Part 1: Techniques for Drying Out Oil-Filled Transformers While Still In-Service
By Michel Bélanger
When it becomes necessary to dry out a transformer, several procedures are available. This 2-part article presents you with a technique that allows for the extraction of water from the windings while the equipment is in service.
The basic insulating materials used in the majority of power transformers are mineral oil and cellulose (KRAFT paper) based products.
When the cellulose is adequately impregnated with oil, it offers the user a material with insulating and mechanical properties of remarkable suppleness. The ready supply of cellulose and mineral oil has, therefore, made these the materials of choice for nearly a century. It is harder to find good quality crude oil since 1970 and industry in general is pushed to optimise its assets. As a result, more emphasis is put on preventive oil and paper maintenance.
Referring to Figure 1, the characteristics of insulation A depend on the properties of the oil, where as insulation B and C depend on a combination of the properties of the oil and cellulose. When the cellulose in B and C have degraded, the basic impulse level (BIL) and short circuit endurance are affected negatively and increase the risk of failure of the transformer.
Figure 2 illustrates the use of cellulose in a transformer insulating system. In this system, the cellulose serves two purposes: as a structural element, such as the tubes on which the primary and the secondary coils are wound and as an insulant, such as the paper wrapping each conductor.
Often, the coil turns are composed of several wires in parallel. Wrappings of two or three rows of paper 2-3 mm thick isolate each conductor. See Figure 3 for more details of the insulation.
Even if the use of cellulose necessitates maintenance, the combination of oil-cellulose, for the most part, is a reliable and economic insulant. Still, manufacturers are looking for a combination of materials that would be more stable at higher temperatures. Aramid paper (NOMEX¨) seems to be a promising material that could be used in a hybrid insulating system.
This system would use the aramid paper in the hottest areas of the windings with the cellulose used everywhere else. Real cases have shown the applicability of this technique .
Cellulose Insulation -- the Weakest Link in the Transformer
The weakest link of any transformer insulation system is the cellulose. In the majority of cases, the failure of a transformer is related to an insulation breakdown, and the reason for the failure is usually mechanical -- under the stress of physical forces, the insulation gives way.
The adjacent surfaces of the spires have to support high mechanical stress. This physical stress comes principally from continuous vibrations and short-circuits. Under the effect of a short-circuit, windings undergo enormous pressure. These forces cause vibrations in all parts of the windings.
If the cellulose is in good condition, the insulation will support the effect of mechanical stress, if not, the latter will be damaged or destroyed, causing a latent default or a transformer failure.
In general, when the insulation is damaged, the transformer has reached the end of its useful life. A rewind then becomes necessary.
Insulation Controls Transformer Life Span
Practically speaking, the life span of the transformer is really the life span of the cellulose insulation. Cellulose cannot be restored to new without a complete repair, unlike the oil which can be replaced as needed.
Survey statistics reveal a lot on this subject. Let's take the data from two survey cases . Both surveys conclude that the insulation is involved in a majority of cases of transformer failure.
Table 1 relates the failure rates between transformer components.
These insulation failures are often caused by a change of chemical bonds between the cellulose cells , where water is one of the principal deterioration agents. When this happens, the insulation system no longer has the necessary capacity to support the stress that is imposed. This will eventually lead to an irreversible breakdown, for example, under the constraint of an external short circuit.
Table 2 draws attention to the main reasons for failures. It indicates that the majority of problems come from manufacturing defects or inadequate maintenance. This table validates the need for: a) use of purchase specifications, b) start-up verification and testing and c) setting up a preventive maintenance programme.
It is therefore possible to optimise the management of transformers by paying particular attention to the cellulose insulation. This management will allow for a longer life expectancy of all the equipment owned.
Factors Affecting the Durability of the Cellulose
Cellulose is responsible in large measure for the behaviour of the insulation under voltage impulses (BIL), short circuit and shrinkage of the windings (control of the deformation due to compression forces of the clamping devices). 
Factors involved in the deterioration of cellulose are water, oxygen and pyrolysis (temperature).  These elements contribute to: the formation of water (the degradation of the cellulose generates water eventually), the degradation of mechanical properties of and reduction of the dielectric rigidity of the cellulose. The degradation of the paper is strongly correlated with the degradation mechanism of thermal origin.
The Achilles heel of cellulose comes from its affinity for water and the compounds of oil degradation. Cellulose is vulnerable to oxygen and to excessive heat. When these elements are present at the same time, the aging process is accelerated. For example, a 1 per cent water content will age the cellulose 10 times faster than if the water content was 0.1 per cent. If we combine the effect of oxygen in the presence of water, we can obtain an aging effect in the order of 25 times . This is why transformer manufacturers began making sealed transformers and transformers with expansion reservoirs equipped with a bladder. These technologies aim at countering the migration of the water and the oxygen towards the main tank.
In conclusion, aging of cellulose in transformers causes the following effects:
Practical recommendations to counter these effects are:
- reduction of the paper strength
- reduction of BIL withstand,
- reduction of short circuit withstand,
- reduction of the solidity of the windings
- reduce the temperature of the transformer either by leaving the cooling system running or by reducing the cooling set point (usually set by the manufacturer at 70°C);
- regularly conduct a comparative evaluation of the cooling capability of the radiators with the help of an infrared camera;
- reduce the water content in the windings;
- specify the maximum water content in the purchase specifications of the transformer. (Wait a few weeks after start up before verifying);
- verify leaks in the reservoir, silica gel circuit, gasket, etc.;
- maintain a minimum concentration of antioxidant.
Usually, the transformer leaves the factory with a water content which is a function of BIL and the operating voltage. For transformers where the operating voltage is less than 230kV and their design BIL level is not reduced, the suggested water content is 0.5 per cent. Table 3 illustrates how to determine if the transformer is operating under reduced BIL values. For transformers which are more than 500kV or where their BIL level is reduced, the water content should be lowered to 0.25 per cent.
The water content should be validated on site with the help of this chart and the humidity evaluation procedure explained in this article.
Astonishingly, the evaluation of the drying level of the windings is not a current practice at transformer start up, or directly afterwards. The evaluation is conducted solely on the dissolved water, in conditions where the obtained values are less credible. Two common errors are worth mentioning.
The first problem is measuring the concentration of water when the transformer is out of service or practically at ambient temperature. First, humidity homogeneity between the elements constituting the insulating system, including the oil and cellulose is never attained if the transformer is below 30°C. This is because the transformer temperature changes with ambient and at this low temperature the water is constantly chasing an equilibrium. Second, the low concentration of water which is normally found in a transformer operating near ambient temperature, implies large measuring errors. For example, at 30°C, 0.5 per cent humidity, the concentration of water in the oil is inferior to 1 ppm. The water concentration measurement could easily reach 3-5 ppm when considering the sample contamination and the errors of the Karl Fisher method (D1533)!
To obtain a reasonable valid value in this example, you have to take a sample when the oil temperature is above 70°C, or use a direct reading sensor immersed in the main tank of the transformer.
The second is measuring the water concentration immediately after dry-out. The humidity in the insulation system is in this case unbalanced and the concentration of water in the oil is deficient. Moreover, a humidity gradient was created in the cellulose components (components of several millimetres of thickness), during the drying process, and this gradient must readjust before an equilibrium can be attained between the oil and the cellulose. In these conditions, the evaluation of dissolved water is not credible. For a transformer at an ambient temperature, the equilibrium is achieved after about 10 days whereas at 50°C, this equilibrium is achieved in fewer days. 
Evaluating the humidity level in the transformer should be part of the conditions of purchase/manufacture/start-up and will directly influence its life span and the BIL.
In the transformer's preventive maintenance program, the humidity in the cellulose and the dissolved water in the oil should be monitored. For a transformer in service, the highest humidity limit should be approximately 1 per cent and 0.5 per cent (reference values only, manufacturers did not confirm these numbers) respectively for transformers operating at a voltage lower than 230kV and higher than 500kV. To do this we have to set a high limit for the concentration of dissolved water, such as those given in Table 4.
Let's illustrate the water evaluation with an example. Let us use the concentration of dissolved water as the sole criterion for a transformer operating at 230 kV. So, 20 ppm, at 50°C, would give a moisture content of 2 per cent. This is too high for a transformer operating at this voltage level. In these conditions, the effective BIL would probably be lower than that inscribed on the nameplate. In conclusion, testing for dissolved water in oil to evaluate the condition of a transformer is not enough; the moisture in the cellulose must be included too.
The concentration of water in the transformer is directly dependent on the average temperature of the oil and the windings, whereas the humidity in the cellulose stays relatively constant. Keep an eye on your data, the concentration of water in the oil is usually lower in winter and higher in summer and this does not come from a migration of atmospheric humidity towards the transformer tank.
The reason is that water migrates between the cellulose and the oil according to the temperature. The water content in the tank is divided in a ratio of 50/1 between the cellulose and the oil. Therefore, the concentration of water in the oil does not represent more than a fiftieth of the total water content in the tank.
In this way, we see why the practice of passing the oil through the "degasifier" a couple of times in order to remove the water has no effect because we removed only a fiftieth of the totality of water contained in the interior of the tank. If we measured the concentration of water several weeks after the treatment, we would notice that the concentration of water has returned to the level before treatment.
Humidity Evaluation of the Transformer
To determine the relative humidity contained in the cellulose in the transformer, you must measure the dissolved water and record the top oil temperature at the moment of sampling. The water concentration in ppm is evaluated using the Karl Fisher method (D1533) from an oil sample taken from the main tank. The sample should be representative of the oil volume. Finally, determine the humidity in the transformer, with the help of Figure 4. For example, to know the humidity of the cellulose at 20 ppm and 50°C top oil temperature, we assign these two values respectively on the horizontal and vertical axis to locate the point in the graph. This point intercepts the 2 per cent on the humidity curve.
It is also possible to obtain online values of both the dissolved water in oil and an evaluation of the moisture content in the cellulose from a sensor immersed in the transformer tank .
To effect an adequate evaluation of the humidity in the transformer, we suggest sampling oil with a syringe, because in a jar, the sample has a tendency to contaminate itself with the humidity in the atmosphere, especially at low dissolved water concentration.
Carry out a sampling campaign during the hottest period of the summer. When the transformer operates at its highest temperature, the dissolved water concentration should be at its highest, rendering the analysed oil sample data more precise.
The precision of the measurement of water by the D1533 method (Karl Fisher) is in the order of +/- 2 ppm. To illustrate the effect of the lack of precision on the measurements, let us take the example of a transformer where the humidity is 1.5 per cent and the oil is sampled at 32°C and 55°C.
With the help of Figure 5, we can identify a concentration of 5 and 20ppm. If we factor in the lack of precise measurement (+/- 2 ppm) to these quantities, we find the limit of the possible values (using D1533) as respectively 3-7 ppm and 18-22 ppm and the humidity limit corresponding at 1.2-2.1 per cent and 1.4-1.7 per cent.
At 32°C, the imprecision effect is unacceptable, when at 55°C the limit produces a value from which it is possible to make decisions.
Finally, make sure that the transformer temperature is relatively constant during the days preceding the taking of oil samples. This precaution will assure you that the transformer is in a quasi-balanced state, where the concentration of water measured will be valid.
Michel Bélanger owns and operates SEIDEL. He can be reached at (418) 822-3561. Part 2 will appear in the next issue of Electricity Today. ET
- IEEE Tutorial Course, Application for Distribution and Power Transformers, #76 CH 1159-3-PWR (1959), pg. 28.
- Kelly J.J., A Guide to Transformer Maintenance, Transformer Main-tenance Institute, pg. 75, SD Myers, 1981.
- Annual survey, Doble Engineering Company.
- Report of Transformer Reliability Survey -- Industrial Plants and Commercial Buildings, James W. Aquilino, IPSD 80-7.
- Test performed on a generator step up transformer, 25MVA, WOJ, 13.2/161 kV, 31800 litres, August 2000- May 2001.
- Sensors to monitor dissolved water in oil and the cellulose moisture level inside a transformer, SEIDEL, 2001.
- How To Reduce The Rate of Aging of Transformer Insulation, L. R. Lewand, P. J. Griffin, Neta World, Spring 1995.
- Insulation Thermal Life Consider-ations for Transformer Loading Guides, W.J. McNutt, Transaction on Power Delivery, Vol. 7 No.1, January 1992.
- Background Information on High Temperature Insulation for Liquid-Immersed Power Transformer, Working Group Report, IEEE PES Transformer Committee, Insulation Life Subcommittee, Working Group on High Temperature Insulation for Liquid-Immersed Power Transfor-mer, IEEE Transaction on Power Delivery, Vol. 9, No. 4, October 1994.
- The Effects on Windings Clamping Pressure due to Changes in Moisture, Temperature and Insulation Age, T. Prevost, W.J. Woodcock, C. Krause, Weidmann, 67th Annual International Confer-ence of Doble Clients, March 2000.
- 'Aging - A Perspective', T.J. Lewis, IEEE Electrical Insulation Mag-azine, July 2001, Vol. 17, Number 4.