An experimental assessment of the intermittent heating and cooling operation effect on power transformer insulation

Report: An experimental assessment of the intermittent heating and

cooling operation effect on power transformer insulation

Stockholm, April 19, 2021

Partner Hitachi ABB Power Grids Authors Daniil Danylov

Yanick Patrick Frei Leonardo Colombo

Raimundo Montalba Mesa Zhanzhan Liu

Patrick Janus

Kateryna Morozovska

ISBN: 978-91-7873-820-5

Series: TRITA-EECS-RP

Abstract

In the future grid with a high share of renewables, grid flexibility is becoming an important issue to solve. Dynamic Transformer Rating (DTR) is a technology that allows to unlock additional grid capacity and provide extra flexibility to power system operation. However, since DTR is a fairly novel operation method, the effect of this type of operation on the rate of insulation ageing has not been studied in detail yet. Dynamic change of load and insulation temperature can introduce new extreme to the operation and lower the insulation lifetime.

Therefore, this work studies the impact of dynamic rating on transformer insulation when exposed to different heating cycles.

This work proposes an experimental evaluation of the impact that different heating cycles can have on insulation degradation. The setup consists of the control loop, heating element and paper-oil samples. The main focus is on the moisture exchange between insulating paper and oil when the samples are exposed to different temperature cycles. Moisture has been measured for both oil and paper and compared to a control group.

This report presents published data from the first performed lab test. The experiment has to be repeated several times to increase the validity of the results. Consequently, the experiment has been documented carefully in order to ensure its replicability.

It is shown that the limited interaction of the inner paper layers with the oil has forced the moisture to stay there. Moreover, part of the samples exposed to the highest temperatures is observed to be damaged.

Key words: Dynamic transformer ratings; Temperature cycle; Paper and oil insulation;

Degradation

Contents

1. Theoretical background 6

1.1. Paper . . . . 6 1.2. Oil . . . . 7 1.3. Interaction of paper and oil . . . . 8

2. Experimental setup 9

2.1. Sample description . . . . 12 2.2. Sample preparation . . . . 13 2.3. L

V

-software . . . . 14

3. Measurement procedure 17

3.1. Sample preparation . . . . 17 3.2. Titration analysis . . . . 18

4. Data processing 20

4.1. Oil data processing . . . . 20 4.2. Paper data processing . . . . 20

5. Results and discussion 22

5.1. Results . . . . 22 5.2. Discussion of results . . . . 22

6. Conclusion 26

A. M

scripts 29

B. Required material 43

C. Measurements data 44

Terminology (alphabetical)

Control — Software (in this case L

V

) that controls the experiment by reading tem- perature values and turning heaters on and off.

Cycle — A cycle consists of a heating up and cooling down phase with defined duration.

DAQ — Data acquisistion system.

DGA — dissolved gas analysis.

DP — degree of polymerization.

DTR — Dynamic transformer ratings.

Group — A group consists of the samples exposed to the same load pattern.

Load pattern — A defined temperature (or dissipated power) curve over time.

Oil — If nothing else is stated refers to transformer insulation oil.

Paper — If nothing else is stated refers to transformer insulation paper.

ppm — parts per million.

PWM — pulse-width modulation.

RTFM — rapid temperature floating measurements.

Sample — A sample is a single test subject, which is exposed to one of the load patterns.

Setup — The setup consists of 15 samples and therefore refers to the whole experiment.

SSR — Solid state relay; used to switch on and off the heaters.

List of Figures

1. The three different load patterns under study . . . . 9

2. Linearized temperature function . . . . 10

3. Switching and temperature measurement of one sample . . . . 11

4. Components of a sample . . . . 12

5. Experimental setup; (a), electrical connection and sample disposition in the oven; (b), temperature profile; (c) overall experimental setup. . . . 13

6. How the control circuit interacts with the heaters . . . . 15

7. L

V

block that controls the SSR of one load pattern [1] . . . . 15

8. L

V

block that measure and save temperature [1] . . . . 16

9. Samples separated into different groups . . . . 18

10. Coulometric KF Titrator C10/C20S/C30s [2] . . . . 18

11. The measured datapoints (mean value of three measurements); the gray lines represent the control groups (two for paper, one for oil) . . . . 20

12. Moisture distribution in paper and oil; the grey lines represent the control groups (two for paper, one for oil) . . . . 22

13. Temperature pattern (5 min average) . . . . 25

List of Tables 1. Diffusion rates and time constants for different temperatures . . . . 10

2. Weight of the paper samples . . . . 14

3. Moisture measurements for groups of paper and oil . . . . 23

4. Measurement data: control samples . . . . 44

5. Measurements data for each sample . . . . 44

1. Theoretical background

In this section, the two main parts of a transformers insulation system are introduced, namely, insulation paper and oil. It is clear that the main purpose of this experiment is to study the degradation of these insulation materials and thus the focus in the following descriptions is on how these degrade, how does this manifest in a measurable manner and a brief description of how to conduct these measurements.

1.1. Paper

Solid insulation (e.g. paper) is mostly made from processed wood pulp where cellulose is an essential end product [3]. Cellulose is a polymer that consists of long chains of glucose rings and has a formula of [C

H

O

]

, where “n” is the number of linked rings. One cellulose chain can consist of up to 15 000 glucose rings, which are held together mainly by hydrogen bonds and hydrophobic interaction.

Transformer windings temperature is not uniformly distributed and has the hottest section, which is commonly referred to as a hot spot. Every 6-10

C rise in temperature increases the degradation rate of the paper increases by a factor of two [4]. Moreover, heat accelerates the degradation process done by oxygen content as it acts as a catalyst in the chemical reac- tion. Oxygen content causes oxidation, meaning oxygen attacks carbon atoms in the cellulose chain. This process leads to the formation of acids and the release of water. The increase of water content both due to oxidation and other reasons (e.g. leaks in tank etc.) increases the degradation rate because moist insulation may lead to the formation of water vapour bubbles at elevated temperatures. This formation can lead to a partial discharge or even insulation break- down. Therefore, it is vital to maintain the winding temperature below a specified threshold.

As mentioned before, cellulose fibres are made of long chains of glucose rings. Over time they undergo a depolymerization process, which leads to their shortening. The length of the fibres has a direct effect on the mechanical properties of the cellulose, such as tensile strength and elasticity. They both govern the ability to withstand mechanical stresses and, if the cellulose chain length is reduced enough, ultimately it can no longer withstand short- circuits, or even vibrations from normal operations [5]. For measuring the length of glucose ring chains, a Degree of Polymerization (DP) is used. It corresponds to the average number of glucose rings in the molecule and can be used to assess retained functionality of transformer insulation as it ages in service. For example, a new paper has a DP of 1 200. After being prepared in the factory drying process, installed paper in a transformer has a DP of about 1 000. DPs in the range 950 to 500 does not significantly affect the strength of the paper. The useful lifetime of paper is usually set to DP 200 since it is around the point where the paper has become brittle and has lost the majority of its mechanical strength [6].

The best way to assess the paper deterioration is to perform tensile and elongation tests on

paper strips and test bursting strength. However, this is generally not possible because this

requires obtaining an insulation sample from a working transformer which is both impractical

and destructive. Instead, this is usually done by indirect measurements such as sampling and

analysis of transformer oil through dissolved gas analysis (DGA). The key gases for assessing

paper degradation are carbon oxide (CO) and carbon dioxide (CO

). However, both amount

of oxygen (O

) and consumption of it (if present) is used to interpret the severity of paper

degradation.

Another test that can be used to assess paper degradation is the furan analysis [7]. When a cellulose molecule breaks, a chemical compound - furan - is formed. In the industry, five furan compounds are measured:

• 2-furaldehyde (2FAL)

• 5-methy-2-furaldehyde (5M2F)

• 5-hydroxymethyl-2-furaldehyde (5H2F)

• 2-acetyl furan (2ACF)

• 2-furfurol (2FOL)

These compounds are partially dissolved in oil, where they can be identified and measured using high-performance liquid chromatography. However, different types of paper give rise to different sets of furans, and the concentration of the furanic compounds depends on the ratio between oil and cellulose. There are known uncertainties regarding the furanic compounds’

stability for the typical condition of transformer temperatures and dissolved oxygen concen- trations. 2FAL is usually considered the primary compound in the analysis due to its relatively higher generation rate and stability [8]. It is also a characteristic component of paper degra- dation and is hence directly related to the DP of the paper. Even though furan analyses are conducted as part of transformer diagnostics, the results investigate whether paper degradation has occurred. The furan content in the oil provides a mean value from the paper degradation of the total amount of paper. The degradation may have occurred locally, and thus the DP-value may differ considerably between different places in the solid insulation.

1.2. Oil

Oil for transformer insulation is oil, which remains durable at high temperature. The trans- former oil used in most transformers is a type of petroleum-based hydrocarbon, which has been chosen due to its high dielectric strength and chemical stability [9]. If we talk about the normal service life of an oil-filled material, the transformer oil inside the apparatus is subjected to worsening due to several causes. Transformer oil deterioration is a number of changes in the insulation fluid that occur during the operation of a transformer. These usually result in the degradation of the physical, chemical, and operational properties of the oil.

When the oil comes in contact with oxygen, an acid is formed due to oxidization. The acid will form a sludge that settles on the transformer windings resulting in reduced heat dissipation. The windings will run hotter, thereby creating more sludge, which will create even more heat. The high acid content and increased temperatures will accelerate the deterioration of the insulating qualities of the oil and, if left untreated, may cause a transformer failure.[10]

A dissolved gas analysis (DGA) is the most commonly used tool to diagnose transformer oil health. Even under normal operation, a power transformer generates a certain amount of gases. The rate at which these gasses are generated can be used to trace the existence of an operation fault [6].

A proactive approach should be taken if transformer oil has a high acid content. Any sludge

formed by the acid must be rinsed out of the transformer with hot oil to remove the sediment.

There is a cost-saving if the oil is reclaimed in the early stages of the acid build-up before the sludging occurs, as the oil will retain its quality longer under normal operating conditions.

Transformer oil can hold water particles in suspension depending on the temperature of the oil. If the oil is at its saturation point, freed water is likely to collect at the bottom of the transformer. The oil dielectric strength decreases if water is present, which may require a degasification of the oil. If the water content is high, a hot oil dries out should be considered.

Although more costly than degasification, this will also remove any water that the core and coil assembly may contain [11].

A significant amount of the water content present in the transformer gets dissolved into the oil. The solubility of water in the oil increases with temperature. Thus more water will be present in the oil if operating at higher temperatures, leading to higher water content and weaker dielectric strength [12]. The amount of water in oil may be measured utilizing Karl Fischer titration, a method that uses reagents to determine unknown concentrations in a solu- tion by measuring the dielectric breakdown voltage of the solution [13].

1.3. Interaction of paper and oil

The degradation of paper and oil complement each other, increasing the degrading process of each other. The degradation of paper is leading to the release of water which increases the ratio of potentially soluble particles dissolving in the oil and reduces its dielectric capabili- ties. Simultaneously, the sludge formed in the oil leads to an increase in temperature, which accelerates the release of even more acids and water from the paper. [10]

The composition of the water molecules dictates their behaviour at the interface of the pa- per and oil. Cellulose, being a hydrophilic material, attracts water molecules and holds them, while on the other hand, oil being a hydrophobic material, repels water molecules naturally.

How strong these characteristics are is dictated in part by temperature. The moisture equi- librium between different mediums will change, getting closer to an even distribution as the temperature rises [14]. These effects can be seen when considering the diffusion time con- stants, which are explained later in section 2. It is then expected that with different heating cycles, a different degree of degradation will be observed. An expected result is a higher de- gree of degradation (higher concentration of water in oil) with an increasing number of the heating cycle.

When talking about transformer temperatures, one usually refers to the concept of Tem-

perature Rise. Transformer temperature rise is defined as the average temperature rise of the

windings above the ambient temperature when the transformer is loaded at its nameplate rat-

ing [15]. For liquid insulated transformers, the temperature rise is typically 55

C or 65

C,

which means an operating temperature at rated conditions of around 100

C, with overloads

occurring whenever the transformer windings operate at higher temperatures. In this experi-

ment, the sample will be loaded with different loading cycles, and the final moisture content

of oil and paper will be measured.

2. Experimental setup

The original experimental setup is described in detail in [3]. The experiment consists of a total number of 15 samples that are exposed to load patterns as indicated in figure 1. The samples are heated using cartridge heaters to T

= 130

C, for a duration of at least t

. Value of t

is based on the diffusion time constant τ

and has strong dependency on the temperature as shown in (2). Afterwards, the sample is cooled down to ambient temperature T

.

T

t

T

t

T

t

T

T

T

T (t)

t

Figure 1: The three different load patterns under study

The three load patterns differ in the number of cycles they are exposed to. To achieve one, two or three cycles, the corresponding heating and cooling times have to be adjusted.

The 15 samples (they will be elaborated a little further) are divided into three groups with five samples. The setup is completed by at least one control sample that is not exposed to any temperature cycles. The samples of one group are connected all in parallel to the control loop.

To obtain the desired maximum temperature T

= 130

C, the voltage is reduced with a transformer. A parallel connection is a modification to [3].

The time constant was modified compared to the original case to shorten the experiment time [3]. The diffusion time rate is (1) and time constant can be calculated as (2) [16][17][18].

D = D

· e

 m

s



(1)

τ

= 4 · d

π

· D [snb] (2)

where E

= 8074 is the temperature reference point, [K]; D

= 1.34·10

is the diffusion

rate at T

, [

]; T

= 298 K; C = 0.5 % is the moisture concentration, percentage by weight.

[3] [18].

These values have already been calculated in [3] for 22

C and 130

C, however this study has assumed 40

C in order to reduce the diffusion time constant. The results from (1) and (2) are summarized in Table 1.

Temperature D τ

22

C 103.4 · 10

m

/s 793.6 · 10

s≈ 9.2 days 40

C 498.3 · 10

m

/s 164.8 · 10

s≈ 1.9 days 130

C 157.5 · 10

m

/s 521.2 s≈ 8.7 min

Table 1: Diffusion rates and time constants for different temperatures

Knowing the maximum temperature that will be reached and the time constant τ

, it is possible to model the temperature rise as:

T (t) = T

+ (T

− T

) · h

1 − e

i

(3) However, given the linear temperature rise as shown in Figure 2, (3) can be simplified to (4) assuming the heating power is much higher than the radiated power at T

= 130

C.

T

t

Real temperature function

Linear temperature approximation

Tol.band

Figure 2: Linearized temperature function

T (t) ≈ T

+ P

· t

C

(4)

where P

=

= 50 W is the power provided by the heater. This is due to the fact that the heater used in the experiment has a rated power of 200 W, but the voltage applied to it is only half to its rated value.

Thermal capacity for the sample can be calculated as a sum of its components, namely oil, cartridge heater and aluminium block, but the thermal capacity of the paper is not taken into account:

C

≈ m

· C

+ m

· C

+ V

· ρ

· C

= 1 049 [J/K] (5)

where C

= 460.5 J/(kg·K) [19] is steel’s specific heat capacity; C

= 921.1 J/(kg·K) [19] is the specific heat capacity of aluminiun; C

= 1 860 J/(kg·K) [20] is the specific heat capacity of oil; ρ

= 877 kg/m

[21];V

= 42 ml is the volume of oil in the sample [3];

m

= 0.060 kg is the mass of the cartridge heater; m

= 1.034 kg is the mass of the aluminium block.

Using the calculation from (5) the heating phases will have a duration of ∆t

(with ∆T

being the value of the temperature tolerance band defined in figure 2 ).

∆t

= ∆T

· C

P

= 20.97 s · K (6)

By setting the value of the temperature tolerance band at 1

C, the resulting heating cycles will have a duration of approximately 20 s. In this equation, the effect of radiation cooling of the sample is neglected because the main interest of the study is the minimum switching time.

The switching is done by the control loop in L

V

. To further reduce the power of the heaters, the program switches the heaters using a PWM (Pulse Width Modulation) pattern. The program also collects data on temperature change with temperature sensors. The schematic of the control loop is shown in Figure ??.

+

T

L

V

Figure 3: Switching and temperature measurement of one sample

2.1. Sample description

Figure 4 represents structure of the test samples, consisting of cartridge heaters, paper sample, mineral oil and temperature sensor which are placed in a glass tube sealed with rubber lid.

Glass tube is subsequently placed into aluminium block to ensure even heat dissipation from the sample.

Aluminium block Transformer oil Temperature sensor (between paper layers) Transformer paper Cartridge heater Glass tube Sealing lid

Figure 4: Components of a sample

Cartridge heater: The cartridge heater represents the heat source inside a transformer. It heats up the sample if required and keeps the temperature at a high level. The voltage of the cartridge heater is reduced to reduce the generated heat, which decreases power with P ∝ U

ratio.

Transformer paper: The transformer paper is one of the test subjects. It is wrapped in 3 layers around the cartridge heater.

Temperature sensor: A temperature sensor is placed between the outer and middle layer of the test paper. The sensor must always be installed in the same position for all samples to get reliable results. The temperature sensor has to be immersed in oil completely.

Transformer oil: The transformer oil is a second test subject. The interaction between oil

and paper samples happens as moisture particles are exchanged between the two components

following the temperature change.

Glass tube: Thermally upgraded glass is a preferable material because it has low interaction with the test subjects and has high-temperature resistivity.

Sealing lid: Sealing lead helps to prevent additional moisture from penetrating into the tube during the experiment. The lid has holes for the electric connection of both cartridge heater and temperature sensor; the holes are covered with epoxy resin to prevent further air leakage.

Aluminium block: The glass tube containing the experiment is placed inside an aluminium block. Aluminium has high thermal conductivity, which helps to ensure uniform temperature distribution outside the test tube.

2.2. Sample preparation

A detailed procedure of sample preparation is described below. Figure 5 shows the experi- mental setup with climate chamber, computer screen and a general view.

(a) (b)

(c)

Figure 5: Experimental setup; (a), electrical connection and sample disposition in the oven; (b), temperature profile; (c) overall experimental setup.

1. Transformer paper (Munksjö Thermo 70 natur [22]) is cut into stripes with equal width and length, respectively 2 cm and 15 cm.

2. The stripes are weighted on a high-precision digital scale Mettler AE260 to be of an equal weight of 1.8 g. This weight value is considered suitable for the decided paper- oil-ratio [3]. Weights 15 samples and a control sample are summarized in Table 2.

Weight Weight Weight Weight Weight

sample 1 sample 2 sample 3 sample 4 sample 5 Group 1 1.8063 g 1.8079 g 1.8042 g 1.8088 g 1.8057 g Group 2 1.8063 g 1.8033 g 1.8053 g 1.8088 g 1.8023 g Group 3 1.8073 g 1.8073 g 1.8067 g 1.8026 g 1.8034 g

Table 2: Weight of the paper samples

3. The 15 samples are wound manually on the heat cartridges so that 3 insulating paper stripes create one layer of insulation. A complete sample has 3 layers of insulating paper. In order to keep the paper in position, small strips of teflon are used both on the top and bottom of last insulation layer. Thermocouples are then inserted between the second and third layers in order to lower their exposition to the heat source.

4. While the paper is wound around the cartridges, a batch of transformer oil (Nynas Nytro 10 XN [21]) has been dried (this oil was used to fill the glass tubes where the wound cartridges were allocated). The drying process takes 24 hours in a vacuum oven. After that, it has to be promptly transmitted into clean glass bottles and tightly sealed to reduce its contact with air in order to avoid moisture absorption.

5. After the oil has dried out, the wounded cartridges are needed to follow the same drying process, except that they must be soaked into oil to help remove moisture from the paper.

6. After the paper has dried out, the complete experimental setup is assembled. Each of the 16 glass tubes is filled with 60 g of a dry oil, and the heat cartridges with transformer paper on them are soaked into it. Then the glass tube is placed into an aluminium block and sealed with a nitrile cork. Each cork has a hole through which power cables and thermocouple wires are passed. Finally, the cork and the hole in it are sealed with epoxy (Loctite EA3430 and 3450 [23]) in order to minimize the air exchange and possible atmospheric moisture contaminations.

2.3. L AB V IEW -software

The design of the experimental set-up is shown in figure 3 and figure 6. DAQ system from NI

controls the whole system, serving the function of control and data acquisition. The design

aims to control simultaneously the three load patterns illustrated in figure 1 and saves values

from temperature sensors in the sample. The software has – with minor adjustment – been

taken over from a previous work [1].

R

115 VAC

Relays

L

V

Figure 6: How the control circuit interacts with the heaters

Figure 7: LABVIEWblock that controls the SSR of one load pattern [1]

The control system is programmed in L

V

, a system-design platform for a visual programming language. The block diagram for the programming structure of one load pattern and the temperature measurement is depicted in Figure 7 and Figure 8.

Five parts are included in the SSR control program. The whole structure is a while-loop.

An open-loop setting achieves the main control of temperature by using the on-off switch time specified in parts 2 and 3.

• Part 1 contains a large while loop, reflected in the first frame of the main body sequence structure. It also contains a smaller sequence structure, including part 2 and part 3, to switch the on-off state. The while loop contains an elapsed time block named ‘LP1 ON’, which controls the load pattern’s on-state running time. A display function of elapsed time and time left is also created in this part.

• In part 2 and part 3, a PWM waveform is created with a certain on-off time. The waveform is assigned to the digital output port by the DAQ assistant to trigger SSR.

Note that the maximum digital output of NI BNC-2120 is +4.4 V with a drive current 13mA [24].

• Part 4 is similar to part 1. The while-loop and sequence-structure control the off-time of

the load pattern and is activated once time has elapsed in part 1. Once the time elapses

Figure 8: LABVIEWblock that measure and save temperature [1]

in this part, the sequential structure will activate part 1 once again. A DAQ assistant configured with a digital output (0.6 V, 8 mA/1.6 V, 24 mA DC) [24] blocks the SSR for an off-state in this part.

In the system shown in Figure 7, in a cycle, the constant value in Part 1 is 2, indicating 2 heating up days; the constant value in Part 4 is 10, illustrating 10 cooling down days.

Thus, a complete cycle in this instance is 12 days.

• Part 5 consists of lamp display functions controlled by the local variables of the while- loops that communicate with the front panel indicating if the load pattern is active or inactive and if the SSR is on or off.

In figure 8 [1] the temperature measurement structure is presented. It collects the temper- atures of the experimental setup and logs them to the computer. As the control is open-loop, there is no connection to the switching parts.

• Element 1 is a DAQ assistant configured with a thermal data acquirement and connected to a waveform chart to the plot. It acquires the temperature measurements from the DAQ system and calculates with different thermal types.

• Element 2 is a signal collector; element 3 is a writing function. These elements collect the temperature measurements and save them in a text-format file.

• Element 4 is a delay section that determines the measuring frequency.

3. Measurement procedure

The measurements have been focused on determining the water content of each layer of paper.

The oxidation process of the cellulose chains results in the release of water, and the moisture migration increases the ageing of the paper and oil insulation. Samples have been divided into groups of oil samples and paper samples.

The Karl Fischer method for the water content determination is one of the most frequently used titration methods. The water content determination is based on the reaction described by Smith, Bryanz and Mitchell[25]. Numeric definition details in Section 5.2.

I

+ SO

+ 3Py + H

O → 2Py − H

I

+ Py · SO

Py · SO

+ CH

OH → Py − H

CH

SO

(7)

There are two methods used in Karl-Fischer titration volumetric method and the coulomet- ric method. The selection of the appropriate technique is based on the estimated water content in the sample: coulometric method is for cases when the water content is around 1-5 ppm[26].

Thus, coulometric Karl-Fischer titration is applied to our measurement.

It is not possible to directly titrate solid samples using KF coulometry[27]. External extrac- tion is suitable for insoluble solids( paper in this case). The organic solvent methanol is most commonly used for insoluble organic solids[28]. The external extraction for paper is carried out in three steps:

• Step 1: Blank value determination of methanol

• Step 2: Weigh-in methanol and sample

• Step 3: Extraction (Shake)

3.1. Sample preparation

Moisture content in the paper and mineral oil samples is determined by Karl Fischer titration.

Each of the studied samples components are separated into 3 big groups, which are analyzed separately. The division into groups is shown in Figure 9:

• Innermost paper: this is the paper that has been wound directly on the cartridge heater.

In total, it constitutes one big group where there are 15 samples from 3 load pattern groups. This group will also be referred to as batch 1 or ‘B1’.

• Middle and outer paper: this is the paper that has been wound on the innermost paper from batch 1. This group also consists of 15 samples from 3 load pattern groups. This group will also be referred to as batch 2 or ‘B2’.

• Oil: this is one group where 15 samples of oil from 3 load pattern groups are present.

Each sample from the groups ‘B1’, ‘B2’ and oil are measured three times to increase the accuracy of measurements. This is also valid for control samples for paper, methanol and oil.

After that, an average of three sample measurements is taken and is used in further analysis.

Figure 9: Samples separated into different groups

3.2. Titration analysis

This section describes the steps followed while performing the coulometric KF titration. The following figure depicts the Coulometric KF Titrator C10/C20S/C30s series, which was used to perform the titration analysis on the samples.

Figure 10: Coulometric KF Titrator C10/C20S/C30s [2]

However, before starting to perform the measurements, it was necessary to perform a pre- titration. This procedure aims to eliminate the presence of water in the reagent to avoid a bias in the titration results. It was possible to start the measurements when the drift value determined by the titrator was lower than a fixed threshold. The steps followed to measure the water content in the samples are summarized below:

1. Fill a syringe with the samples’ content.

2. Weight the syringe the first time, it is recommended to reset the scale to 0.

3. Tap Start ⇒ add the sample into the titrator.

4. Inject approx. 0.5 to 2 g of sample into the measuring cell.

5. Place the syringe on the weight scale and calculate the weight of the sample injected into the titrator.

6. Enter the sample weight on the touch screen and tap OK. ⇒ The analysis starts; Once

the titration is complete, the results dialogue is displayed.

4. Data processing

Further analysis is performed using measurements and utilizing data collected using the data acquisition system shown in Figure 3(c). The scripts written for data processing are presented in Appendix A. The measured data points are summarized in Figure 11 and shown in detail in Appendix C.

G1B1 G2B1 G3B1 G1B2 G2B2 G3B2

0 200 400 600 800 1 000 1 200 1 400 1 600 1 800 2 000

Methanolmoisturecontent(ppm)

Moisture measurement for paper

G1B1 G2B1 G3B1

0 5 10 15 20 25

Oilmoisturecontent(ppm)

Moisture measurement for oil

Figure 11: The measured datapoints (mean value of three measurements); the gray lines represent the control groups (two for paper, one for oil)

4.1. Oil data processing

Oil data processing is done in the following way: in total, there are 15 samples from 3 load pattern groups; each of the samples has been measured three times. In order to get the added moisture content, we subtract the average of the oil control-sample measurements from the average of the oil test-sample measurements as follows:

M

= O

− O

, (8)

where M

is the added moisture in oil after heating,[ppm]; O

is the average moisture content of the oil test-sample, [ppm]; O

is the average moisture content of the oil control- sample, [ppm] and n is the sample number.

4.2. Paper data processing

Karl-Fischer titration is a process that can determine the moisture content in liquids, not solids.

Therefore it is necessary to extract moisture from paper samples first, which is achieved by

submerging paper samples into methanol for several hours. The original moisture content

of the dry methanol has to be taken into account when determining the moisture content

from paper samples. In this experiment, two different types of methanol were used, which

introduced an extra degree of data processing.

1. The moisture level of the control sample of methanol is subtracted from the test sam- ple’s moisture level. As a result, an unbiased ppm sample value is obtained. However, this value indicates the moisture level of both methanol and paper. Therefore further processing is needed to find out only paper moisture. If two methanol mixtures have been used, they have to be subtracted by their proportion:

U

= S

− C

, (9)

where U

is an unbiased sample value, [ppm]; S

is an test-sample moisture level, [ppm]; C

is the original moisture content of methanol for the sample, [ppm]

and n is the sample number.

2. The bias from the control paper samples is found. For that one finds the difference between the paper control-sanple ppm and the corresponding methanol control-sample ppm and divides it by the mass of the control-sample:

Bias

= C

− C

M

, (10)

where Bias

is the bias from the control paper samples, [ppm]; C

is the paper control-sample, [ppm] and M

is the weight of the paper control-sample, [g].

3. Finally, one founds moisture in the paper as a ratio of the unbiased sample value to the bias of the control paper samples:

P

= U

Bias

, (11)

where P is the moisture content in the paper test-sample, [ppm].

5. Results and discussion

In this section, the results obtained from the experiment are presented.

5.1. Results

From Figure 13, which shows the temperature as a function of time, the following results are obtained:

• G1 is heating for 6 days and cools down for 6 days; average group temperature equals to 99.6

C during the heating time.

• G2 is heating for 4 days and cools down for 4 days; average group temperature equals to 122.0

C during the heating time.

• G3 is heating for 2 days and cools down for 2 days; average group temperature equals to 96.1

C during the heating time.

In general, it can be said that the temperature distribution varies within one group but also is comparable with the other groups. The measured data has several outliers, which may add additional variability to the distribution.

Following the procedures presented in the previous section, temperature patterns and moisture distribution results in oil and paper for different groups are obtained. They are shown in Figure 12 and Table 3.

G1B1 G2B1 G3B1 G1B2 G2B2 G3B2

0 200 400 600 800 1 000

Moisturecontent(ppm)

Moisture content in paper

G1B1 G2B1 G3B1

0 5 10 15 20 25

Moisturecontent(ppm)

Moisture content in oil

Figure 12: Moisture distribution in paper and oil; the grey lines represent the control groups (two for paper, one for oil)

5.2. Discussion of results

Starting with the paper from Figure 12 (left), the median moisture levels inside the paper of

groups G1 B1, G2 B2, G1 B2, G2 B2 and G3 B3 are lower than in the paper control sample

P a p e r

Median (ppm) Maximum (ppm) Minimum (ppm)

G1 B1 372.87 441.39 181.54

G2 B1 311.36 914.21 190.06

G3 B1 403.78 590.32 245.51

G1 B2 218.75 236.64 167.25

G2 B2 253.38 340.83 181.58

G3 B3 169.68 290.51 148.6

Control 1 186.85 189.11 183.35

Control 2 402.11 414.93 392.04

O i l

Median (ppm) Maximum (ppm) Minimum (ppm)

G1 14.1 15.1 9.2

G2 16.8 20.5 9.1

G3 12.2 14.9 9.3

Control 14.5 15.6 12.2

Table 3: Moisture measurements for groups of paper and oil

with 402.11 ppm, which shows that there is a relatively active moisture exchange between the paper and the oil. On the contrary, the median paper moisture in group G3 B1 exceeds the control level and is higher than for groups G1 and G3 of ‘B1’. The explanation for this can be derived when looking at Table 1. According to it, the diffusion time constant τ

equals 9.2 days for a temperature of 22

C and 1.9 days for a temperature of 40

C. Since the sample temperature during the cooling process is around 25-26

C and the cooling time for group G3 is equal to 2 days, there has not been enough time for the moisture to diffuse from paper into the water and reach equilibrium. Therefore, some moisture has been trapped inside the paper samples.

Regarding the oil, in Figure 12 (right), the moisture levels of the groups G1 B1 and G3 B3 are equal to or less than the oil control moisture level. However, the median oil moisture for group G2 is higher than those of G1 and G2. The explanation for this can be derived when looking at the temperature pattern in Figure 13, where the average temperature difference between the groups G1, G3 and group G2 is around 22

C. Moreover, during the sample preparation for Karl Fischer titration, it is discovered that samples in the second group, especially the innermost ones, have burnt. Naturally, the higher temperature for group G2 did not easily let moisture from oil diffuse into the paper, forcing moisture to stay outside in the oil. It holds true mainly for the hot phase. During hot stages, it can be assumed that the equilibrium is reached based on (2).

Another interesting result is that the moisture level of the medium layer is higher than the

moisture level of the innermost and outer wraps. The explanation is that moisture passes from

the inner layer to the outer layer while heating and from the outer layer to the inner layer

during cooling. Therefore, there is constant (but might be different in strength for different

temperature gradients) flow through the medium layer, keeping moisture inside. Most prob-

ably, there will always be some minimum moisture level inside the medium wrap, even for

prolonged heating or cooling phases.

Overall, it can be seen that an increasing number of cycles over the same period of time leads to moisture accumulation inside the paper (namely in the middle layer, which is of primary interest). The difference in the moisture level inside the paper for a different number of cycles is not very drastic. However, the experiment itself is conducted over the span of 12 days, while the transformers are designed for 30-40 years of service. Therefore, one might assume that on a long time scale, transformers operating with a frequent number of start-ups and shut- downs will have a higher chance to experience insulation breakdown due to the effects of high moisture accumulation inside the paper isolation.

In Figure 13, the temperature profiles are quite different, together with the statement that four to six degrees temperature increase double the rate of degradation it becomes even more crit- ical [4]. The results may still reveal some general schemes but have to be treated carefully, keeping in mind that some effects may also be originated because of the different tempera- tures.

Some additional issues regard the “cold” of the temperature patterns. The used time constant for diffusion τ

is calculated based on a temperature of 40

C as shown in (2), which does not match the temperature that can be seen in the colder stages (and therefore the τ

be- comes larger).

Although these issues, the measurement procedure works well: no significant difference is

found among the different measurements of a single sample, which yields a possible assump-

tion that a random error (but not a systematic one) can be assumed to be absent. Furthermore,

a way to handle data has been presented. Unfortunately, the validity of this way cannot be

proven, as the input data might already be erroneous.

1224364860728496108120132144156168180192204216228240252264276 20 40 60 80 100 120 140Temperature Group 1 (^◦C)

Sampletemperaturefordifferentloadpatterns(5minaverage).S1...S2...S3...S4...S5 1224364860728496108120132144156168180192204216228240252264276 20 40 60 80 100 120 140Temperature Group 2 (^◦C)

01224364860728496108120132144156168180192204216228240252264276 20 40 60 80 100 120 140

Time(h)

Temperature Group 3 (^◦C)

Figure 13: Temperature pattern (5 min average)

6. Conclusion

This report shows the procedure for testing the effect of different heating cycles on the mois- ture migration of transformer insulation. The procedure to be followed is internally valid and may be used as a benchmark. The results should be further evaluated, and the procedure must be repeated to draw generalizable conclusions.

The results described in this report show that some samples have been exposed to overheating, causing physical damage to the test sample. Even though the duration of the experiment is shortened, it is possible to observe that the samples exposed to higher numbers of cycles are the ones with the higher moisture content in the middle layer of the insulation.

It has to be added that the dynamic transformer rating effect on transformer insulation is a new

research topic, and therefore, it is difficult to draw conclusions on this test. Continuing the

research on this topic is important since it would allow gaining additional knowledge about

transformer thermal limits.

References

[1] Christos Stefanou, “Investigation of the effect of moisture in transformers on the aging of the solid insulation for dynamic rating applications,” Master’s thesis, KTH Royal Institute of Technology, 2018.

[2] Coulometric KF Titrator C10/C20S/C30S: Operating Instructions, Mettler-Toledo GmbH, Analytical, Dec. 2015.

[3] Patrik Gustafsson, “Design of Experimental Setup for Investigation of Effect of Moisture Content on Transformer Paper Ageing during Intermittent Load,” Master’s thesis, KTH Royal Institute of Technology, 2018.

[4] U.S. Department of the Interior, Bureau of Reclamation, “Transformers: Basics, Main- tenance, and Diagnostics,” Hydroelectric Research and Technical Services Group, 2005.

[5] L. E. Lungaard, W. Hansen, D. Linhjell, T. J. Painter, “Ageing of Oil-Impregnated Paper in Power Transformers,” IEEE Transaction of Power Delivery, vol. 19, no. 1, pp. 230–

239, Jan. 2004.

[6] G. C. Stevens, A. M. Emsley, “Kinetics and Mechanics of the low temperature degrada- tion of cellulose,” Cellulose 1, pp. 26–56, 1994.

[7] L. Cheim, D. Patts, T. Prevost, S. Xu, “Furan Analysis for Liquid Power Transformers,”

IEEE Electrical Insulation Magazine, vol. 28, no. 2, pp. 8–21, 2012.

[8] H. C. Sun, Y. C. Huang, C. M. Huang, A Review of Dissolved Gas Analysis in Power Transformers. Elsevier, 2011.

[9] S. V. Kulkarni, S. A. Khaparde, Transformer Engineering: Design and Practice. CRC Press, 2004.

[10] (2020) Transformer Oil Deterioration. Globe Core. Retrieved 2020-09-18. [Online].

Available: https://globecore.com/transformer-oil-deterioration/

[11] T. V. Oomen, T. A. Prevos, “Cellulose Insulation in Oil-Filled Power Transformers: Part II – Maintaining Insulation Integrity and Life,” IEEE Electrical Insulation Magazine, vol. 22, no. 2, pp. 5–14, 2002.

[12] L. Lewand, “Understanding Water in Transformer Systems,” Neta World, 2002.

[13] IEC 60814: Insulation liquids – Determination of water by automatic colulometric Karl Fischer titration, IEC Std., 1997.

[14] Y. Du, M. Zahn, B. Lesieutre, A. Mamishev, and S. Lindgren, “Moisture equilibrium transformer paper-oil system,” Electrical Insulation Magazine, IEEE, vol. 15, pp. 11–

20, 02 1999.

[15] (2020) Temperature Rise and Transformer Efficiency. Copper Development Association In. Retrieved 2020-12-14. [Online]. Available: https://www.copper.org/environment/

sustainable-energy/transformers/education/trans_efficiency.html

[16] P. F. Ast, “Movement of Moisture through A50P281 Kraft Paper (Dry and Oil Impreg- nated),” General Electric, 1966.

[17] E. K. Steele, “Moisture Redistribution in Simulated Transformer,” General Electric, 1970.

[18] S. D. Foss, “Power Transformer Drying Model,” General Electric Company, Large Transformer Operation, 1987.

[19] (2020) Specific Heat Capacity of Metal Table Chart. Engineer Edge LLC. Retrieved 2020-09-19. [Online]. Available: https://www.engineersedge.com/materials/specific_

heat_capacity_of_metals_13259.htm

[20] Dmitri Kopeliovich. (2012, Jun.) Mineral transformer oil. SubsTech. Retrieved 2020- 09-19. [Online]. Available: https://www.substech.com/dokuwiki/doku.php?id=mineral_

transformer_oil

[21] (2012, Apr.) Super Grade Nytro 10XN Maximum performance insulating oil.

Nynas AB. Retrieved 2020-09-19. [Online]. Available: https://www.nynas.com/de/

product-areas/transformer-oils/oils/nytro-10xn-iec/

[22] (2020) Insulation paper for power transformers. Ahlstrom- Munsjö. Retrieved 2020-11-05. [Online]. Available: https:

//www.ahlstrom-munksjo.com/products/insulation-paper-and-specialty-pulp/

electrotechical-paper/insulation-paper-for-power-transformers/

[23] (2020) Loctite EA 3430 – Structural Bonding Adhesive – Epoxy. Henkel Adhesives.

Retrieved 2020-11-22.

[24] N. instrument, Installation Guide BNC-2120,Connector Accessory for E/M/S/X Series Devices, 2012.

[25] Good Titration Practice in Karl Fischer Titration, Mettler-Toledo GmbH, Analytical, Sep. 2009.

[26] Introduction to Karl Fischer Titration, Mettler-Toledo GmbH, Analytical, Aug. 2012.

[27] Taking Samples for Karl Fischer Titration, Mettler-Toledo GmbH, Analytical, Aug.

2012.

[28] Sample Preparation for Karl Fischer Titration, Mettler-Toledo GmbH, Analytical, Aug.

2012.

A. M ATLAB scripts

A.1. Improved version

1

2 %% Moisture samples plotting

3 % Following order is performed: first measurment, second measurement, third

4 % measurement, average of them. ’p’ - paper, ’o’ - oil, ’w’ - weight, ’m’ -

5 % moisture

6 % ’B1’ - innermost layer measurements; ’B2’ - medium & outermost layer measurements

7

8 %% Group one paper - oil

9 % First paper batch

10

11 G1_pw_B1_meas = [1.6035 1.6918 1.1970 1.4974; % S1

12 1.2400 1.2253 1.3724 1.2792; % S2

13 1.4127 1.7450 1.6182 1.5920; % S3

14 1.4905 1.3655 1.7777 1.5446; % S4

15 1.5020 1.6376 1.4424 1.5273]; % S5

16

17 G1_pm_B1_meas = [478.9 481.2 501.7 487.3; % S1

18 513.6 518.0 526.5 519.4; % S2

19 200.5 203.8 241.3 215.2; % S3 pure 500

20 1151.0 1161.8 1209.9 1174.2; % S4 pure 1600

21 490.5 497.4 489.6 492.5]; % S5 pure 500

22 23 % Oil

24

25 G1_ow_B1_meas = [1.8003 0.9832 1.0590 1.2808; % S1

26 1.9352 1.7550 1.6081 1.7661; % S2

27 1.8347 1.6773 0.6028 1.3716; % S3

28 1.7146 1.6489 0.9155 1.4263; % S4

29 1.6046 0.9695 1.3720 1.3154]; % S5

30

31 G1_om_B1_meas = [12.5 16.9 12.7 14.8; % S1

32 14.1 12.8 14.1 13.7; % S2

33 15.7 14.5 0 15.1; % S3

34 9.40 9.40 8.70 9.20; % S4

35 11.5 7.10 23.7 14.1]; % S5

36

37 % Second paper batch

38

39 G1_pw_B2_meas = [1.7265 1.4189 1.7151 1.6202; % S1

40 1.6321 1.4883 1.6333 1.5882; % S2

41 1.7367 1.3709 1.6470 1.5849; % S3 pure 500

42 1.5764 1.6263 1.5724 1.5917; % S4

43 1.4671 1.6478 1.6501 1.5883]; % S5 pure 500

44

45 G1_pm_B2_meas = [490.6 476.8 531.5 499.6; % S1

46 462.2 458.3 472.1 464.2; % S2

47 529.1 541.6 551.6 540.8; % S3

48 601.8 557.2 542.4 567.1; % S4

49 563.4 577.0 614.8 585.1]; % S5

50 51

52 % Group 2 paper - oil

53 %First paper batch

54

55 G2_pw_B1_meas = [1.5435 1.6663 1.3795 1.5298; % S1

56 1.3445 2.5308 2.3310 2.0688; % S2

57 1.0296 1.4544 1.2290 1.2377; % S3

58 1.7268 1.5232 1.0717 1.4406; % S4

59 1.6587 1.8715 1.2121 1.5808]; % S5

60

61 G2_pm_B1_meas = [264.9 259.2 262.3 262.1; % S1 pure 500

62 779.4 778.7 810.1 789.4; % S2 pure 500

63 400.0 400.3 405.2 401.83; % S3 pure 500

64 1720.9 1730.6 1723.6 1724.9; % S4 pure 1600

65 902.3 914.6 920.7 912.5]; % S5

66 67 % Oil

68

69 G2_ow_B1_meas = [1.9360 1.7177 1.6947 1.7828; % S1

70 1.6427 1.7376 1.7383 1.7062; % S2

71 1.7142 1.7223 1.4303 1.6222; % S3

72 1.2077 1.5796 1.5008 1.4294; % S4

73 1.5355 0.7933 0.7508 1.0265]; % S5

74

75 G2_om_B1_meas = [21.2 18.4 21.9 20.5; % S1

76 18.3 17.4 15.7 17.9; % S2

77 13.7 13.3 15.0 14.0; % S3

78 17.5 16.7 16.1 16.8; % S4

79 10.8 10.1 6.50 9.10]; % S5

80

81 % Second paper batch

82

83 G2_pw_B2_meas = [1.5152 1.5782 1.6735 1.5890; % S1

84 1.5238 1.5083 1.6531 1.5617; % S2

85 1.6993 1.6809 1.4393 1.6065; % S3

86 1.6277 1.5630 1.5719 1.5875; % S4

87 1.6869 1.6340 1.4700 1.5970]; % S5

88

89 G2_pm_B2_meas = [583.7 578.5 615.8 592.7; % S1

90 801.7 836.9 917.3 852.0; % S2 pure 500

91 440.8 437.7 440.4 439.6; % S3 pure 500

92 789.0 795.6 784.7 789.8; % S4 pure 500

93 583.0 575.7 579.6 579.4]; % S5

94

95 % Group 3 paper - oil

96 % First paper batch

97

98 G3_pw_B1_meas = [1.9291 1.5688 1.2597 1.5859; % S1

99 2.0363 1.5526 1.7608 1.7832; % S2

100 1.8963 1.5577 1.1418 1.5319; % S3

101 1.7728 1.5158 1.6268 1.6385; % S4

102 1.7022 1.5561 1.5035 1.5873]; % S5

103

104 G3_pm_B1_meas = [392.7 400.2 406.3 399.7333; % S1

105 911.0 910.6 917.0 912.8667; % S2

106 720.3 719.1 711.8 717.0667; % S3

107 1525.5 1519.8 1520.6 1521.6333; % S4

108 1487.8 1496.8 1492.4 1492.3333]; % S5

109 110 % Oil

111

112 G3_ow_B1_meas = [2.0129 1.7291 1.5461 1.7627; % S1

113 1.8929 1.8122 1.5035 1.7362; % S2

114 1.8377 1.6813 1.6582 1.7257; % S3

115 2.0140 1.3456 1.1282 1.4959; % S4

116 2.1588 1.8818 1.6754 1.8818]; % S5

117

118 G3_om_B1_meas = [9.10 8.30 10.2 9.30; % S1

119 10.6 10.3 7.90 9.60; % S2

120 17.5 14.2 13.1 14.93; % S3

121 11.5 10.5 14.5 12.2; % S4

122 15.9 11.6 12.6 13.4]; % S5

123

124 % Second paper batch

125

126 G3_pw_B2_meas = [1.7480 1.5395 1.5458 1.6111; % S1

127 1.7705 1.5110 1.6783 1.6533; % S2

128 1.7408 1.5591 1.4088 1.5696; % S3

129 1.6393 1.6451 1.4560 1.5801; % S4

130 1.7119 1.6105 1.4758 1.5994]; % S5

131

132 G3_pm_B2_meas = [424.3 424.2 460.8 436.4; % S1

133 438.8 435.5 437.3 437.2; % S2