This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

Investigation of Distributed Moisture and

Temperature Measurements in Transformers using
Fiber Optics Sensors
Muhammad A. Ansari, Student Member, IEEE, D. Martin, Member, IEEE, and T. K. Saha, Fellow, IEEE

Abstract--The measurement of water inside a transformer is FBG_M fiber bragg grating wavelength of moisture sensor (after
essential to maintain its reliable operation, as the water content (T corr) removing temperature effect)
FBG_T temperature measurement using fiber bragg grating
can increase as the transformer ages. This measurement is
sensors
frequently performed by a utility as part of its periodic FBG_wcp water content of paper estimated using fiber bragg
maintenance activities. In this paper, the application of optical grating sensors (% by mass of dry paper)
fiber sensors as an online method to measure water content of Vaisala_T temperature measured by Vaisala probe
paper inside an operating transformer is presented. There are Vaisala_wcp water content of water measured by Vaisala probe
several advantages of using these sensors, which includes their
size, lightweight, flexibility, multiplexing, no-electromagnetic I. INTRODUCTION
interference etc. A unique benefit of fiber optic technology is
that sensors can be embedded into the winding structure and
operate in areas of high electric field strength. For this work,
fiber optic water and temperature sensors are investigated by
T he presence of water in transformer paper insulation is a
concern for an asset owner as it degrades the equipment
health in numerous ways, such as: decreased life expectancy
comparing their measurements to a commercial water activity of paper insulation, decreased dielectric strength of oil,
probe immersed in oil. The study found that the determination increased partial discharge, and the possibility of failure at
of water content by fiber optic sensors compared closely with
other techniques, which leads to the conclusion that they can be high load caused by bubble evolution [1]–[5]. The
an effective technology for water measurement in an measurement of moisture in transformer insulation is gaining
operational transformer. importance and is becoming beneficial for asset owners as
transformers in many countries are reaching or surpassing
Index Terms—fiber optics, moisture, bragg gratings, their useful life [6], where the moisture management is very
condition monitoring, water activity, water content of paper, important as old transformers may have wet insulation. It is
transformers, oil, paper, sensor, insulation
therefore of great importance for a utility operator to have
the correct measurement of water when operating the
NOMENCLATURE
equipment during normal and overload conditions.
𝜆𝐵 bragg wavelength (nm)
Karl Fischer titration (KFT) is a traditional technique to
𝜂𝑒𝑓𝑓 effective reflective index
𝛬 grating period (nm) measure the water content of solid transformer insulation.
𝑝 vapour pressure of water in oil (Pa) However, it requires direct access to transformer winding to
𝑝𝑜 vapour pressure of pure water at same temperature (Pa) collect the paper samples which is rarely possible. Another
𝑃𝑒 photo-elastic constant of fiber (Pa-1)
𝜀 strain induced on the fiber
difficulty is preventing changes in the water content of the
α thermal expansion coefficient of the fiber (°C−1) sample on its way to the test laboratory as it may adsorb
ζ thermo-optic coefficient of the fiber (°C−1) water from the atmosphere. Therefore, several indirect
𝑆(𝑇) solubility of water in oil at temperature “T” (ppm mass techniques to estimate transformer moisture have been
of water/mass of oil)
𝐴, 𝐵 solubility coefficients developed over the last few decades.
𝑆𝑎𝑤 water activity sensitivity (pm) The moisture estimation techniques can be broadly
𝑆𝑇 temperature sensitivity (pm/°C) categorized as oil chemical analysis, indirectly by dielectric
𝑆𝑟ℎ sensitivity of grating relating to moisture
𝑆𝑡 sensitivity of grating relating to temperature (°C−1) response measurements on transformers, online water
𝑃𝑣 vapour pressure of water in oil (Pa) activity measurements in oil and by optical techniques [7].
𝑎𝑤 water activity (p.u.) The chemical and dielectric measurements have been well
𝑊𝐶𝑂 water content of oil (ppm by mass of dry oil)
𝑊𝐶𝑃 water content of paper (% by mass of dry paper)
established [8], [9]. Dielectric response measurements have
𝐸𝑅𝐻 equilibrium relative humidity (%) the disadvantage that these tests are performed offline [10]
𝑅𝑆 relative saturation (%) where the transformer must be disconnected from the
FBG_M fiber bragg grating wavelength of moisture sensor network, and transformer operators prefer to minimize
(response includes temperature and moisture effect)
outages. Online water activity probes are available, however,
The authors are with the School of Information Technology and
algorithms are required to calculate the water content of
Electrical Engineering, The University of Queensland, St Lucia, QLD, paper from an oil measurement. Fiber optic based sensors
Australia, 4072. (Email: [email protected], [email protected] and have the advantage that they are both online and immune to
[email protected]) high electromagnetic fields, so they can be integrated into the

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

winding structure of a transformer. These sensors, however, transformer parameters e.g. temperature measurement [24]–
are in the research and development stage for power system [26], transformer oil ageing [27], insulating oil breakdown
applications, although they have been successfully used in [28], hydrogen detection [29], [30], clamping and vibration
various engineering and medical applications such as force [31] etc. Only a few limited studies for moisture
structural, aerospace, medical, soil etc. [11]. measurement in oil only [32]–[34] have been done. We also
One common and traditional practice of determining the performed some initial testing of the bragg grating sensors in
water content of solid insulation is based on using adsorption air, mineral and Envirotemp FR3 insulating fluids to measure
isotherms along with the water content of oil [12]. The oil the optical sensors response and their potential [35], [36].
sample is taken from the sampling valve and the transformer The measurements were performed in a small test vessel at
temperature is recorded to correctly use the adsorption different temperatures and moisture level using different
isotherm. However, this method introduces several saturated salt solutions and demineralized water. However,
uncertainties in the results due to sample handling and in an operating transformer, the environment is entirely
moisture dynamics. Ward et al presented that uncertainties different with the paper, oil and several other materials.
can be as high as 200% [13]. Two example reasons are that Most importantly, the temperature and moisture rarely
oil-paper curves are only valid for equilibrium conditions reaches an equilibrium state, hence it is essential to measure
[14], [15], and, that adsorption properties of oil-paper the dynamic response of sensors in an oil-paper insulation
insulation system change with age. Due to the limitation of system. Therefore, these studies are limited in their scope
this method, time and frequency domain dielectric response and do not conclude any concrete findings on their
techniques were introduced. The dielectric measurements are application for electrical insulation in an operating
performed offline and take on average 6~8 hours to environment.
complete, making it difficult for utilities to perform these In this article, an investigation on using polymer-coated
tests regularly. Moreover, these measurements are highly fiber bragg grating sensors to measure the water content of
dependent on the temperature of the insulation, which varies solid insulation is presented. The measurements were
during the measurement period and has a gradient along the performed using distribution voltage class transformers
transformer insulation. insulated with mineral and vegetable (Envirotemp FR3) oils.
The introduction of water activity probes in the 1990s [4], The transformers were subjected to cyclic temperature
[16] based on capacitance measurement, led to a real-time loading conditions realistic of normal operation to see the
moisture measurement of transformer insulating oil. The dynamic response of optical sensors, and then the water
probes measure the water activity of insulating oil based on content determined using the fiber optic sensor was
the vapor pressure of water dissolved in insulating oil. compared to other measurements. The results are compared
Several mathematical models are developed that estimates to a commercial water activity probe, and in general, found
paper moisture level based on the water activity level such as to be in close agreement with that of a water activity probe
Fessler [17] and Piper [18]. The key advantage of a water (Vaisala MMT330).
activity probe is its real-time and continuous monitoring
application. The moisture trend over the whole asset life can II. TRANSFORMER WATER CONTENT
be observed, which will reflect the true moisture status
A. Water Activity
instead of a single dielectric or chemical measurement. The
concern of non-equilibrium condition can be sorted out by The water content of an insulating medium can be
taking a long-term average of the probe output [19] or using quantified by its water activity level. Water activity is the
digital filtering [20]. However, installing the probe at a ratio of the vapor pressure of water dissolved in a fluid to that
correct location is important and it should be in the path of of pure water at the same temperature in equilibrium
oil flow [21]. A temperature compensation [22] is required condition (1). When water activity of air space is measured
for estimation of water content of paper to be valid, as water it is also called the equilibrium relative humidity [7].
𝑝
activity probe cannot be installed close to a high voltage 𝑎𝑤 = (1)
𝑝𝑜 (𝑇)
winding due to the risk of compromising the insulation and
causing dielectric breakdown. B. Vapour Pressure of water
Because of the limitations of current techniques, and need The water vapor pressure of a fluid can be calculated
to cope with future challenges, it is important to explore new directly from its measured water activity level using (1) and
technologies that can be utilized to measure the water content (2). Equation (2) was proposed by Buck [37] to calculate the
of paper insulation. Fiber optic based sensors have been used saturation vapor pressure of water.
over the years to measure winding temperature, and thus a 17.502×𝑇
similar technology to measure insulation water content 𝑝𝑜 (𝑇) = 6.1121 × 𝑒 (240.97+𝑇 ) (2)
would be highly beneficial. Their miniature size, lightweight, C. Water Content of Transformer Insulation Paper
flexibility, immunity to electromagnetic interference and Two models proposed by Fessler [17] and Piper [18] are
ruggedness make them an ideal sensor device [23]. Some commonly used in industry. Both models are based on the
researchers have demonstrated the application of polymer vapor pressure of water in oil and corresponding
coated bragg grating based optical sensors for various

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

temperature. The Fessler’s model (which also takes into Berruti is found to be suitable and hence used in our work.
account Piper) was used in this work to estimate the moisture The bragg grating is coated with a suitable polymer
distribution in the distribution transformer, which is given by material to absorb surrounding moisture and are then
equation (3) as: characterized with respect to moisture and temperature level.
4725.6
𝑊𝐶𝑃 = 2.173 × 10−7 × 𝑝0.6685 × 𝑒 𝑇 (3) Various hygroscopic materials are used for increasing
sensitivity such as polymethyl methacrylate (PMMA) [33],
III. FIBER BRAGG GRATING BASED MOISTURE polyvinyl acrylate (PVA) [34] or polyimide [41]. A brief
MEASUREMENTS description of various humidity sensing material can be
found in [45]. In this work, polyimide coated sensors are
A. Working Principle used due to their effective hygroscopic properties. The
The fiber bragg grating sensor operates analogous to an polyimide absorbs moisture and puts pressure on the bragg
optical filter, where most wavelengths of light pass through gratings only. The optical fiber sensors used has a polyimide
freely while certain wavelengths are reflected, which is the coating of around 30µm thickness. Fig. 1 presents the typical
bragg wavelength. The working principle is illustrated in Fig. dimensions of a sensor used in this study.
1. According to the coupling mode theory [38], the reflected
bragg grating wavelength is given by:
𝜆𝐵 = 2𝜂𝑒𝑓𝑓 𝛬 (4)
The 𝜂𝑒𝑓𝑓 depends on the optical fiber core material and the
grating period 𝛬 represents the sensing element. The grating
period varies with the change of strain.
The change in bragg wavelength is due to the stress on the
sensor and its temperature where their relationship is fairly
linear [11], [39]. Yeo [40], [41], Kronenberg [42] and Berruti Fig. 1. Fiber bragg grating sensing and dimensions
et al [43], [44] proposed various relationships for the bragg
B. Measurement System and Sensor Specifications
wavelength shift due to temperature and strain variations as
follows respectively: 1. Measurement Setup
𝛥𝜆𝐵
= (1 − 𝑃𝑒 )𝜀 + [(1 − 𝑃𝑒 )𝛼 + 𝜁]Δ𝑇 (5)
𝜆𝐵
𝛥𝜆𝐵
= 𝑆𝑟ℎ 𝛥𝑎𝑤 + 𝑆𝑡 𝛥𝑇 (6)
𝜆𝐵
𝛥𝜆𝐵 = 𝑆𝑎𝑤 𝛥𝑎𝑤 + 𝑆𝑇 𝛥𝑇 (7)
Yeo presented a relationship that depends on the material
properties, strain and temperature variations around the
sensor while Kronenberg reported a linear relationship with
air humidity and temperature with the per unit wavelength
change. However, Berruti suggested a direct relationship
between the moisture, temperature and bragg wavelength as
given in (7). As the bragg wavelength shift was found to be
linearly dependent on moisture and temperature surrounding
the sensors, therefore the direct relationship proposed by
Fig. 2. Fiber bragg measurement system

Capacitive Probe
Fibre Optics
Moisture and
Temperature
Sensors

Fibre Bragg
Gratings
High Voltage and
Low Voltage
Bushings
Active
Insulating Oil Part

Fig. 1 . Fig. 2.

Fig. 3. (Left) Schematic diagram showing the location of sensors, (Right) showing the actual transformer top with sensors connected

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

A complete measurement system is required for the fiber (SAR). The temperature class of the transformer paper
optics system, which includes a light generation and insulation is “A”. The insulating fluids used were mineral oil
detection device, known as the data acquisition unit. The and vegetable oil. The nameplate data of these units is given
smartfibres interrogator is used for FBG sensors, which is in Table II.
shown in Fig. 2. The interrogator unit generates a light signal TABLE II
with a wavelength in the range of (1528-1568 nm), and has Test Transformers Nameplate Data
4 optical channels, and each channel can measure signals Parameters Value
from 16 sensors simultaneously [46]. The interrogator unit is Rating (kVA) 5
then connected to a LabVIEW based software for data LV @ No load (V) 230
LV rated current (A) 21.74
acquisition and further processing. HV @ No load (kV) 2
HV rated current (A) 2.5
2. Sensor Specifications and Dimensions Frequency (Hz) 50
For this work, FBG based moisture and temperature Phase Single
Top Oil / Winding rise (K) 60 / 65
sensor array were fitted into two 5 kVA transformers by Oil Volume (l) 63
wrapping them around and through the winding structure in
oil ducts. Since the moisture sensing FBG requires its
In addition to the fiber optics sensors, each transformer
temperature to be known, the temperature sensors were
was equipped with a water activity probe (MMT-330)
installed as close to each other as possible (within few mm).
immersed in the oil volume. The probe was fitted through the
The mineral oil filled transformer was fitted with ten
top cover of transformers. The test transformers have an
temperature and ten moisture sensors, while the vegetable oil
additional controllable heater at the bottom to elevate the
(Envirotemp FR3) filled transformer was fitted with ten
insulation temperature to a level even higher than what
temperature sensors and two moisture sensors. The
would be reached during the rated load operation.
schematic arrangement of different sensors is shown in Fig.
3. The sensors bragg grating wavelengths are summarized in 2. Transformer Connections
Table I. Two transformers were connected in series via their high
TABLE I voltage side. One transformer’s low voltage side is
Transformers Optical Sensors Bragg Grating Wavelength connected to a supply voltage of 230V, and the other is
Bragg wavelength (nm) connected to a resistive load. The connection diagram of the
Sensors # Mineral Oil Vegetable Oil transformers is shown in Fig. 4.
Temperature Moisture Temperature Moisture 2 kV 230 V
TF 1 TF 2
FBG 1 1559.9898 1560.0 1559.9257 1550.0 Mains Supply
FBG 2 1557.3341 1557.5 1557.3567 1545.0 230 V
FBG 3 1554.8310 1555.0 1554.9175
FBG 4 1552.3068 1552.5 1552.2949
Load
FBG 5 1549.8333 1550.0 1549.8070
FBG 6 1547.4333 1547.5 1547.3853
FBG 7 1544.7852 1545.0 1544.7654 Fig. 4. Test transformers connection
FBG 8 1542.3549 1542.5 1542.3249
FBG 9 1539.7761 1540.0 1539.7927 B. Moisture Dynamic Experiment
FBG 10 1537.3044 1537.5 1537.1692 For this experiment, the test transformers were subjected
All the sensors, in both transformers, were installed in the to constant electrical and cyclic thermal loading. Moisture
oil ducts and a heater was enabled during the experiment. dynamics in the transformers is assessed using fiber optic
Therefore, no large variations in the moisture measurement sensors and water activity probe. Normally the transformers
were possible with either using 10 or 2 sensors. Therefore, undergo daily, weekly and seasonal cyclic loading pattern.
having 2 sensors in the Envirotemp FR3 transformers did not Therefore, a sinusoidal thermal loading to simulate actual
affect the measurements and is further discussed in Section transformer loading was applied with heaters and a constant
V. resistive load of 26 Amperes was connected with the
transformer. The experiment was repeated three times with
IV. EXPERIMENTS AND MEASUREMENTS the following sequence to see the repeatability and
consistency of sensor output.
A. Transformer Connections
1. Initially, transformers were kept at a constant
1. Transformers Data temperature of 50⁰C for a week to reach a thermal
The transformers used in this study are specially designed equilibrium condition and assess its initial condition
sealed distribution transformer having a 5 kVA rating. They before cyclic loading was started.
were designed to have a similar oil/paper insulation ratio as 2. Transformers were then subjected to a sinusoidal
that of a power transformer. The insulation material of temperature loading for a week. The cyclic period was
prototype test transformers includes diamond dotted paper 24 hrs to simulate daily loading pattern of field
(DDP), moldable pressboards (MPB) and spacers all round transformer. The loading pattern was controlled by a

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

PLC setup for sinusoidal temperature variation, observed to be around 3⁰C. Fig. 5 (Bottom) shows the
which was varied from 30⁰C to 80⁰C. temperature measurements by fiber optics sensors and water
3. After sinusoidal loading, the transformers were then activity probe. The temperature gradient (from top of the
kept at a constant temperature of 50⁰C for a week to winding to the bottom) for mineral oil filled transformer was
attain moisture equilibrium for estimating its moisture noted to be only around 2.5⁰C. The thermal gradient in the
condition. test transformer is low as the active part of the transformer is
only 265mm in height. However, it can go up to 20⁰C in large
4. Step # 2 and Step # 3 were repeated twice to see the
power transformers and higher the temperature variations,
repeatability and reproducibility of sensors output.
more thermal driven moisture gradient can be expected.
C. Mineral Oil Filled Transformer
2. Moisture Measurements:
1. Temperature Measurements: The moisture distribution in a transformer is not only a
Ten FBG sensors were installed in the winding structure function of temperature, but also depends upon the location
very close to the moisture FBG sensors. The temperature and moisture gradient in a large power transformer insulation
measured by the FBGs were compared to those from the body. One key advantage of a distributed FO moisture
thermocouples and water activity probe. The different sensing array is that it will monitor the water content at
temperature sensors installed were: multiple points. The moisture readings from this array will
i. Thermocouples measuring temperatures axially be compared with other measurements to ascertain
on top, middle and bottom of the unit. They were effectiveness and any moisture gradients within the system,
installed on the body of the transformer. which is shown in Fig. 6. In general, the optical sensors
ii. Water activity probe measuring the temperature showed very encouraging results and have a characteristic
of top oil. similar to the water activity probe. The variation in the
iii. FBG sensors measuring the temperature along moisture measurement is likely to have been caused by their
active part of units. different locations and corresponding temperatures.

Fig. 6. Mineral oil transformer moisture measurement by optical and water

activity probe

Fig. 5. Mineral oil transformer thermocouples (Top), fiber optics and water
activity probe temperature measurements (Bottom)
Fig. 7. Water content of paper estimated by moisture sensor # 1 in Mineral
The temperature measurements provided by all the sensors oil filled transformer
are consistent with each other. Fig. 5 (Top) provides the
temperature measurements using thermocouples. The For ease of understanding only the outcome of two FBG
temperature difference between the top and bottom unit was moisture sensors, 1 & 5 along with their corresponding

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

temperature, is presented in Fig. 7 and Fig. 8. The change in before the start of the testing and at the end of the experiment.
the water level measured by FBG sensors is evident from the It was found that the average distributed moisture content
change in its temperature by both sensors. When the measured by the optical sensors is a little higher than that
temperature increases, the oil moisture solubility increases measured by water activity probe, however, the Karl Fischer
and therefore its water activity level fall [22], [47]. Both the titration shows a lower moisture content in case of mineral
FBGs provided a similar output as that of the commercial oil filled transformer.
water activity probe.
D. Vegetable Oil (Envirotemp FR3) Filled Transformer
1. Temperature Measurements:
Similar to the mineral oil transformer, ten temperature
sensors were installed in vegetable oil filled transformer. The
same cyclic thermal loading was applied to the vegetable oil
transformer and response from different temperature sensors
is shown in Fig. 10. The maximum temperature variations
between the different optical temperature sensors were found
to be around 5⁰C, while the top and bottom thermocouples
temperature variation was 6⁰C. All the sensors response
showed similar variations and can be observed in Fig. 10.

Fig. 8. Water content of paper estimated by moisture sensor # 5 in Mineral

oil filled transformer

As it is discussed in Section III about the working and

sensitivity of fiber bragg grating on temperature and
moisture. In order to demonstrate the effect of temperature
on bragg grating sensors, sensor # 1 output is presented in
Fig. 9. The complete response of the sensors (FBG_M1)
includes its temperature and moisture effect and bragg
wavelength follows the temperature loading, as the
temperature has a more prominent effect on the change of
bragg grating [35], [36]. Therefore, when we remove the
effect of the daily cycling temperature from the overall
response, a steady WCP determination can be made
(FBG_M1 (T_corr)).

Fig. 10. Envirotemp FR3 oil transformer thermocouples (Top), fiber optics
and water activity probe temperature measurements (Bottom)

2. Moisture Measurements:
Fig. 9. Bragg wavelength of moisture sensor # 1 (Before and after the
temperature correction) in Mineral oil filled transformer Two moisture sensors were installed in the vegetable filled
oil transformer in addition to the water activity probe which
A summary of moisture level estimated by various was installed from the top cover of transformer. Fig. 11
measurement techniques is given in Table III. These presents the moisture response from all the sensors during
different techniques were: the fiber optics sensors, water the cyclic loading of transformer. The moisture measurement
activity sensor immersed in oil and Karl Fischer titration trend by fiber bragg grating sensors were found to be in
method (using pressboard samples placed inside the agreement with commercial Vaisala probe. This shows that
transformers). The Karl Fischer titration was performed only the optical sensors will measure moisture in different
insulating fluids, and can be effectively used in Envirotemp

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

FR3 filled transformer as well. after its temperature correction. The cyclic measurement by
the sensor shows the correct working of the sensor based on
the water activity measured in transformer. A measurement
summary similar to the mineral oil transformer for moisture
estimation during the steady state period by the sensors is
compared in Table III.

Fig. 11. Envirotemp FR3 oil transformer moisture measurement by optical

and water activity probe

Fig. 12 and Fig. 13 shows the fiber bragg grating sensors

response and their corresponding temperatures. The figures
present the moisture level calculated by the sensors during
Fig. 14. Bragg wavelength of moisture sensor # 1 (Before and after the
thermal loading of transformer. The fiber bragg grating temperature correction) in Envirotemp FR3 oil filled transformer
sensor # 2 has shown some deviations during the decrease in
thermal loading around its measurement values. Further V. FIBER OPTICS SENSORS PERFORMANCE
investigation is being carried out to remove this noise using
digital signal processing techniques to smooth the sensor A. Measurement Performance:
output. The moisture measurement from each sensor is compared
in Table III, after the transformers reach equilibrium
condition as mentioned in section IV. The table shows the
average water content estimated by each sensor during the
last hour once the transformer reaches equilibrium.
TABLE III
Estimated Moisture (%) in Transformers after Each Cyclic Loading
Avg. WCP Avg. WCP Avg. WCP
Avg. WCP
after 1st after 2nd after 3rd
Sensor # before Cyclic
Cyclic Cyclic Cyclic
Loading
Loading Loading Loading
Mineral Oil Transformer
FBG_wcp1 2.58 2.38 2.30 2.27
FBG_wcp2 2.68 2.17 2.51 2.33
FBG_wcp3 2.63 2.15 2.65 2.33
FBG_wcp4 2.54 2.19 2.62 2.31
Fig. 12. Water content of paper estimated by moisture sensor # 1 in
FBG_wcp5 2.68 2.95 3.03 2.58
Envirotemp FR3 oil filled transformer
FBG_wcp6 2.61 2.85 2.33 2.29
FBG_wcp7 2.72 2.51 2.35 2.29
FBG_wcp8 2.67 2.44 2.93 2.90
FBG_wcp9 2.60 2.66 2.82 2.59
FBG_wcp10 2.63 2.19 2.49 2.30
Avg. FBG_wcp 2.63 2.45 2.60 2.42
Water activity
2.60 2.35 2.30 2.30
probe
Karl Fisher 2.41 - - 0.66
Vegetable (Envirotemp FR3) Oil Transformer
FBG_wcp1 4.28 3.67 2.83 2.39
FBG_wcp2 4.34 3.25 2.84 2.42
Avg. FBG_wcp 4.31 3.46 2.84 2.41
Water activity
4.20 2.90 2.95 2.50
probe
Karl Fisher 3.40 - - 2.80
Fig. 13. Water content of paper estimated by moisture sensor # 2 in
Envirotemp FR3 oil filled transformer The table shows that the moisture estimated by the fiber
optics sensors is in close agreement with that determined
Fig. 14 presents the optical sensor # 1 output before and

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

using the commercial water activity probe. The Karl Fischer during the cyclic thermal loading of transformers. The results
titration methods were consistent however in one case, after obtained with the use of these sensors correlates very well
the 3rd cyclic loading it gave a very low result. The with a water activity probe and they have demonstrated to be
inaccuracy in measurements from different techniques is effective in different insulating oils. The accuracy of the
expected and have been reported in earlier studies [48]. measurement is also acceptable and shows close results to
that of water activity probe. The application of these
B. Measurement Accuracy:
localized distributed moisture measurement will provide
In order to assess the performance and accuracy of the more precise moisture measurement and it’s distribution
fiber optics sensors, standard deviations for the water content inside a transformer. The fiber optics sensor’s immunity to
measurement in table III is calculated. Table IV presents the electromagnetic interference introduces the possibility to
standard deviations of each sensor during each measurement install the sensors virtually anywhere inside a transformer.
period. For the case of mineral oil transformer sensor, the This flexibility can offer new insights for a system operator
overall average standard deviation for fiber optics sensors is to better manage the assets during normal and overloading
0.118 (minimum deviation of 0.02 and a maximum deviation conditions. The application of digital filtering techniques
of 0.498) while the commercial water activity probe showed will be explored in future to make the sensor’s output more
an average deviation of 0.030. For the case of Envirotemp stable.
FR3 insulating transformer, the overall average standard However, one major limitation of these sensors is their
deviation for fiber optics sensors was 0.055 (minimum installation which can only be performed during
deviation of 0.031 and a maximum deviation of 0.44) while manufacturing or refurbishment stage. Another challenge is
the commercial water activity probe showed an average the long-term stability of these sensors, which is yet to be
deviation of 0.069. Although some sensors have shown large tested. However, it can be anticipated that these sensors will
deviations, which could be because of the temperature be stable, as they use the same polyimide coating that is used
variations and sensors stability. Further investigation will be in water activity probes. Therefore, it can be concluded that
done on the application of digital filtering to stabilize the the proposed optical fiber sensors can be used for real-time
sensor output. moisture measurements in transformers.
TABLE IV
Standard Deviations (SD) of the Measurements
VII. ACKNOWLEDGMENT
SD of SD of SD of
SD of WCP The authors would like to thank the Australian Research
WCP after WCP after WCP after
Sensor # before Cyclic st nd rd
1 Cyclic 2 Cyclic 3 Cyclic Council (ARC) and industry partners AusGrid, Ergon
Loading
Loading Loading Loading
Energy, Powerlink Queensland, TransGrid and Wilson
Mineral Oil Transformer
Transformer for the financial support. Thanks to Insulect for
FBG_wcp1 0.090 0.057 0.057 0.066
FBG_wcp2 0.079 0.051 0.109 0.232
providing FR3 dielectric ester fluid and the Australian
FBG_wcp3 0.033 0.030 0.102 0.246 Government for supporting the HDR program through APA
FBG_wcp4 0.087 0.057 0.127 0.157 scholarship.
FBG_wcp5 0.110 0.037 0.135 0.142
FBG_wcp6 0.093 0.070 0.078 0.127 VIII. REFERENCES
FBG_wcp7 0.045 0.073 0.063 0.075
[1] H. Borsi, “Dielectric behavior of silicone and ester fluids for use
FBG_wcp8 0.106 0.076 0.212 0.498 in distribution transformers,” IEEE Trans. Electr. Insul., vol. 26,
FBG_wcp9 0.098 0.026 0.211 0.364 no. 4, pp. 755–762, 1991.
FBG_wcp10 0.041 0.020 0.152 0.321 [2] Z. Liu et al., “Partial discharge behaviour of transformer liquids
Avg. FBG_wcp 0.078 0.050 0.125 0.2228 and the influence of moisture content,” in 2014 IEEE 18th
Water activity International Conference on Dielectric Liquids (ICDL), 2014, pp.
0.033 0.000014 0.046 0.0424 1–4.
probe
Vegetable (FR3) Oil Transformer [3] F. Heinrichs, “Bubble Formation in Power Transformer Windings
at Overload Temperatures,” IEEE Trans. Power Appar. Syst., vol.
FBG_wcp1 0.032 0.078 0.031 0.040
PAS-98, no. 5, pp. 1576–1582, Sep. 1979.
FBG_wcp2 0.032 0.077 0.040 0.110 [4] CIGRE Working Group A2.30, “Moisture Equilibrium and
Avg. FBG_wcp 0.032 0.078 0.036 0.075 Moisture Migration within Transformer Insulation Systems,”
Water activity 2008.
0.140 0.028 0.061 0.050
probe [5] CIGRE Task Force D1.01.10, “Ageing of Cellulose in Mineral Oil
Insulated Transformers,” 2007.
[6] D. Martin, J. Marks, and T. Saha, “Survey of Australian power
VI. CONCLUSIONS transformer failures and retirements,” IEEE Electr. Insul. Mag.,
In this work, a polymer coated optical fiber sensors based vol. 33, no. 5, pp. 16–22, Sep. 2017.
[7] CIGRE Working Group D1.52, “Moisture Measurement and
moisture measurement system is proposed and tested in this Assessment in Transformer Insulation – Evaluation of Chemical
experiment to measure the moisture inside a real transformer. Methods and Moisture Capacitive Sensors,” 2018.
Two different type of insulated oil filled transformers, [8] T. K. Saha, “Review of Modern Diagnostic Techniques for
Assessing Insulation Condition in Aged Transformers,” IEEE
mineral oil, and Envirotemp FR3 oil, have been used. Trans. Dielectr. Electr. Insul., vol. 10, no. 5, pp. 903–917, 2003.
The results obtained from the new optical sensors in both [9] D. Martin and T. Saha, “A review of the techniques used by
the transformers have been compared with other techniques utilities to measure the water content of transformer insulation

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

paper,” IEEE Electr. Insul. Mag., vol. 33, no. 3, pp. 8–16, May etched fiber Bragg grating,” in 2015 IEEE International
2017. Instrumentation and Measurement Technology Conference
[10] CIGRE Working Group D1.01(TF14), “Dielectric Response (I2MTC) Proceedings, 2015, pp. 422–427.
Diagnoses For Transformer Windings,” 2010. [31] A. Marinescu, G. Opran, M. Teodorescu, I. Dinu, and L. Tascau,
[11] L. Alwis, T. Sun, and K. T. V. Grattan, “Optical fibre-based “Fibre optic based clamping force monitoring system for power
sensor technology for humidity and moisture measurement: transformers,” Proc. Int. Conf. Optim. Electr. Electron.
Review of recent progress,” Measurement, vol. 46, no. 10, pp. Equipment, OPTIM, pp. 282–286, 2012.
4052–4074, Dec. 2013. [32] D. K. Mahanta and S. Laskar, “Water Quantity-Based Quality
[12] IEC, Insulating liquids - Oil-impregnated paper and pressboard - Measurement of Transformer Oil Using Polymer Optical Fiber as
Determination of water by automatic coulometric Karl Fischer Sensor,” IEEE Sens. J., vol. 18, no. 4, pp. 1506–1512, Feb. 2018.
titration. . [33] W. Zhang and D. J. Webb, “PMMA Based Optical Fiber Bragg
[13] B. W. Ward, T. V. Oommen, and J. A. Thompson, “Moisture Grating for Measuring Moisture in Transformer Oil,” IEEE
estimation in transformer insulation,” in Technical presentation Photonics Technol. Lett., vol. 28, no. 21, pp. 2427–2430, Nov.
for the IEEE/PES Transformers Committee. 2016.
[14] Y. Du, M. Zahn, B. C. Lesieutre, A. V. Mamishev, and S. R. [34] J. H. Rodriguez-Rodriguez, F. Martinez-Pinon, J. A. Alvarez-
Lindgren, “Moisture Equilibrium in Transformer Paper Oil Chavez, and D. Jaramillo-Vigueras, “Polymer optical fiber
Systems,” pp. 11–20, 1999. moisture sensor based on evanescent-wave scattering to measure
[15] T. V. Oommen, “Moisture Equilibrium Charts for Transformer humidity in oil-paper insulation in electrical apparatus,” in 2008
Insulation Drying Practice,” IEEE Trans. Power Appar. Syst., vol. IEEE Sensors, 2008, pp. 1052–1055.
PAS-103, no. 10, pp. 3062–3067, Oct. 1984. [35] M. A. Ansari and D. Martin, “Preliminary assessment of using
[16] F. V. Helsinki, “HMP228 Moisture and Temperature Transmitter optical fibre sensors to measure moisture in transformers,” in
for Oil User’s guide,” 2002. 2017 Australasian Universities Power Engineering Conference
[17] W. A. Fessler, T. O. Rouse, W. J. McNutt, and O. R. Compton, (AUPEC), 2017, pp. 1–5.
“A Redefined Mathematical Model For Prediction of Bubble [36] M. A. H. Ansari and D. Martin, “Feasibility of Optical Fibre
Evaluation In Transformers,” IEEE Trans. Power Deliv., vol. 4, Sensors for Moisture Diagnosis in Transformer Insulation,” in
no. 1, pp. 391–404, 1989. 2018 Condition Monitoring and Diagnosis (CMD), 2018, pp. 1–
[18] J. D. Piper, “Moisture Equilibrium Between Gas Space and 6.
Fibrous Materials in Enclosed Electric Equipment,” Trans. Am. [37] A. L. Buck, “New Equations for Computing Vapor Pressure and
Inst. Electr. Eng., vol. 65, no. 12, pp. 791–797, Dec. 1946. Enhancement Factor,” J. Appl. Meteorol., vol. 20, no. 12, pp.
[19] D. Martin, C. Perkasa, and N. Lelekakis, “Measuring Paper Water 1527–1532, 1981.
Content of Transformers: A New Approach Using Cellulose [38] K. O. Hill and G. Meltz, “Fiber Bragg grating technology
Isotherms in Nonequilibrium Conditions,” IEEE Trans. Power fundamentals and overview,” J. Light. Technol., vol. 15, no. 8, pp.
Deliv., vol. 28, no. 3, pp. 1433–1439, Jul. 2013. 1263–1276, 1997.
[20] D. Martin, O. Krause, and T. Saha, “Measuring the Pressboard [39] J. Ascorbe, J. Corres, F. Arregui, and I. Matias, “Recent
Water Content of Transformers Using Cellulose Isotherms and the Developments in Fiber Optics Humidity Sensors,” Sensors, vol.
Frequency Components of Water Migration,” IEEE Trans. Power 17, no. 4, p. 893, Apr. 2017.
Deliv., vol. 32, no. 3, pp. 1314–1320, Jun. 2017. [40] T. L. Yeo, T. Sun, K. T. V. Grattan, D. Parry, R. Lade, and B. D.
[21] D. Martin, T. Saha, T. Gray, and K. Wyper, “Determining water Powell, “Characterisation of a polymer-coated fibre Bragg grating
in transformer paper insulation: effect of measuring oil water sensor for relative humidity sensing,” Sensors Actuators B Chem.,
activity at two different locations,” IEEE Electr. Insul. Mag., vol. vol. 110, no. 1, pp. 148–156, Sep. 2005.
31, no. 3, pp. 18–25, May 2015. [41] T. L. Yeo et al., “Polymer-coated fiber Bragg grating for relative
[22] D. Martin, T. Saha, C. Perkasa, N. Lelekakis, and T. Gradnik, humidity sensing,” IEEE Sens. J., vol. 5, no. 5, pp. 1082–1089,
“Fundamental concepts of using water activity probes to assess Oct. 2005.
transformer insulation water content,” IEEE Electr. Insul. Mag., [42] P. Kronenberg, P. K. Rastogi, P. Giaccari, and H. G. Limberger,
vol. 32, no. 3, pp. 9–16, 2016. “Relative humidity sensor with optical fiber Bragg gratings,” Opt.
[23] K. T. V. Grattan and B.T. Meggitt, Optical Fiber Sensor Lett., vol. 27, no. 16, p. 1385, Aug. 2002.
Technology. Boston, MA: Springer US, 1998. [43] G. Berruti et al., “A Comparative Study of Radiation-Tolerant
[24] N. Haque, J. V. Ittiarah, T. K. Gangopadhyay, and S. Chakravorti, Fiber Optic Sensors for Relative Humidity Monitoring in High-
“Temperature monitoring of power transformer using fiber-optic Radiation Environments at CERN,” IEEE Photonics J., vol. 6, no.
sensor,” in 2015 1st Conference on Power, Dielectric and Energy 6, pp. 1–15, Dec. 2014.
Management at NERIST (ICPDEN), 2015, pp. 1–5. [44] A. Makovec et al., “Radiation hard polyimide-coated FBG optical
[25] Yi Jiang, Shuang Liu, Li Xiao, and Wei Li, “Fiber Bragg grating sensors for relative humidity monitoring in the CMS experiment
sensors for temperature monitoring in oil-immersed at CERN,” J. Instrum., vol. 9, no. 03, pp. C03040–C03040, Mar.
transformers,” in 2016 15th International Conference on Optical 2014.
Communications and Networks (ICOCN), 2016, pp. 1–3. [45] D. Lo Presti, C. Massaroni, and E. Schena, “Optical Fiber
[26] X. Zhang, S. Yao, R. Huang, D. Hou, W. Huang, and M. Zheng, Gratings for Humidity Measurements: A Review,” IEEE Sens. J.,
“Oil-immersed transformer online hot spot temperature vol. 18, no. 22, pp. 9065–9074, Nov. 2018.
monitoring and accurate life lose calculation based on fiber Bragg [46] Smart Fibres Ltd, “SmartScan.” [Online]. Available:
grating sensor technology,” in 2014 China International https://www.smartfibres.com/products/smartscan. [Accessed: 19-
Conference on Electricity Distribution (CICED), 2014, no. Ciced, Dec-2017].
pp. 1256–1260. [47] I. Atanasova-Höhlein et al., “Experience with capacitive on-line
[27] P. Zubiate, J. M. Corres, C. R. Zamarreno, I. R. Matias, and F. J. sensors for moisture evaluation in transformer insulation,” IEEE
Arregui, “Fabrication of Optical Fiber Sensors for Measuring Electr. Insul. Mag., vol. 35, no. 2, pp. 18–26, Mar. 2019.
Ageing Transformer Oil in Wavelength,” IEEE Sens. J., vol. 16, [48] D. Martin, N. Lelekakis, J. Wijaya, and K. Williams, “Water
no. 12, pp. 4798–4802, Jun. 2016. uptake rates of transformer paper insulation impregnated with
[28] D. K. Mahanta and S. Laskar, “Investigation of transformer oil vegetable oil,” IEEE Electr. Insul. Mag., vol. 29, no. 5, pp. 56–61,
breakdown using optical fiber as sensor,” IEEE Trans. Dielectr. Sep. 2013.
Electr. Insul., vol. 25, no. 1, pp. 316–320, Feb. 2018.
[29] G.-M. Ma, C.-R. Li, R.-D. Mu, J. Jiang, and Y.-T. Luo, “Fiber
bragg grating sensor for hydrogen detection in power
transformers,” IEEE Trans. Dielectr. Electr. Insul., vol. 21, no. 1,
pp. 380–385, Feb. 2014.
[30] J. Jiang, H. Song, G. Ma, C. R. Li, Y. Luo, and H. Wang,
“Dissolved hydrogen detection in power transformer based on

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2924271, IEEE
Transactions on Power Delivery

10

IX. BIOGRAPHIES

Muhammad A. Ansari (S’17) received his B.E.

from NED University of Engineering and
Technology, Pakistan and M.Sc. from King Fahd
university of Petroleum and Minerals, Saudi Arabia
in 2009 and 2015, respectively. Before joining here,
he was working in the industry as a substation design
engineer. Currently, he is a PhD candidate in the
School of Information Technology and Electrical
Engineering at The University of Queensland, St Lucia, QLD, Australia. His
research interests include condition monitoring of equipment, substation
assets, and measurement techniques.

Daniel Martin (M’12) received the B.Eng. degree in

Electrical and Electronic Engineering (with study
abroad in Germany) from the University of Brighton,
UK in 2000. He has a Ph.D. degree in electrical
insulation from the University of Manchester, UK.
He investigated the suitability of using vegetable oils
and synthetic esters as substitutes for mineral oil
within large power transformers, graduating in 2008.
He is currently a lecturer in power and energy
systems at the University of Queensland, Australia,
with his research interest being on asset management of power transformers.

Tapan Kumar Saha (M’93–SM’97) was born in

Bangladesh in 1959 and immigrated to Australia in
1989. He received the B.Sc. degree in electrical and
electronic from the Bangladesh University of
Engineering and Technology, Dhaka, Bangladesh, in
1982, the M.Tech. degree in electrical engineering
from the Indian Institute of Technology, New Delhi,
India, in 1985, and the Ph.D. degree from the
University of Queensland, Brisbane, Australia, in
1994. He has had visiting appointments for a semester with both the Royal
Institute of Technology, Stockholm, Sweden, and the University of
Newcastle, Australia. He is currently a Professor of Electrical Engineering
with the School of Information Technology and Electrical Engineering,
University of Queensland. His research interests include condition
monitoring of electrical plants, power systems, and power quality. He is a
fellow of the Institution of Engineers, Australia.

0885-8977 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.