Health Index Based Condition Monitoring of

Distribution Transformer
Gajanan C. Jaiswal Makarand Sudhakar Ballal
Additional Executive Engineer, M.S.E.D.C.L. Associate Professor
Research Scholar Department of Electrical Engineering
Department of Electrical Engineering Visvesvaraya National Institute of Technology, Nagpur
Shri Ramdeobaba College of Engineering & [email protected]
Management, Nagpur
[email protected]

D.R. Tutakne
Associate Professor
Department of Electrical Engineering
Shri Ramdeobaba College of Engineering & Management,
Nagpur, [email protected]

Abstract— Distribution transformer is equipment which plays and indirect cost of industrial and commercial electricity user.
very important role in the power distribution system. Because of, The proper management of such type of assets is discussed in
presence of the large number of transformers and various [1] and [2]. A diagnostics and data interpretation skill has used
components over a wide area in power systems, the condition to develop the decision making and assets management model
monitoring of health and automatic controlling of distribution with the use of fuzzy logic techniques. Cost effective
transformers are important issues. This paper presents fuzzy optimization skills, management activities, maintenance and
logic approach towards monitor as well as diagnose the condition combination of real time condition monitoring could be used
of distribution transformers, like load current, transformer oil for improvement in reliability of system is describe in [3].
and voltages. The suggested scheme will help to detect the
Health index technique is very important tool for
internal fault of distribution transformer and also diagnose the
faults with the help of defined range of parameters. On the
integrating all of the current statistics of transformer. Single
performance of these parameters health of distribution quantitative index which shows its total health is obtained with
transformer is observed and decisions taken for whether to keep the use of these health index techniques. These health index
distribution transformer in service or go for maintenance. techniques is used to show the long-term degradation level, to
determine the health condition which is not easily possible to
Keywords— Distribution transformer; fuzzy logic; condition determined by routine inspection and to measures the
monitoring system (CMS); health index. transformer aging such as winding resistance and furans
content. The position of any particular portion of a transformer
I. INTRODUCTION with respect to repair does not shown by health index.
In present era of competitive market, the main aim of Quantitative health index is described in [4] and [5].
power distribution utility is to provide the stable power Transformer health monitoring and providing the transformer
supply at a least cost, maximum profit and extension of life health index which shows its total health is different from the
span. Being a utility engineer, it is observed that failure occurs traditional health condition monitoring system. In traditional
in distribution transformer is due oil leakage, over loading and health condition monitoring techniques, interaction between
unbalance loading in remote rural area. To avoid such problem attributes have not considered.The total health status of the
proper condition monitoring of distribution transformer power transformer is shown with the help of health condition
parameters plays an important role. by considering the interaction between various health
Condition monitoring is the process or technique of condition monitoring tests for transformer are discussed in [6].
monitoring the operational characteristics of distribution The incipient faults are under the severe fault category
transformer. The condition monitoring of distribution that may result in damage to the transformer and simulation in
transformer can be done in such a way that, changes in the power system using Matlab/Simulink software for 100 MVA,
monitored operational characteristics can be used for 138/13.8 kV power transformer is discussed [7]. The classical
predictive maintenance in advance to avoid any serious DGA interpretation methods are used to examine the collected
deterioration or breakdown. Condition monitoring system can DGA results. These results are used for evaluating the
be used to estimate the distribution transformers health status exactness and consistency of each interpretation techniques.
from remote end. The failure of these types of assets, results in The results of investigation have been utilized for
creation of badly negative impact on the social life of development of fuzzy logic model is described in [8]. Analysis
electricity consumers. This also directly affects on the direct of distribution transformer with and without condition
monitoring in terms of financial loss to utility as well as to

978-1-4673-8888-7/16/$31.00 ©2016 IEEE

consumer and analysis of reasonable aspects of distribution of DT is implemented in C++ in Raspberry Pi2. The application
transformer with remote condition monitoring system is software developed consists of windows. In window,
discussed in [9]. In [10], the issue of monitoring and parameters values are shown. The window also shows the
evaluation of health status for distribution transformer with ON/OFF button and with this the distribution transformer can
remote condition techniques is described. be connected in circuit or disconnected from circuit.
Detection of an interturn fault at the initial stage in power The available test data are divided into three sets: voltages,
transformers with transformer interturn fault detection system load current and transformer oil level. Fuzzification of all data
(TIFDS) is discussed in [11]. In [12], transformer no-load and sets is done by using two types of membership functions. The
light load current harmonic analysis to detect the presence of output of fuzzification process is three linguistic variables for
interturn fault at the incipient stage is discussed. In [13] all data sets. For mapping these three linguistic input values to
probabilistic power system health index (PSHI) is discussed one linguistic output, twelve numbers of expert rules are used.
which describes power system health indices based on The health index of distribution transformer is represented as
adequacy and security. output. Membership functions utilized to show the distribution
Fuzzy logic approach is used in this paper to monitor the transformer health index (output) are trapezoidal and
distribution transformer performanance and for that, three triangular type. Defuzzification process is performed by using
numbers of input considered such as voltage, load current and the membership functions to find out crisp value for health
transformer oil level whereas one number of output considered index of distribution transformer. Functional block diagram
such as health index of distribution transformer which and experimental set-up of proposed system for calculating
indicates the performanance of distribution transformer and monitoring the health index of distribution transformer are
according to changes in input parameters. The prime objective given in Fig.1 (a) and (b).
of the research work proposed in this paper is for utilization of Development of the membership functions is on the basis
fuzzy logic techniques for evaluation of distribution of IEEE standards and selection based to achieve the desired
transformer health index on the basis of real time data. The output.
actual field data is utilized for determination of health index
using real measurements.
II. EXISTING METHOD USED BY UTILITY
Distribution transformer plays vital role for distribution of
electricity to end user, hence the health monitoring of these
transformer is very important to maintain reliability of supply.
The utility (Maharashtra State Electricity Distribution
Company Limited) measures parameter like oil leakages,
percentage loading, identification of low voltage pockets
(location), megger results, BDV results for calculation and
monitoring of health status of distribution transformer. The
health status of distribution transformer was improved by
proper tightening of bushing nut bolts and cleaning of dust on
distribution transformer bushing.

III. PROPOSED METHOD OF CALCULATING A

TRANSFORMER HEATH INDEX
A distribution transformer having capacity 400/400 V,
15 kVA is constructed for testing this system. Special
arrangements have been made for incorporating all sensors and
to change the internal conditions. All electrical parameters can
be sensed from energy meter data read out which is a serial port
using Rs 485. An HPL energy meter is used for this purpose.
Load current, voltage, power, power factor and other tampering
data are available in energy meter memory. It is interfaced with
SBC and communicated to the remote end. ADC’s are used for
converting the analog signal output of sensors to digital values.
Then serial communication using Rs 485 passes this data to
SBC when oil level moves up and down, ultrasonic signal varies
and thereby raspberry pi2 process the variation in oil level. The
output of the ultrasonic sensor that will be a voltage in
accordance with oil level and it can be interfaced with SBC after
passing through ADC. The ADC of MCP-3802 is used here for
performing the required operation. The ultrasonic sensor is used Fig. 1: (a) Functional block diagram and (b) experimental set-up of proposed
for observation of oil level variation in the distribution model.
transformer tank. The algorithm for health index determination
 An operational voltage level for all the distribution

Degree of membership
transformers under test are 11 kV/433V, the membership
functions generated in the paper are developed for these
voltage levels. The proper method for the selection of
membership function is used to achieve the desired output
results.
A. Membership Functions of Voltage
The input voltage variation to distribution transformer
affects the distribution transformer health, such as increase in
aging rate, dielectric strength reduction and bubbling problems
on high temperatures. The membership functions of input
voltage of distribution transformer were divided into three
linguistic terms: low, medium and high. Three membership
functions of trapezoidal type therefore used to model the
levels of voltage to obtain the accurate result, are shown in
Fig. 2 (a).

B. Membership Functions of Load Current

The increase in load current continuously, affects the life Fig. 2: (a) Membership functions of voltage and (b) Membership functions of
load current.
span of the transformer oil. Due to increase in heat inside the
transformer, the deterioration of the insulation increases with
increase in load current beyond limit of the transformer
capacity. Degree of membership
The membership functions of load current of distribution
transformer were divided into two linguistic terms: medium
and high. Two membership functions of trapezoidal type
therefore used to model the load current to obtain the accurate
result, as shown in Fig. 2 (b).
C. Membership Functions of Oil Level
Quality of the insulating oil is depends on the ability to
withstand voltage stresses. Partial discharges increases due to
Degree of membership

low oil strength and partial sparks results in increase of

transformer ageing. The transformer oil level is divided into
two linguistic terms: low and high. Two membership functions
of triangular and trapezoidal type therefore used to model the
transformer oil to obtain the accurate result, as shown in Fig. 3
(a).

D. Membership Functions of Output Health Index

The membership functions indicating the health index
Fig. 3: (a) Membership functions of oil level and (b) Membership functions of
(poor, fair and good) of the transformer according to the input health index.
condition are given in Fig. 3 (b). The health index of
transformer are divided into three linguistic variable: poor, fair E. Expert Rules
and good and three membership functions of trapezoidal type Twelve number of expert rules are constructed by using
is used to model the health index to achieve good results. The fuzzy logic model to find out the accurate results of
membership function for good condition covers health indices distribution transformer health index. First priority has given
from 75% and above, health indices for fair condition is from to the oil level due to its reliable results for evaluating the
25% - 80% and health indices for poor condition is from 0% - solid insulation health. The results of load current has given
35%. second priority, while input voltage is allocated lower priority.
The constructed expert rules to obtain desired output are
shown in Table I.
 TABLE II: HEALTH INDICES OBSERVED FROM UTILITY AND
TABLE I: FUZZY LOGIC RULES HEALTH INDICES CALCULATED FOR DISTRIBUTION
TRANSFORMER 1-15
Rule Rules
No. S. Year of Voltage ratio (kV) Health Health
1 If voltage is low and load current is medium No. installation ; Power rating Index Index
and oil level is low then health is poor. (kVA) calculated observed
(method from utility
2 If voltage is low and load current is high and oil
based on data in %
level is low then health is poor. fuzzy logic)
3 If voltage is low and load current is medium in %
and oil level is high then health is good. 1. 2000 11/0.433; 100 82.5 – G 81 – G
4 If voltage is low and load current is high and oil 2. 2001 11/0.433; 200 85.6 – G 84 – G
level is high then health is fair. 3. 2002 11/0.433;63 30.0 – P 29 – P
4. 2000 11/0.433;200 39.6 – F 80 – G
5 If voltage is medium and load current is
5. 2003 11/0.433;63 15.8 – P 22 – P
medium and oil level is low then health is poor.
6. 2005 11/0.433;200 89.2 – G 87 – G
6 If voltage is medium and load current is high 7. 2000 11/0.433;100 52.4 – F 53 – F
and oil level is low then health is poor. 8. 2001 11/0.433;63 45.0 – F 21 – P
7 If voltage is medium and load current is high 9. 2005 11/0.433;200 79.6 – G 81 – G
and oil level is high then health is fair. 10. 2000 11/0.433;100 16.4 – P 19 – P
8 If voltage is high and load current is medium 11. 2002 11/0.433;200 88.5 – G 74 – F
and oil level is low then health is poor. 12. 2003 11/0.433;100 38.2 – F 42 – F
13. 2006 11/0.433;200 53.3 – F 56 – F
9 If voltage is high and load current is high and
14. 2001 11/0.433;200 87.8 – G 84 – G
oil level is low then health is poor.
15. 2002 11/0.433;200 92.4 - G 93 – G
10 If voltage is high and load current is high and oil
level is high then health is fair.
The difference in the health condition of the distribution
11 If voltage is high and load current is medium
and oil level is high then health is good.
transformer should be due to result of loading pattern,
difference in maintenance schedule, variation in number of
12 If voltage is low and load current is medium
and oil level is low then health is poor.
faults, or problems in transformer oil cooling system. Those
distribution transformer are in poor condition are likely to fail.
The distribution transformers are in fair condition have a
IV. RESULT AND DISCUSSION great chances of failure but not as those in poor condition. The
replacement of these type distribution transformers should be
Fuzzy logic is the method to formulate the mapping of scheduled in near future. Distribution transformers having
three numbers of input values into the health index value. health index in good condition does not require any special
Mamdani max-min fuzzy interfacing method is used in this attention.
paper for converting the numerical values for voltage, load
current and oil level into linguistic expressions by using
corresponding sets of membership functions. The linguistic V. COMPARISON OF RESULTS WITH UTILITY
expression is substituted in the developed fuzzy rules to fire a The results obtained were compared with the results
subset of rules. Resulting membership function of output observed from utility data. The presented health index in the
(health index) for each rule fired at a minimum membership paper is on the basis of fuzzy-logic technique whereas the
value for all the antecedents are truncated by using mamdani utility health index is on the basis of a crisp number. A
max-min method. Finally total membership function has been variation in health index is clearly shown in Table II. As per
obtained by considering the union of all truncated output Table II the interpreted linguistic expressions of both indices
membership functions of the fired rules. Centroid method is are similar, but the crisp numbers of health indices are not
used for defuzzification. The centroid method is used to find perfectly matched. The implementation of the fuzzy logic
out the crisp value for output. method to calculate the health index, required creation of a
The input data, data of health indices observed from different number of health condition categories like good, fair,
utility and the data of calculated health indices of 15 number poor and more. The categories of health index are good, fair
sample distribution transformers from the 29 number of and poor as per the method used by utility. In order to make
working distribution transformer are shown in Table II. On the comparison easy, the health index conditions good (G), fair
basis of Fig. 2 (a), Fig. 2 (b), Fig. 3 (a) and Fig. 3 (b), the (F) and poor (P) provided by the fuzzy-logic decisions are
corresponding conditions of health index for three input values compared with the good, fair and poor health index observed
with a degree of membership have developed and rule number from utility data. The results of methods presented and the
3, 7 and 9 are fired to show the health condition. A total 7 of data observed from utility are 80% matched.
15 number of distribution transformers were installed between
the year 2000 and 2001. As per the samples of distribution
transformers taken, 7 are in good health condition, 4 are in fair
VI. CONCLUSION
health condition and the remaining 4 are in poor health
condition. Health index based condition monitoring of distribution
transformer with the use of fuzzy set theory is presented in the
research work. Distribution transformer health dependent [13] J. Lim, J. Choi, H. Choi, J. Joo and J. Cha, “Expert criteria based
parameter like voltage, load current and oil level are probabilistic power system health index model and visualization,”
Probabilistic Methods Applied to Power Systems (PMAPS), 2014
considered as inputs for fuzzy logic model for calculation of International Conference on, Durham, 2014, pp. 1-6.
distribution transformer health index. Twelve numbers of
expert rules and two types of membership functions are
utilized for mapping the three input values into output
decisions to produced accurate health index of transformer.
The method used in this research work is tested by using field
data for 15 number of sample working distribution
transformers. Results obtained were compared with the
observed results from utility data. The comparison of these
result shows that 80% of the results are matched. The method
used by utility is based on observation of field data while the
method mentioned in this paper is based on membership
functions and hence more fruitful. The rules used for the
calculation are easy and clear for utilization to the user. As
compared with the method used by utility for data observation,
the presented method can be used directly by utilities and
industries in order to find out the health indices of distribution
transformers, which is significant advantage.

REFERENCES
[1] M. Wang, A. J. Vandermaar, and K. D. Srivastava, “Review of condition
assessment of power transformers in service,” IEEE Elect. Insul. Mag.,
vol. 18, pp. 12-25, Nov./Dec. 2002.
[2] A. E. B. Abu-Elanien and M. M. A. Salama, “Asset management
techniques for transformers,” Elect. Power Syst. Res. J., vol. 80, pp.
456-464, April 2010.
[3] Muhammad Arshad, Syed M. Islam, A. Khaliq, “Fuzzy logic approach
in power transformers management and decision making,” IEEE
Transcations on Dielectrics and Electrical Insulation, vol. 21, pp. 2343-
2354, 2014.
[4] A. Jahromi, R, Piercy, S. Cress, J. Service and W. Fan, “An approach to
power transformer asset management using health index,” IEEE Elect.
Insul. Mag., vol. 25, pp. 20-34, Mar./Apr. 2009.
[5] A. Naderian, S. Cress, R. Piercy, F. Wang, and J. Service, “An approach
to determine the health index of power transformers,” in Proc. Conf.
Rec. IEEE Int. Symp. Elect. Insul., pp. 192-196, Jun. 2008.
[6] J. Singh, Y. R. Sood, and R. K. Jarial, “Condition monitoring of power
transformers-bibliography survey,” IEEE Elect. Insul. Mag., vol. 24, pp.
11-25, May/Jun. 2008.
[7] K. Ramesh and M. Sushama, “Inter-turn fault detection in power
transformer using fuzzy logic,” Science Engineering and Management
Research(ICSEMR), 2014, International conference, pp. 1-5, 2014.
[8] A. Abu-Siada, s. Hmood and S. Islam, “A new fuzzy logic approach for
consistent interpretation of dissolved gas-in-oil analysis,” IEEE
Transcations on Dielectrics and Electrical Insulation, vol. 20, pp. 2343-
2349, 2013.
[9] A. Avinash Nelson, Gajanan C. Jaiswal, Makarand S. Ballal,
“Economical aspects of remote condition monitoring system for
distribution transformer,” Power, Automation and Communication
(INPAC), 2014 International conference, pp. 45-49, Oct. 2014.
[10] A. Avinash Nelson, Gajanan C. Jaiswal, Makarand S. Ballal, D.R.
Tutakne,“Remote condition monitoring system for distribution
transformer,” 2014 Power System Conference (NPSC), pp. 1-5, Dec.
2014.
[11] M. S. Ballal, H. M. Suryawanshi, M. K. Mishra and B. N. Chaudhari,
“Interturn Faults Detection of Transformers by Diagnosis of Neutral
Current”, IEEE Transactions on Power Delivery, vol. 31, no. 3, pp.
1096-1105, June 2016.
[12] P. A. Venikar, M. S. Ballal, B. S. Umre, H. M. Suryawanshi, “A Novel
Off-line to On-line Approach to detect Transformer Inter-Turn Fault”,
IEEE Transactions on Power Delivery, vol. 31, no. 2, pp. 482-492, April
2016.