Proceedings of 2008 International Symposium on Electrical Insulating

Materials, September 7-11, 2008, Yokkaichi, Mie, Japan A-1

Condition Monitoring and Diagnosis of Power Transformers

E. Gockenbach*, H. Borsi
Leibniz Universität Hannover, Schering-Institut
Callinstr. 25 A, 30167 Hannover, Germany
* E-mail : [email protected]

Abstract: Power transformers are important determined mainly by two parameters, the faults should
components of all electric networks. They are usually be recognized in a time period, which allows to take
very expensive and therefore diagnosis and monitoring actions to avoid critical conditions, and the request to
systems will be valuable for preventing damages. A do some maintenance work should only take place if the
continuous control of the status of the transformer condition of the plant requires it.
condition is worthwhile and will support the appropriate
decision in case of a problem. The transformers are The advantages of a monitoring system, which should
geographically spread and with the aid of Internet, it is be part of the complete system and therefore very often
possible to collect appropriate information from these a component of the manufacturer service concept, in
transformers to a central node for diagnostic purposes combination with a reasonable diagnosis are:
under the supervision of high voltage engineering
x extension of the remaining life
experts. The aim of this paper is to show the main
problems related to power transformers and to review x improvement of overloading possibility
mitigation methods for the monitoring and diagnosis of x higher availability and service reliability
power transformers. x condition-based maintenance and repair
x prevention of loss and destruction
INTRODUCTION
Diagnosis contains interpretation of data to determine
Power transformers are important equipment in electric the current condition of a transformer based on the
energy networks. The majority of these devices have monitored data. A precise diagnosis is required because
been in service for many years under different a wrong diagnosis costs money in any case, if the
environmental, electrical, and mechanical conditions. diagnosis is to pessimistic the investment for repair
Fig. 1 shows as an example the service years of and/or replacement is actually not necessary. If the
transformers from a large German utility. diagnosis is to optimistic an unexpected failure causes
80 most probably high costs. Therefore numerous research
70
activities are ongoing to use all the available
information form the monitoring systems for a reliable
60
diagnosis. Advanced technologies may have the
potential to reduce time and to increase the reliability of
number of transformers

50

transformer diagnosis. The following techniques can be

40
used to get worthwhile information out the large amount
30 of monitored data:
20
x Expert Systems
10 x Case-Based Reasoning (CBR)
0 x Model-Based Reasoning (MBR)
x Artificial Neural Network (ANN)
x Fuzzy Logic
service age in years
x Knowledge-Based Systems
Fig. 1: Transformer service age within one utility [1]
x Genetic Algorithm
A CIGRE study shows clearly that associated with
equipment repair or replacement, the capital loss of an The experiences up to now have shown that more than
accidental power transformer outage is often counted in on technology has to be used in order to get reliable
million dollars for output loss only [2]. Because of this results, because the sensitivity of some parameters is
economic motivation, on-line monitoring and diagnosis different for the various methods.
systems are of benefit to predict fault conditions and
maintenance of the high voltage transformers. FAILURE CAUSES
In order to monitor the relevant parameters it is
The selection of the monitoring functions can be necessary to know the main failure causes within a

−16−
 1000
transformer. In the following the main failure causes of
the insulation system in a transformer should be stressed,

relative depolymerisation velocity

80 °C
even if other failures may also play a dominant role. 100 °C
The failures can be caused by the following stresses: 100 120 °C

x electrical
x electromagnetic
10
x dielectric
x thermal
x chemical
1
0 1 2 3 [%] 4
These failures may need a certain time until they will
water content in paper
lead to a breakdown in the insulation system. One of the
problems in monitoring and diagnosis is the fact, that in Fig. 3 Relative depolymerisation velocity at different water
content and temperature
the first years of service the failure rate is relatively low
and will not require a monitoring or diagnosis system.
Later the increase of the failure rate may be very strong,
but information from earlier times is necessary to
improve the diagnosis. Fig. 2 shows a general curve electrodes
representing the failure rate of transformers as function
of the service time.
60

50

40
failure rate in %

30
impregated paper
20

10

0 liquid stream
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49
service years
Fig. 4 Sensor for humidity [4]
Fig. 2 Example of failure rate as a function of transformer age [3]
If the system has reached its equilibrium a simple
measurement of the capacitance of the measuring
Electrical causes
electrodes and the conductivity of the liquid will give
information about the humidity content in the paper.
Inside the transformers, insulation systems (made of
insulating oil and cellulose), can be locally overstressed,
whereby the aging process of the insulating system will Electromagnetic causes
be accelerated, leading to acid generation and pollution
in the oil, an increase in the water content, gas and mud High forces can be caused by a short circuit and this
formation. As a consequence, the breakdown voltage will lead to a deformation of the windings within the
will decrease. The increase of the water content will transformer. In consequence the paper insulation may
also accelerate the depolymerisation of the cellulose also be influenced due to mechanical stress. This will be
which leads again to a reduction of the breakdown the initial point for partial discharges which lead after a
voltage, because the paper aging process is primarily certain time to a complete breakdown of the insulation
dependent on the water content and the temperature. system. The evaluation of the mechanical changes in the
Furthermore the existence of oxygen will also transformer is usually difficult, because there is no
remarkably increase the rate of depolymerisation by a direct access to the windings. The measurement of the
factor between 2.5 and 10. Fig. 3 shows the so called Transfer Function is a way to detect
depolymerisation activity as a function of water content mechanical deformation. Fig. 5 shows an example of a
in the paper and the temperature as parameter and it transfer function, measured on a 400 kV transformer.
should be noted that the y-axis has a logarithmic scale. The transfer function of the three phases shows some
differences, probably caused due to the non symmetrical
A sensor for detection of humidity in the insulation arrangement in the transformer vessel and some
system is shown in Fig. 4. The main components of the mechanical changes within the windings. It should be
sensor are the paper, the liquid stream and the noted that the measurement of the transfer function is a
measuring electrodes, whereby the paper should be very power tool bur needs some precaution concerning
same material as it used for the other insulation parts. the influence of the measuring circuit on the results.

−17−
 available mathematic tools. The well known method of
Dissolved Gas Analysis (DGA) [6] can have some more
reliable result using newly developed software [7].

Generally the use of software can improve the reliability

and provide facilities to analyze the test data. It can also
be used to extract information and knowledge that is not
available and not visible from the data directly using
advanced information technology methods such as Data
Mining. The advancement in artificial intelligent (AI)
modelling techniques has enabled power engineers and
researchers to develop useful artificial intelligent
software for diagnosis.
Fig. 5 Transfer function of a transformer
Expert systems will support the interpretation of the
Dielectric causes data. An expert system is a computer program that
performs a complex decision making task within a
The quality of the paper insulation in a transformer particular narrow problem domain that is normally done
depends on the polymerization degree (number of the by a human expert. It is based on the idea of taking the
glucose rings and chains) of the paper. During aging knowledge from a specialist and expressing it in an
process, the chains are up-smelled consisting of glucose appropriate representation to exploit the knowledge in
rings and develop water, gases (CO, CO2), groups of the same way as the human expert does and above all
carboxyl’s (organic acids). The ageing speed depends with the same result. The use of expert systems for
thereby on different parameters such as temperature, transformer diagnosis offers the potential of reducing
water content, oxygen content, number and kind of the overhead required by substations for the
temperature cycles and material properties. maintenance of transformers.

Thermal Causes In contrast to laboratory measurements, on-site

measurements are very often disturbed by noise due to
As a thermal cause, the losses result from the electromagnetic waves and severe environmental
demagnetization of the core (eddy current losses) and conditions. However on-site monitoring gives actual
the Ohm's resistance of the coil inevitably. Heating up data, which allows a very fast reaction and measures.
and cooling processes of the insulation with high Furthermore the data represent the real service
temperature gradient, represent additional load of the condition and can be used for further evaluation if they
insulation system which influences the life span of the are recorded over a certain time in order to detect a
transformer insulation. trend in the parameter development. Thus, slow changes
can be detected and a warning message or command to
Chemical Causes immediate disconnection can take place, if given limit
values are crossed [8].
Organic acids are developed as aging by-products of the
solid and liquid insulation, which attack in particular the Chemical Procedures
paper insulation and accelerates the aging. In addition
metals such as copper, iron, aluminium and zinc which With the help of chemical procedures, some failures, in
exist in each transformer are catalytic, and have an the transformer insulation can be determined. One of
additional accelerating effect on the aging processes of the standard investigations is the so called “gas in oil
the insulation. analysis (DGA)”, with which a sample of the insulating
liquid is taken from the transformer and examined. As
TOOLS FOR MONITORING AND DIAGNOSIS an evaluation criterion, in particular, the triangular
method after Duval is used. The disadvantage of these
Tools and methods for monitoring and diagnosis of high procedures is that only an integral evaluation is
voltage power transformers are actually available [5]. possible.
These are traditional diagnostic methods that have been
used for many years, where the results are worthwhile Further procedures are the furan analysis (HPLC) and
but in many cases not sufficient, new methods that the analysis of free gases collected in Buchholz relay.
range from methods that are starting to be used and While with the furan analysis, vital information on the
methods that are still at the research stage from which a quality of the solid insulation is obtained, the analysis of
higher quality of diagnosis will be expected. the collected free gases provides information about the
range of an available failure. The evaluation of the gas
Also the traditional methods can be improved by using rate will improve the analysis.

−18−
Electrical Procedures REFERENCES
[1] M. Stach: Betriebswirtschaftliche Gesichtspunkte im
The electrical procedure is mainly limited to partial Asset-Management im Zeitalter der Fusion, Micafil
discharge (PD) measurement because this method can Symposium, 20-21.März 2002, Stuttgart
be used for non destructive tests and diagnosis [9]. Here [2] CIGRE-WG 12-05: An international survey on
the transfer function can also be use to determine the failures in large power transformers in service, Electra
PD source within the transformer winding. A detailed No.88 (1983), pp. 21-48
description of the method can be found in [10].
[3] C. Bengtsson, J.-O. Persson, M. Svenson:
Dielectric Procedures Replacement and Refurbishment Strategies for
Transformer Populations, CIGRE 2001 Colloquium,
The dielectric methods are used as offline procedure Dublin SC 12.20
and permit a more exact view of the insulating system, [4] E. Gockenbach: A Diagnosis and Condition
if an extended statement about their condition is Monitoring System for the Insulation System of Power
required. The following procedures are mainly used: Transformers, CMD 2006, 2006, Changwon, Korea
x •Recovery Voltage Measurement (RVM) [5] M. Wang, A. J. Vandermaar: Review of condition
x •Polarization and Depolarization Current (PDC) assessment of power transformers in service, IEEE El.
x •Frequency Dielectric Response (FDS) Insul. Magazine, Vol. 18, No. 6, pp. 12-25
[6] Guide for the Sampling of Gases and Oil from
CONCLUSIONS Oil-Filled Electrical Equipment and for the Analysis of
Free and Dissolved Gases,” IEC Publ. 60567, 1992
x Monitoring and diagnosis systems for power
transformers are available. [7] A. Akbari, A. Setayeshmehr, H. Borsi, E.
x The most influencing parameters and the failure Gockenbach: A Software Implementation of the Duval
sources and causes are well known. Triangle Method, International Symposium on
x To monitor the various effects different tools should Electrical Insulation ISEI 2008, p. 124 -127
be used in parallel taking the single advantage of [8] Y. Han, Y.H. Song, Y.H.: Condition monitoring
each tool. techniques for electrical equipment-a literature survey,
x The procedures are mainly on chemical, electrical IEEE Transactions on Power Delivery, Volume: 18,
and dielectric measurements. Issue: 1, Pages:4 - 13, Jan 2003
x The tendency of the different parameters can be [9] A. Akbari, et al.: Transfer Function-Based Partial
recorded however the evaluation of the failure risk Discharge Localization in Power Transformers: A
and the estimation of the residual life time from the Feasibility Study, IEEE Electrical Insulation Magazine,
recorded parameters are not sufficient solved up to
Vol. 18, No. 5, 2002, p 22 – 32
now.
x Using more and more monitoring and diagnosis [10] E. Gockenbach, H. Borsi: Transfer Function as
systems on power transformers like Fig. 6 will Tool for Noise Suppression and Localization of Partial
increase the data base and improve the diagnosis. Discharges in Transformers during On-site
Measurements, ISH 2008, Beijing, China, paper K2-11

Data Acquisition Data Mining Diagnosis

Feature Extraction

PD, FRA, Bushing Clustering ANN

Pre-processing

Online
& Fuzzy
Voltages, Currents Discrimination
Offline
Sensors CBR
Information
Dissolved Gas in Oil
Fusion MBR

Fig. 6 Block diagram of monitoring and diagnosis system for power transformers

−19−