A NEW TECHNIQUE TO ESTIMATE FURAN CONCENTRATION

THROUGH SPECTRAL RESPONSE OF TRANSFORMER OIL

A. Abu-Siada and S. Islam
Electrical and Computer Engineering Department
Curtin University, Australia
ABSTRACT
Furans are the major degradation product of insulating paper in transformer oil. Hence the concentration
of furans in oil can be used as a good indicator of paper deterioration. Furan concentration in transformer
oil is currently measured using High-Performance Liquid Chromatography (HPLC) or Gas
Chromatography-Mass Spectrometry (GC/MS). Both methods provide accurate and reliable results in
detecting furan concentration. However, the two methods need very expensive equipment and take long
times to perform tests on oil samples which must be chemically treated before conducting the test.
Moreover, both methods require an expert to perform the test and to interpret its results. This paper
introduces a novel technique for detecting furan concentration in transformer oil through measuring oil
spectral response. The Ultraviolet-to-Visible (UV-Vis) spectral response of transformer oil can be
measured instantly with relatively inexpensive equipment and does not need an expert to conduct the
test. Results show that there is a good correlation between oil spectral response and its furan contents.
The effect of ambient temperature and conducting materials dissolved in transformer oil on its spectral
response is also investigated. A fuzzy logic model to estimate the relationship between furan
concentration and spectral response parameters of transformer oil is presented. The technique is proved
to be reliable, fast, cost effective, and can be implemented online.

INTRODUCTION
Power transformers represent a vital link in any transmission or distribution network. In the event that a
failure occurs in service, the impact can be far reaching; not only causing extended outages, but can
result in costly repairs and potentially serious injury or fatality. Therefore, it is essential to adopt reliable
condition monitoring and diagnostic techniques to decrease maintenance and improve reliability of the
equipment [1]. Since the entire energized and high temperature transformer components are immersed in
dielectric oil, transformer oil is a key source to detect incipient and fast developing faults, and oil quality
generally reflects the health condition of the transformer. Paper insulation used for electrical winding is
composed of approximately 90% of cellulose, 6-7% hemi-cellulose and 3-4% of lignin. Due to electrical
and thermal stresses that an operating transformer experiences, oil and cellulose decomposition occurs,
and gases will be evolved. As a result of this decomposition, the heat dissipation capability of the oil and
the mechanical strength of insulation paper will decrease [2]. When degradation of paper insulation
occurs, the cellulose molecular chains get shorter and chemical products such as furan derivatives are
produced and dissolve in the oil. Practical measurements show that the 2-furfural is the most prominent
component of furan derivatives [3, 4].
An increase in furan concentration in transformer oil corresponds to a decrease in the tensile strength,
and the degree of polymerization (DP) of the insulation paper in a transformer can be correlated with
paper DP as a function of furan level, and therefore an in-service assessment of the mechanical strength
of the paper insulation can be made. De Pablo reported the following relation between furfural and
Degree of polymerization based on viscosity (DPv) [5] :

DPV

7100
8.88 2 FAL

(1)

where 2FAL is the 2-furfural concentration in mg/kg of oil.

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved

Oil samples can be easily collected from operating transformers; however, it is impractical to acquire
paper samples from in-service transformers., so The concentration of 2-furfural dissolved in transformer
oil can be used as a good indicator of paper deterioration. It has been estimated that new paper, under
normal operating conditions, will generate furfural at the rate of 1.7 ng/g of paper per hour. The rate of
production increases with increasing degradation levels [6-8]. Furan concentrations are measured using
High-Performance Liquid Chromatography (HPLC) based on ASTM method D5837 or Gas
Chromatography-Mass Spectrometry (GC/MS) [9]. Both methods provide accurate and reliable results in
detecting the concentration of all furan derivatives: 2- furfural (2-FAL), 2-furfurol (2-FOL), 5-hydroxy
methyl-2-furfural (5-HMF), 5-methyl-2-furfural (5-MEF), and 2-acetylfuran (2-ACF). However, these two
methods need very expensive equipment and take long times to get the results for oil samples which
must go through a sample preparation step before conducting the test. Moreover, it requires an expert
person to perform the test and interpret its results. This paper presents a novel technique to determine
furan concentration in transformer oil through measuring its spectral response. The Ultraviolet-to-Visible
(UV-Vis) spectral response of transformer oil can be measured instantly with relatively inexpensive
equipment and does not need an expert person to conduct the test, and there is no sample preparation
required for the oil samples.

Laboratory aged oil

The study has been performed on in-service as well as laboratory-aged transformer oil. Laboratory-aged
insulating oil was prepared by utilizing the heating process available in IEC 61125 (International Electrotechnical Commission) [10]. Sections of new Kraft paper (20mm x 280mm) were cut and wrapped around
copper strips (3mm x 10mm) and then impregnated in 25ml of new transformer oil (Shell Diala B). All
samples were maintained at 100C in a thermostatically-controlled aluminum alloy block heater for 7
days. Oxygen flow at a rate of 1 l/hr was supplied into each dry tube to further accelerate the aging
process.

FURAN ANALYSIS
All samples were prepared in accordance to standard ASTM D5837 and tested using GC/MS system for
furan derivatives identification and quantification [9, 11]. Each sample was pre-treated with acetonitrile
prior to extraction and analyzed using GC/MS. Table 1 shows furan derivative concentrations in parts per
million (ppm) by weight for five oil samples using GC/MS. Results show that the 2-FAL is the most
prominent component of paper decomposition. Therefore, the level of 2-FAL in transformer oil can be
used as an indicator for paper deterioration.

Table 1
Furan Concentration Results using GC/MS

Test
Sample
New Oil
Sample
1
Sample
2
Sample
3
Sample
4

<0.01

2FOL
<0.01

25ACF MEF
<0.01 <0.01

5HMF
<0.01

3.1

<0.01

0.01

0.01

<0.01

5.1

<0.01

0.02

0.01

<0.01

10.0

<0.01

0.03

0.03

<0.01

15.0

0.01

0.05

0.05

<0.01

2-FAL

Oil samples at other 2-FAL concentrations were produced by diluting oil samples of known furan
concentration. Oil samples at different furan contents were prepared in a 10ml volumetric flask that was
placed on a scale at an ambient temperature of 20C. The required 2-FAL concentration was calculated
based on transformer oil density of 0.889g/ml at 20C as shown in the equations below:
st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
2 of 10

A B 8.89 g
C
A 8.89
D

(2)
(3)

Where A was the weight of original oil sample which has 2-FAL concentration of D ppm, B is the weight of
the new oil sample in grams and C is the required 2-FAL concentration to be produced.

Table 2
DP and 2-Furfuraldehyde (2-FAL) Correlation
2-FAL (ppm)
0-0.1
0.1-1.0
1-10
> 10

DP Value
1200-700
700-450
450-250
< 250

Significance
Healthy Insulation
Moderate Deterioration
Extensive Deterioration
End of Life Criteria

The correlation between 2-FAL and DP with respect to the health condition of insulation paper is given in
Table 2. When a DP test reveals a value of 250 or less, the paper is considered to have lost all its
mechanical strength, and the transformer has reached its end of life [12].

UV-VIS SPECTRAL RESPONSE

UV-Spectrophotometry is an accurate and sensitive method to analyze impurities in the transformer oil
using light absorbing properties of a sample. Light transmitted through the oil sample containing various
contaminants is decreased by that fraction being absorbed and is detected as a function of wavelength
[13]. A spectrophotometer measures the transmission, absorption, or reflection of the light spectrum for a
given wavelength. Absorption spectroscopy provides a measure of how much light is absorbed by the oil
sample and can be calculated as [5]:

S D

A log
R D

(4)

where A is the absorbance, S is the sample intensity, D is the dark intensity, and R is the reference
intensity at wavelength .

The same oil samples used for furan concentration measurement using GC/MS, along with the diluted
samples, were tested using a laboratory grade spectrophotometer for absorption spectroscopy at a
temperature of 20C. The experimental procedure was set up in reference to ASTM E275 [14]. Figure 1
shows the lab set up for measuring spectral response for one oil sample. The light source sends light at
different wavelengths via an input fiber into an oil sample which is placed in a cuvette located in a cuvette
holder. The light interacts with the oil sample and the output fiber carries light from the sample to the
spectrometer which is connected to a PC. Figure 2 shows the spectral response (absorbance) for
different oil samples with different furan concentrations. The encircled number on each plot in Figure 2
represents the 2-FAL concentration in parts per million (ppm).

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
3 of 10

Lab set up for UV-Vis Spectroscopic Measurement

Figure 1
It can be shown from Figure 2 that the new oil exhibits its characteristics in a bandwidth between 200 and
350nm with maximum absorbance at 225nm wavelength. However, in-service and laboratory aged oil
samples exhibit their respective characteristics in the range of 200nm and 470nm wavelengths. Results
show that absorbance as well as bandwidth increases by a significant and easily observable margin with
oil contamination which is indicated by the furan concentration level in the oil. The UV spectrum shows
considerable noise for contaminated oil at wavelength 350nm. This may also be attributed to the variety
of oil contaminants including very high carbon and water content.
Figure 2 reveals that the more furan concentration in the oil sample, the more absorbance and broader
wavelength span its spectral response will exhibit. The correlation between furan concentration in
transformer oil and its spectral response shown in Figure 2 indicates that the spectral response
parameters can be used as an alternative method to GC/MS to determine the furan concentration in
transformer oil.

UV/Vis Spectrum (Absorbance) for Different Oil Samples with

Different Furan Concentrations
Figure 2
Due to the fact that the spectral response of transformer oil depends on the instruments ambient
temperature [14], UV-Vis spectral response was conducted on the same samples at different ambient
st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
4 of 10

temperatures . The maximum variation in wavelength and maximum absorbance was found to be 3% as
can be seen in Figures 3 and 4.

Spectral Bandwidth Variation at Ambient Temperature of 20C and 25C for Different
Furan Concentrations
Figure 3

Maximum Absorption Variation at Ambient Temperature of 20C and 25C for Different
Furan Concentrations
Figure 4
The above two figures show that, for the same 2-FAL concentration, an ambient temperature incremental
change from 20C to 25C expands the bandwidth by generally 10nm, whereas absorption peaks for the
same temperature increments are overlapped at different furans concentration.
To investigate the impact of electrically conducting materials on the spectral responses of transformer oil,
transformer oil with specific furan concentration was scanned and analyzed by using UV-Vis
Spectroscopy. About 2.1 gram weight of each oil sample was placed into a cuvette for scanning . The
absorbance level and the bandwidth of the oil spectral response was recorded as references. 10mg of
copper powder was then added and dissolved into each oil sample which was rescanned and analyzed
immediately after the stirring process. The oil spectral response was observed and analyzed every 5
minutes for one hour. Each step was repeated with 20mg, 30mg, 40mg, 50mg, 60mg, 80mg and 100mg
of conducting material dissolved in the oil sample.

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
5 of 10

(a)

(b)

(c)

UV-Vis Spectral Response with Additional Conducting Material for a) instantaneous

effects (b) after 5 minutes, and (c) after 10 minutes
Figure 5
As shown in Figure 5(a), the incremental change in both the maximum absorbance and wavelength is
proportional to the amount of conducting material added to the oil sample. Figure 5(a) shows that the
st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
6 of 10

maximum absorbance level exceeds 2.5 after 100mg of copper powder was added to the oil sample.
After 5 minutes of adding the copper powder to the oil sample, the wavelength and maximum absorbance
for all oil samples was significantly reduced, and become close to the original oil sample as shown in
Figure 5(b). Five minutes later, spectral responses for all samples almost coincide with the spectral
response of the original oil sample result as shown in Figure 5(c). Theoretically, the conducting material
will settle to the bottom of the oil sample after a while, hence the spectral response is expected to have
similar characteristics of oil samples without any additional conducting material.

Fuzzy logic model

A fuzzy logic model was developed to estimate the correlation of furan contents in transformer oil and its
spectral response parameters. Input variables for the model are the values of bandwidth wavelength and
the maximum absorption for different oil samples collected from the test set up shown in Figure 1.
Variation of these parameters at different ambient temperatures as found in Figures 3 and 4 were
included in the model design. The output from the model then estimated the furan content in ppm. The
model was built using a graphical user interface tool provided by MATLAB. Each input was fuzzified into
five sets (normal to significant) of Gaussian combination membership functions (MF) governed by the
following equation:

f ; , c e

c 2
2 2

(5)

The function set developed for each input is a combination of and c parameters, where first function,
=1, specified by 1 and c1 determines the shape of the left-most curve; the second function, =2,
specified by 2 and c2 determines the shape of the right-most curve. The corresponding Gaussian
function curve for the input variable set is shown in Figure 6.

Input Variable MFs (Band Width and Maximum Absorbance)

Figure 6
Results of fuzzification from each input was then applied with fuzzy operator in the antecedent and
related to the consequence, by application method. Centre-of-gravity, which is widely used in fuzzy
models, was used for the defuzzification method where the desired output z0 is calculated as [15]:

z0

z. ( z)dz
( z)dz
c

(6)

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
7 of 10

where c(z) is the membership function of the output.

The membership functions for the output variables (furan) were considered on the scale from 0 to 15 ppm
(normal to significant criticality) and is shown in Figure 7.

Output Variable MF Estimated Furan Contents

Figure 7
The variation range in wavelength and maximum absorbance at different room temperatures and the
corresponding furan concentration was used to develop a set of fuzzy logic rules in the form of (IF-ANDTHEN) statements to relate the input variables to the output as shown in Figure 8. The fuzzy model to
estimate furan concentration for any set of input variables is shown in Figure 9 where BW and Ab
(spectral response bandwidth and peak absorbance) represent the input variables and t is the step time
for the fuzzy logic model simulation.

Fuzzy Rules - Spectrum-Furan Correlation Model

Figure 8

Furan Measurement using Spectral Response Fuzzy Model

Figure 9
st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
8 of 10

The fuzzy logic model was tested with various oil samples of known furan concentrations using GC/MS
and the model was able to estimate the furan concentrations with more than 95% accuracy compared to
measured results [16].

Comparison between ASTM D5837 and UV-Vis spectroscopy with fuzzy logic
ASTM D5837 is widely adopted for accurate identification and quantification of furan derivatives in
transformer oil. However, testing with HPLC or GC/MS requires trained personnel, tedious pre-treating of
test samples and very expensive equipment. By proposing UV-Vis spectroscopy and a fuzzy logic model,
utilities could use this alternative to estimate the overall furan concentration in the oil before it is found
necessary to be sent for a standard test. The comparison of both methods is summarized in Table 3.

TABLE 3
Comparison between ASTM D 5837 and UV-Vis Spectroscopy with Fuzzy Logic

Method

Test method

ASTM D5837

UV-Vis with fuzzy logic

-Requires trained
person to prepare the
sample and handle the
experiment.

-Doesnt require trained

person to conduct the
experiment.

Oil samples need to be

Oil samples do not have
pre-treated with
to be pre-treated with
chemical reagents.
any chemical reagent.
Time taken to
conduct the test
Price of Equipment
Accuracy

>4hrs

Instantly

>Base Price
Able to identify and
quantify the
concentration for every
single furan derivative.

< Approx. Base Price

Shows the response as
a whole. However, 99%
of furan concentration is
dominated by 2-furfural.

CONCLUSION
This paper introduces a new technique to detect furan concentrations in transformer oil through
measuring its spectral response which can be measured instantly using relatively inexpensive equipment
and does not call for an expert person to conduct the test. Results show that there is a good correlation
between furan concentrations in transformer oil and its spectral response parameters namely; bandwidth
and maximum absorption. A fuzzy logic approach is employed to estimate furan contents in transformer
oil samples using its spectral response parameters. The proposed fuzzy logic model provides excellent
estimation of the furan concentration in transformer oil using oil spectral response parameters. This
technique along with the expert system model can be used as a good alternative to the current technique
that is using HPLC or GC/MS to detect furan concentration in transformer oil. This technique is accurate,
cost effective, easy to implement and can be conducted onsite instantly without the need to send oil
samples to an external laboratory.

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
9 of 10

REFERENCES
[1] A. Abu-Siada and S. Islam, A Novel On-Line Technique to detect Power Transformer Winding
Faults, IEEE Transaction on Power Delivery, Vol. 27, No. 2, pp. 849-857, April 2012.
[2] A. M. Emsley, X. Xiao, R. J. Heywood, and M. Ali: 'Degradation of cellulosic insulation in power
transformers. Part 2: formation of furan products in insulating oil', IEE Proceedings on Science,
Measurement and Technology, 2000,147, pp. 110-114
[3] R. D. Stebbins, D. S. Myers, and A. B. Shkolnik: 'Furanic compounds in dielectric liquid samples:
review and update of diagnostic interpretation and estimation of insulation ageing', Proceedings of the
7th International Conference on Properties and Applications of Dielectric Materials, (3), June 2003, pp.
921-926
[4] A. J. Kachler and I. Hohlein: 'Aging of cellulose at transformer service temperatures. Part 1: Influence
of type of oil and air on the degree of polymerization of pressboard, dissolved gases, and furanic
compounds in oil', IEEE Electrical Insulation Magazine, 2005, (21), pp. 15-21
[5] M. Arshad: 'Remnant Life Estimation Model Using Fuzzy Logic for Power Transformer Asset
Management', PhD thesis, Curtin University of Technology, 2005
[6] Y. Shang, L. Yang, Z. J. Guo, and Z. Yan: 'Assessing aging of large transformers by furfural
investigation', Proceedings of the 2001 IEEE 7th International Conference on Solid Dielectrics, 2001,
pp. 272-274
[7] C. H. Zhang and J. M. K. Macalpine: 'Furfural Concentration in Transformer Oil as an Indicator of
Paper Ageing, Part 1: A Review', IEEE PES Power Systems Conference and Exposition, 2006, pp.
1088-1091
[8] J. M. K. MacAlpine and C. H. Zhang: 'Observations from measurements of the furfural content of oil
samples from transmission transformers', International Conference on Advances in Power System
Control, Operation and Management, 2000, (2), pp. 317-321
[9] ASTM: 'Standard Test Method for Furanic Compounds in Electrical Insulating Liquids by HighPerformance Liquid Chromatography (HPLC)', D5837-05, 2005.
[10] ITC: 'Unused Hydrocarbon-Based Insulating Liquids-Test Methods for Evaluating The Oxidation
Stability', IEC61125, 1992.
[11] http://www.gmu.edu/depts/SRIF/tutorial/gcd/gc-ms2.htm
[12] M. Wang, A. J. Vandermaar, and K. D. Srivastava: 'Review of condition assessment of power
transformers in service', IEEE Electrical Insulation Magazine, 2002, (18), pp. 12-25
[13] J. A. Palmer, W. Xianghui, R. A. Shoureshi, A. Mander, D. Torgerson, and C. Rich: 'Effect of aging
on the spectral response of transformer oil', IEEE International Symposium on Electrical Insulation,
2000, pp. 460-464.
[14] ASTM: 'Standard Practice for Describing and Measuring Performance of Ultraviolet, Visible, and
Near-Infrared Spectrophotometers', ASTM E275-01, ( 03.06), pp. 72-81, 2001.
[15] H. Li and M. M. Gupta: 'Fuzzy Logic and Intelligent Systems', International Series in Intelligent
Technologies, Kluwer Academic Publisher, 1995.
[16] A. Abu-Siada, S. P. Lai, and S. M. Islam, "A Novel Fuzzy-Logic Approach for Furan Estimation in
Transformer Oil," Power Delivery, IEEE Transactions on, vol. 27, pp. 469-474, 2012

st

2014 Doble Engineering Company 81 International Conference of Doble Clients

All Rights Reserved
10 of 10