp 239-241. Technical Advisor to IEC TC 10, Fluids for Electrotechnical
Applications; member of the IEEE Insulating Fluid
[ 5 ] P. J. Griffin, “Criteria for interpretation of data for Subcommittee of the Transformer Committees, and a member
dissolved gases in oil transformer (A Review)”, Electrical on several CIGRE Task Forces.
Insulating oils, STP998, American Society for Testing and
Materials, Philadelphia, 1988, pp 89-106.

[6] R. R. Rogers, “IEEE and IEC Codes to Interpret Incipient-

Faults in Transformers Using Gas in Oil Analysis”, IEEE
Trans. Electr. Insul, Vol. EI-13 No.5. October, 1978, pp 349-
353.