Sei sulla pagina 1di 7

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO.

3, JULY 2013

1433

Measuring Paper Water Content of Transformers: A New Approach Using Cellulose Isotherms in Nonequilibrium Conditions
Daniel Martin, Member, IEEE, Caesar Perkasa, and Nick Lelekakis
AbstractIn this paper, we discuss a better method to measure the paper water content of transformers using cellulose isotherms. The advantage of the new method is that the transformer oil and cellulosic insulation do not need to be in thermodynamic equilibrium. Instead, the temperature needs to be cycling in order to determine when the direction of water migration, whether out of or into the oil changes. The reversal in water migration can be used to determine the vapor pressure of the water adsorbed by the cellulose. This method was tested using 12 different conditions similar to those within a normally operating transformer. The results showed close agreement when this method is used. The signicance is that isotherm equations can now be used even when insulation media are not in equilibrium. Index TermsMoisture measurement, oil insulation, paper insulation, power transformer insulation, power transformers.

Two models that are used by the industry are those by Piper [5] and Fessler [6]. Both of these equations use temperature and vapor pressure to derive the paper water content. The Fessler equation is given (1) where is the concentration of absorbed water as the ratio of the mass of water to the mass of dry cellulose, is the vapor pressure of water in atm, and is the temperature in Kelvin. Within the context of this paper, we have dened , calculated using the Fessler equation, as the apparent water content of paper, because under changing temperatures, the calculated water content will reect the conditions of the oil. Some time is required for to equal the true paper water content, when sorption equilibrium has been achieved. One paper on using cellulose isotherms to estimate transformer cellulose wetness from the oil water content was written by Oommen in 1984 [7]. However, the main drawback is that the different insulation media must be in thermodynamic equilibrium. If they are not, then an inaccurate determination will result. Du investigated the time constants involved for pressboard to reach equilibrium with the surrounding mineral oil [8], [9]. The equation to calculate the time constant for diffusion of water through both sides of cellulose, as a function of water content and temperature, is given [9] (2) where is the thickness of the cellulose and is the diffusion coefcient. The diffusion time constants for 1-mm-thick mineral oil pressboard are reported to be 333 h at 20 C and 6 h at 70 C [9]. The long periods that are required demonstrate that a transformer, in operation, is unlikely to be at the same temperature long enough for equilibrium to occur. One major step forwards with measuring water was the introduction of online water sensors. The water content of a transformer could then be monitored in real time, rather than periodically using a Karl Fischer measurement. Koch proposed using long-term averages of the oil relative water saturation and temperature, with cellulose isotherms, to calculate the mass of water in paper [3]. However, a potential drawback to using averages is that this method does not take into account changing diffusion coefcients. The diffusion coefcients change with temperature, and, the diffusion coefcient for when the cellulose is desorbing

I. INTRODUCTION ELLULOSE isotherms have been used by the utilities to estimate the water content of paper for decades [1]. The vapor pressure of water, dissolved in a volume of oil, is used to determine the concentration of water adsorbed by cellulose. Excessive water can lead to precipitation in the oil causing dielectric failure [2], and causes an increasing rate of aging of cellulosic insulation [1]. Cellulose isotherms only provide an estimation of paper water content because the type and age of the cellulosic material affect the isotherm relationship [3]. Some relevant errors to using equilibrium diagrams, given in [4], are a steep gradient in the low moisture region complicating the reading, scattered results using different equilibrium charts, and temperature gradients within windings not being considered. However, the advantage of using this method is that it is easy to integrate into an online transformer monitoring system. Therefore, if some of the relevant errors can be minimized, this will provide a step forward for the industry.
Manuscript received April 26, 2012; revised August 22, 2012, October 21, 2012, and January 13, 2013; accepted February 14, 2013. Date of publication April 23, 2013; date of current version June 20, 2013. This work was supported in part by Ausgrid, Citipower/powercor, Cooper Power Systems, Dynamic Ratings, Energex, Ergon Energy, FM Global, Powerlink Queensland, TJ H b Analytical Services, Transgrid, Transpower New Zealand, UK PowerNetworks, and Wilson Transformer Company. Paper no. TPWRD-00430-2012. The authors are with Monash University, TransformerLIFE Centre, Clayton VIC 3800, Australia (e-mail: daniel.martin@monash.edu; Nick. lelekakis@monash.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRD.2013.2248396

0885-8977/$31.00 2013 IEEE

1434

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 3, JULY 2013

water is two to three times higher than when the cellulose is adsorbing water [10]. We set out to evaluate the validity of using average conditions to determine cellulose water content, studying the migration of water between cellulose and oil at a constant temperature and when the temperature is cycling. II. SORPTION AND EQUILIBRIUM Equilibrium sorption is when the rate of attachment of vapor molecules balances their rate of loss from the sorbing material [11]. When these rates are not equal, there is a net migration of water either into or out of the cellulose. The rate of change of oil water concentration is given in (3), where in ppm changes proportionally to the net rate of migration . The constant is related to the volume of oil and surface area of cellulose (3) Janssen investigated the diffusion of water vapor in porous materials and used a Fick equation to relate the ow rate to the vapor pressure gradient (4), [12] (4) Consequently, when ow rate is zero, there is also zero vapor pressure gradient between the surface of the cellulose and sensor tip. In practice, there may be a temperature gradient between the cellulose and sensor tip. However, within the context of this system, such a gradient was minimal. Our method is based on the assumption that as the temperature of a transformer is cycling, the ow of water between the insulation media oscillates around the sorption equilibrium position. When the ow is zero, there is no vapor pressure gradient between the online probe and cellulose. III. APPLICABILITY OF FICK EQUATIONS TO CELLULOSE The equilibrium sorption discussed previously only considers the movement of water between the surface of the cellulose and the surrounding uid medium. Water also migrates through the cellulose block. The applicability of using the Fick equations to model movement of water through paper has been discussed by Nyman, who notes that using these equations is valid within the relative humidity range 0%60%, after which a strong nonlinear dependency is observed [13]. When the rate of change of cellulose surface water is equal to zero, then the water concentration gradient between the surface and bulk cellulose should also be zero. This equality assumes that the water within the block can easily migrate to the surface, and is not trapped inside. IV. ERROR CAUSED BY HYSTERESIS The moisture content of wood not only depends on temperature and water vapor pressure, it also depends on its immediate history (i.e., the direction of change of relative humidity). Early examples of hysteresis are given by Ramarao [14] and Haslach [15], who comment on previous work by Seborg [16], Wink [17], and Stamm [18]. The actual values reported were 2.2%2.7% at 10% relative humidity and 4.8%5.7% at 30%

Fig. 1. Vessel used for investigation, showing how cellulose was immersed in the vegetable oil.

relative humidity [15]. Work published by Chatterjee shows a similar, but slightly smaller, hysteresis [19]. At 30% relative humidity, at 23.8 C, the equilibrium moisture content of bleached Kraft paperboard was 6.23% under adsorption conditions and 6.89% when desorbing. Despite a thorough literature review, little was found on measurements relating to hysteresis when the temperature of cellulose is cycled. Investigations have tended to focus on isothermal conditions with the relative humidity cycled, rather than the humidity cycled by changing temperature. We therefore decided to assume that hysteresis had a minimal impact on our model, until such research becomes available. V. EXPERIMENT A. Equipment Setup The vessel was set up to include a paper and pressboard sample holder, water activity probe, and thermometer to control the hotplate, shown in Fig. 1. A magnetic stirrer bar was placed in the oil. The water activity probe (Vaisala MMT 330) was used to monitor the water content and temperature of the vegetable oil. The Buck equation (5) [20] was used to calculate the vapor pressure of water, for (1), from the water activity measurement and temperature in degrees Celsius. Water activity is dened as the ratio of the partial pressure of water vapor to the partial pressure of water vapor above pure water at the same temperature [21]. A conversion factor of was included to change the units from hectopascals into atmospheres, required for use with (1). (5) In order for the probe measurement to be accurate, its sensing tip must be in thermodynamic equilibrium with the surrounding uid. Since we are using sinusoidal temperature waveforms, the system will not reach equilibrium. However, as long as the response time of the sensor is relatively short in comparison to the rate of change of water within the system, we regarded the validity of using this probe to be correct. Three different vessels were set up to run in parallel, each vessel had a different starting water content. Since hydrocarbon-based materials generate water during oxidation, we minimalized oxygen availability. The oil was vacu-

MARTIN et al.: MEASURING PAPER WATER CONTENT OF TRANSFORMERS

1435

Fig. 2. The 2 1-mm strips of pressboard and 40 sheets of paper were used in this investigation. In total, the cellulose had a mass of 23 g.

umed, removing air gases, and the conservator was lled with inert argon. B. Calibrating the Water Activity Probe The output from the water activity probe is converted into parts per million (ppm) of water dissolved in this uid using (6) Coefcients A and B are dependent on both the type of oil and its age; old oil can be more hygroscopic than when it is new [22]. is the temperature in Kelvin. The method used to determine solubility was similar to the one utilized by Lewand [23]. The oil was heated in steps when samples were removed for water measurement using Karl Fischer titration. The temperature and water activity at the time of sampling were recorded and, thus, the solubility at that temperature was calculated. We found A and B to be 6.000 and 881.39, respectively, similar to 5.3318 and 684 reported for this uid by Lewand in [23]. When using water activity probes with different transformers, we recommend determining its oil A and B solubility coefcients. C. Conditioning Materials We used 23 g of cellulose, comprised of two pieces of 1-mm-thick pressboard and 40 sheets of paper. The cellulose was screwed to a plug, shown in Fig. 2, and then suspended in one liter of oil. The mass of solid insulation to volume of oil ratio was 0.023 kg/l. We conditioned three sets of cellulose to have a water content similar to that of in-service transformer insulation, one with water content comparable to that of a new transformer ( 0.5% [2]), one similar to a wet transformer near its end of life (5%6% [24]), and one somewhere in between. The dry cellulose was conditioned in an oven heated to 110 C for 2 h. The cellulose was allowed to cool overnight in a desiccator, kept dry with silica gel. The oil used was degassed by placing a bottle on a hot plate and heating the oil to 80 C overnight under vacuum. The cellulose and oil were quickly transferred to the heating vessel, which was left to stand at 40 C for one week, to give sufcient time for the water to equilibrate between the media. To check that the system was very close to equilibrium, we plotted a trendline of the oil water content over the last 4 h that the system was held at a constant 40 C. In equilibrium, the gradient of the oil water content will equal zero. The slope of the trendline was extremely small, a calculated change of ppm/s over 4 h, and probably within the accuracy of the system. We therefore took the assumption of being in equilibrium as valid. For the wet system, neither the cellulose nor oil were dried before use. The cellulose and oil were transferred to the vessel
Fig. 3. Effect of varying the spline smoothing parameter on the paper water content calculation.

and left to stand at 40 C for one week. For the medium wetness system, the oil was not dried and the cellulose was heated in the oven similar to the dry sample. The cellulose was left to cool outside the desiccator. It was then transferred into the vessel and left to stand at 40 C for one week. D. Temperature Prole We selected temperatures similar to those on which a transformer operates. The IEEE standard for transformers C57.12.00 advises that the temperature rise of the insulating liquid shall not exceed 65 C when measured near the top of the main tank [25]. A program written in Microsoft Visual Basic was developed to continually change the temperature setting of the hotplate via an RS232 link to replicate a sinusoidal temperature prole. The smallest unit that the hotplate could be changed by was 1 C, and the temperature was changed hourly. E. Smoothing Algorithm The mass of water data was processed using a smoothing spline to remove noise. This ltering removes the noise which would affect the differentiation of the signal. The smoothing was accomplished using the curve-tting toolbox in Matlab. A spline requires a smoothing parameter to remove noise from a data set. The sensitive range for the smoothing parameter was calculated using the method proposed by de Boor [26], [27] where (7)

where represents the values in the dataset and is the number of data points in the array. The data were processed using a smoothing spline with the smoothing parameter set to 3.6e-09. This was found to give satisfactory results in all cases. Since there was a certain level of operator skill required to tune the smoothing spline, we investigated the consequence of changing the spline parameter around the value we had chosen. Focusing on the area of interest when the rate of change of oil water content is zero (Fig. 3), varying the spline parameter did not signicantly affect the paper water content calculation. Using the Fessler equation at the time when the derivative for oil water content is zero did not give a signicantly different result when the spline smoothing parameter was varied around the area of interest.

1436

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 3, JULY 2013

TABLE I TEMPERATURE AND AVERAGE CELLULOSE WATER CONTENT OF THE SYSTEM

TABLE II COMPARISON OF CELLULOSE WATER CONTENT

VI. RESULTS AND DISCUSSION A. Temperature and Average Moisture Content The conditions of each test are shown in Table I. First, the system was heated to a constant temperature for one week to facilitate equilibrium. The Fessler equation was used to calculate the equilibrium moisture content (EMC), assuming that equilibrium had been reached. Second, the system was heated with a sinusoidal temperature prole, with an average temperature equal to that of the previous step, for one week. Third, the system was held at the average temperature for another week to determine whether the EMC of the cellulose had changed. The average water content was calculated using measured temperature and water activity averaged over the week period. As can be seen, in most cases, the EMC at the end of the proles was slightly higher than that at the beginning. Since there is some hysteresis involved in the moisture content of cellulose reaching its equilibrium value, the difference may have been caused by this, although further work would be required to be certain. The accuracy of the Fessler equation was investigated by measuring the water content of paper, using Karl Fischer titration, and comparing it to the calculated value, shown in Table II. The titration was performed according to procedure 4.4 of IEC 60814 [28], where a specimen of impregnated paper is heated in an oven next to the Karl Fischer apparatus, and the evolved water is transferred into the titration vessel. Since verication requires that the cellulose is in sorption equilibrium with the oil, the system was given some time to stand on each temperature step. We found that the calculated water content was close to, although slightly higher than, that measured. Each measurement required the vessel to be opened with the cellulose being exposed to the air for a short period while a sample was cut from the paper. Thus, a small quantity of moisture would have been introduced on each sampling. However, the measured moisture contents are all less than calculated, suggesting this error was minimal. Koch investigated the isotherms of new and thermally degraded Kraft paper at temperatures between 21 and 80 C [29]. As the water content of paper rises above 1%, by dry mass, the isotherms begin to diverge dependent on both temperature

Fig. 4. Calculated water content of oil and cellulose.

and age. Thus, modeling with one equation brings in an error. Below 1% by dry mass, the gradient of the moisture in paper versus moisture relative to saturation is high, and a small change in water activity will bring about a signicant change in paper moisture content. The problem this brings is that it is difcult to establish a precise relationship between water activity and mass of water at lower levels. B. Water Dynamics During the Cyclic Temperature Prole For the cycling temperature as the equilibrium sorption position is passed, the direction of water ow changes. The equilibrium point being when the rate of adsorption is equal to the rate of desorption, or . The temperature was cycled between 30 and 60 C at a period of 1 day, similar to typical transformer conditions. The vapor pressure also cycles, being a function of the temperature of and the mass of water dissolved in the oil. When the rate of change of oil water content is equal to zero, occurring twice every temperature cycle, the vapor pressure gradient is zero (4). Fig. 4 shows the oil water content varying as the temperature is changing and the water content of the paper, the dashed trace, was calculated using (1). The crosses on the dashed trace denote when the net ow of water reverses direction, either into or out of the oil, which is when the insulation media are passing their equilibrium point. As can be seen in Fig. 4, the cellulose water content, determined at the equilibrium point, differs depending on whether the oil water content has peaked or bottomed. Even though equilibrium time constants are longer than 24 h at these temperatures, some water still migrates between the oil and surface of the cellulose, which, in this case, was approximately 15 ppm.

MARTIN et al.: MEASURING PAPER WATER CONTENT OF TRANSFORMERS

1437

Fig. 5. Calculated water content of paper compared to measurements made by Karl Fischer titration.

Fig. 6. Paper water content for 30 to 50 C.

This mass being adsorbed and desorbed by the cellulose surface causes the vapor pressure to differ when . To verify that the water content of the paper does not significantly change during one temperature cycle, we repeated the 3070 C prole and measured the paper water content at the beginning in steady state, at a trough, at a peak then again in steady state. The paper water content was measured using Karl Fischer titration. The system was heated to 50 C for one week then cycled. Despite the calculation, using the Fessler equation, showing a result higher than that measured by Karl Fischer titration (Fig. 5), the graph shows that the actual water content of the paper does not change signicantly during the cycle. This implies that the variation in paper water content during one temperature cycle is only apparent, and is not physically occurring. This is expected as the mass migrating between the surface of the cellulose and oil is small. The fall in wcp kf could be due to differing time constants for adsorption and desorption; more water leaves the paper during the hotter temperatures than is readsorbed during the lower ones. If the mass of water moving in and out of the oil during one temperature cycle is , the water content of the cellulose is cycling between , where is the average water content of paper. We therefore took the average of the two points when . C. Comparison of Multiple Temperature Ranges and System Water Content The same methodology was applied to a variety of different temperature and water content conditions. The results are shown in Figs. 68. The dry and medium wetness systems agreed with the calculations for water content using average temperature and water activity. The cellulose water content is based on the surface conditions of the cellulose. Thus, a change can indicate that water is being transferred between the inside and surface of the cellulose. When cellulose is in sorption equilibrium with the surrounding oil, and its temperature is increased causing desorption, water moves from inside the cellulose to its surface and then into the oil. The dispersion of data in the wet system is high, reecting changing surface water content of cellulose. Under these dry

Fig. 7. Paper water content for 30 to 60 C.

Fig. 8. Paper water content for 30 to 70 C.

conditions, water forms a monolayer on the cellulose surface. The thickness of this monolayer is extremely small, only one molecule wide. The overall volume of water being transferred will therefore be much less than the mass of water being transferred when the water content of the entire cellulose block changes. VII. APPLICATION TO AN OPERATING TRANSFORMER The team studied the online dryout of two transformers by a utility [24]. The oil measurement data, along with the water

1438

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 3, JULY 2013

Fig. 9. Temperature and water activity data for a mineral oil-lled transformer, with the measuring sensor installed at the top of the tank.

Fig. 11. Temperature and water activity data for a mineral oil-lled transformer, with a measuring sensor installed at the top of the tank.

Fig. 10. Paper water content at the top of the transformer, using data from Fig. 9, obtained using the new method.

Fig. 12. Paper water content at the top of the transformer, using data from Fig. 11, obtained using the new method.

content of the paper, is shown in Figs. 9 and 10. In order to correlate the results for this method to that of another technology, we used an Omicron Dirana unit, which uses dielectric response [30]. At 0 h, the Dirana recorded a paper water content of 5% and at 3550 h 4%. The rst measurement correlates with that produced by the algorithm. The second measurement recorded by the Dirana was higher. This difference may have been due to the dielectric response method detecting water absorbed inside the cellulosic insulation, whereas the use of water activity sensors only reects the water on the surface of the cellulose. The online dryout method removes water from transformer insulation by drying the oil. The water trapped inside the cellulosic transformer pressboard does not readily move out of the pressboard, compared to water moving from the surface of the cellulose into the oil. Therefore, water is more concentrated within the pressboard than on its surface, which is reected by the differences in measurement technique. In Fig. 9, the increase in water activity at 3000 h was due to the online dryout unit being switched off, and the water migrating out of the cellulose into the oil. Adding a trendline to the series for both temperature and water activity provides their respective long-term average. Taking the start and end values of the long-term averages to calculate paper water content derives a value of 2.5% at the start and 2.1% at the end. Taking the long-term average has resulted in the initially higher oil water activity being smoothed out.

In another project, the team investigated a 10-MVA 22/6.6-kV transformer, also located in Australia. A Vaisala water activity probe was inserted into the top of the transformer tank. The oil temperature and water activity over a 17-day period are shown in Fig. 11. The paper water content is shown in Fig. 12. The water content results are fairly consistent over this time period. The paper water content determined using long-term averages for temperature and water activity varies from 3.4% at the start to 4.1% at 400 h, close to the values in Fig. 12. Unlike the prole in Fig. 9, there are no spikes in temperature. Using long-term averages for temperature and water activity may provide an estimate of the bulk cellulose water content. However, its use will not detect uctuations in the surface water content of cellulose, which can lead to water precipitating in the oil. VIII. CONCLUSION Although diffusion time constants are long, some water can quickly migrate between the surface of the cellulose and oil. A small quantity of water migrating into the oil can cause a relatively large change in its oil water activity, which then affects the determination of cellulose water content. Taking long-term average oil conditions provides a reasonable estimate of cellulose water content. However, water can move relatively quickly between the surface of the cellulose and the oil. Taking an average will not detect this movement. A large movement of water could be dangerous if the local oil

MARTIN et al.: MEASURING PAPER WATER CONTENT OF TRANSFORMERS

1439

volume becomes saturated and precipitation occurs. A dynamic model, which determines the water content of paper based on daily measurements, is superior to the use of a long-term average because sudden changes in the partition of water between the cellulose and oil are actually detected, and are not smoothed out by an averaging function. The values measured using Karl Fischer titration did not equal exactly those determined by the Fessler equation, in agreement with Kochs observation that age and temperature affect the isotherm relationship. Taking Karl Fischer measurements at peak and trough temperatures indicated little change in water content over the cycle. This supports attempts to calculate cellulose water content by assuming that this does not signicantly change during the cycling temperature. ACKNOWLEDGMENT The authors would like to thank P. Cole and M. Gibson, of Ausgrid, and C. Feely, of Citipower/Powercor, for providing the data for algorithm development. REFERENCES [1] T. V. Oommen and T. Prevost, Cellulose insulation in oil-lled power transformers: Part IIMaintaining insulation integrity and life, IEEE Elect. Insul. Mag., vol. 22, no. 2, pp. 514, Mar./Apr. 2006. [2] T. V. Oommen and S. R. Lindgren, Bubble evolution from transformer overload, in Proc. IEEE TD Conf. Paper, 2001, pp. 137142. [3] M. Koch, S. Tenbohlen, and T. Stirl, Diagnostic application of moisture equilibrium for power transformers, IEEE Trans. Power Del., vol. 25, no. 4, pp. 25742581, Oct. 2010. [4] Moisture equilibrium and moisture migration within transformer insulation systems, CIGRE A2-30 Brochure (Ref. No. 349) Jun. 2008. [5] J. D. Piper, Moisture equilibrium between gas space and brous materials in enclosed electric equipment, AIEE Trans., vol. 65, pp. 791797, 1946. [6] W. A. Fessler, O. Rouse, W. J. McNutt, and O. R. Compton, A rened mathematical model for prediction of bubble evolution in transformers, IEEE Trans. Power Del., vol. 4, no. 1, pp. 391404, Jan. 1989. [7] T. V. Oommen, Moisture equilibrium in paper-oil insulation systems, in Proc. Elect./Electron. Insul. Conf., 1983, pp. 162166. [8] Y. Du, M. Zahn, B. C. Lesieutre, A. V. Mamishev, and S. R. Lindgren, Moisture equilibrium in transformer paper-oil systems, IEEE Elect. Insul. Mag., vol. 15, no. 1, pp. 1120, Jan./Feb. 1999. [9] Y. Du, Measurements and Modeling of Moisture in Transformer Insulation. Berlin, Germany: VDM Verlag Dr. Mller, 2009. [10] B. Time, Hygroscopic moisture transport in wood, Ph.D. dissertation, Dept. Bldg. Constuct. Eng., Norwegian Univ. Sci. Technol., Trondheim, Norway, 1998. [11] R. W. Dent, A multilayer theory for gas sorption, Part 1: Sorption of a single gas, Text. Res. J., vol. 47, pp. 145152, 1977. [12] H. Janssen, Thermal diffusion of water vapour in porous materials: True or false?, presented at the Nordic Symp. Building Phys., Tampere, Finland, May/Jun. 2011. [13] U. Nyman, P. J. Gustafsson, B. Johannesson, and R. Hgglund, A numerical method for the evaluation of non-linear transient moisture ow in cellulosic materials, Int. J. Numer. Meth. Eng., vol. 66, pp. 18591883, 2006. [14] B. V. Ramarao, Moisture sorption and transport processes in paper materials, in Studies Surface Sci. Catalysis. New York: Elsevier Science B. V., 1998, vol. 120, Adsorption and its Applications in Industry and Environmental Protection. [15] H. W. Haslach, The moisture and rate-dependent mechanical properties of paper: A review, in Proc. Mechan. Time-Depend. Mater. Conf., 2000, vol. 4, pp. 169210. [16] C. O. Seborg and A. J. Stamm, Sorption of water vapor by paper-marking materials IEffect of beating, Ind. Eng. Chem., pp. 12711275, 1931. [17] W. A. Wink, The effect of relative humidity and temperature on paper properties, Tappi, vol. 44, pp. 171A178A, 1961. [18] A. J. Stamm, Adsorption in swelling versus non-swelling systems, Tappi, vol. 40, pp. 761770, 1957.

[19] S. G. Chatterjee, Comparison of domain and similarity models for characterizing moisture sorption equilibria of paper, Ind. Eng. Chem. Res., vol. 40, pp. 188194, 2001. [20] A. L. Buck, New equations for computing vapor pressure and enhancement factor, J. Appl. Met., vol. 20, pp. 15271532, 1981. [21] Dictionary of Food Science and Technology, 2nd ed. Hoboken, NJ, USA: Wiley, 2009. [22] P. J. Grifn, C. M. Bruce, and J. D. Christie, Comparison of water equilibrium in silicone and mineral oil transformers, presented at the 68th Annu. Int. Conf. Doble Clients, Boston, MA, USA, 1988. [23] L. Lewand, Laboratory evaluation of several synthetic and agricultural-based dielectric liquids, presented at the Doble. Int. Client Conf., Boston, MA, USA, 2001. [24] N. Lelekakis, D. Martin, W. Guo, and J. Wijaya, A eld study of two on-line dry-out methods for power transformers, IEEE Elect. Insul. Mag., vol. 28, no. 3, pp. 3239, Sep./Oct. 2011. [25] IEEE Standard General Requirements for Liquid-Immersed Distribution, Power, and Regulating Transformers, IEEE Standard C57.12.001993, 1993, IEEE Power Eng. Soc. [26] MathWorks, Curve Fitting Toolbox Cubic Smoothing Splines. Feb. 15, 2012. [Online]. Available: http://www.mathworks.com.au/products/curvetting/demos.html?le=/products/demos/shipping/curvet/ csapsdem.html [27] C. De Boor, A Practical Guide to Splines. New York: Springer, 2001. [28] Insulating LiquidsOil-Impregnated Paper and Pressboard-Determination of Water by Automatic Coulometric Karl Fischer Titration, IEC 60814 Ed. 2.0, 1997. [29] M. Koch, S. Tenbohlen, and T. Stirl, Advanced online moisture measurements in power transformers, presented at the Int. Conf. Condition Monitor. Diagnost., Changwon, Korea, 2006. [30] Omicron Electronics GmbH., Dirana Product Brochure. Austria, 2011. Daniel Martin (M12) received the B.Eng. degree in electrical and electronic engineering from the University of Brighton, Brighton, U.K., in 2000. He joined Racal Electronics, London, U.K., which became the international electronics company Thales, working on communication and aircraft systems. He left Thales in 2004 to pursue the Ph.D. degree at the University of Manchester, Manchester, U.K., on the suitability of vegetable oils and synthetic esters as large power transformer dielectrics. In 2007, he joined Monash University, Clayton, Australia, where he now directs industry-leading research into the condition monitoring of transformers for the power industry.

Caesar Perkasa received the B.Eng. (Hons.) degree in electrical and computer systems engineering from Monash University, Clayton, Australia, in 2011, where he is currently pursuing the M.Eng.Sc. (Research) degree in transformer insulation. He is currently investigating the phenomena of bubble evolution in vegetable oil.

Nick Lelekakis received the B.Sc. degree (Hons.) in chemistry from Monash University, Clayton, Australia, in 1995. Since then, he has been with Monash University, conducting research on transformer-related projects, including moisture dynamics in paper-oil systems investigated by Karl-Fischer titration and moisture probes. He has more than 15 years experience in sampling, measuring, and monitoring gases dissolved in electrical insulating oil, using gas chromatography. He has also conducted extensive paper-aging experiments on the degradation of transformer insulation and oils. His interests lie in applying his chemistry background to monitoring, maintaining, and extending the life of power transformers.

Potrebbero piacerti anche