In 2001, the first commercial online power transformer monitoring system in Brazil was installed at Alumar, one of the largest aluminum production complexes in the world. The monitoring system was installed on a three phase 343MVA 230-34.5kV transformer with two OLTCs (On-Load Tap Changers), one on the high voltage and another on the medium voltage side. This article describes the system’s architecture, based on a decentralized data acquisition philosophy and using intelligent devices, and also its monitoring and diagnosis features. We have also mention here some of our practical experiences during the implementation and operation of this system in the last five years, including early detection of a mechanical problem in one of the OLTCs.
1.0 - INTRODUCTION
ALUMAR is considered one of the world’s largest alumina and aluminum industrial complexes, a consortium of the companies Alcoa, Bhpbilliton, Alcan and Abalco. The complex currently has an approximate annual output capacity of 1.5 million tons of alumina and 380 thousand tons of aluminum. The plant is basically split into two areas: Reduction and Refinery.
The Reduction area has received this name because of the alumina electrolytic reduction process required to produce aluminum, which is carried out there. A significant quantity of electric power is required for the alumina reduction to take place. The average demand of electric power in this area of ALUMAR is of 765MW, and its average monthly power usage is 570GWh. ALUMAR is connected to the basic power grid, through which Eletronorte delivers power to it at a voltage of 230kV.
The process of alumina reduction requires a reliable power supply without interruptions, which must be kept steady for over two hours. If this happens, it would freeze the entire production line, resulting in losses which could amount to millions of US dollars.
This is why the company’s electrical system and its power transformers are an essential asset in this line of business, a fact that led ALUMAR to purchase a new 343MVA reserve bay that was added to the existing four in 2001. Each existing bay has a 300MVA 230-34.5kV transformer, with one high-voltage OLTC and a 34.5kV voltage regulator transformer, also equipped with one OLTC. However, for the new bay, the option was to build both transformers in the same tank, resulting in equipment with two OLTCs, one conventional, oil-based arc extinction type, on the 230kV side, and another with vacuum chamber extinction one, on the 24.5kV.
While still being purchased, the equipment was specified in a way to allow installation of an online monitoring system in the future. This system was purchased from, installed on the field and placed in service by Treetech still in 2001, immediately after the transformer had been delivered and commissioned. Therefore, this monitoring system was the first of its kind to be installed and to start operating regularly in Brazil, pioneering the acceptance of online transformer monitoring devices, which has been quickly growing.
2.0 - ONLINE MONITORING SYSTEM ARCHITECTURE
The online monitoring system installed on the transformers at ALUMAR (Sigma, by Treetech) uses a modular and decentralized architecture ,, based on Intelligent Electronic Devices (IEDs) installed on the control panel of the transformer casing, from where data is sent via a serial interface to a computer at the plant’s control room, which runs the software to store the data received, making it available and able to be processed, as shown in Figure 1. These three main parts of the monitoring system architecture are described below.
2.1 - Intelligent Electronic Devices (IEDs)
Some of the IEDs perform primary transformer control functions and are therefore used in the transformer regardless of the presence of a monitoring system. The existing equipment is integrated into the monitoring system through one of its serial communication ports, in order to work simultaneously as sensors to supply data to the system, but without increasing costs.
Other sensors were installed specifically to be used by the monitoring system, but also based on the philosophy of decentralized IEDs integrated to the system through their serial ports. In the few cases where it was impossible to integrate them to the system via serial interface, either because the devices are not intelligent devices or because the manufacturer does not offer open protocols at serial ports (only proprietary protocol), universal data acquisition modules were used, able to receive multiple digital and/or analog signals and digitizing them, thus making them available through open protocol serial ports.
It was then possible to integrate every sensor, both intelligent and conventional, to the monitoring system through serial communication port. This was also a reason why no centralizing equipment was installed on the transformer casing, and simplified both the monitor’s design and installation, reducing the initial costs and, last but not least, also reduced the TCO (Total Cost of Ownership) of the system as it increased its reliability and availability.
Another beneficial feature of a decentralized architecture, with IEDs, is the system’s modularity, allowing free choice of monitoring variables and also facilitating future expansions, by simply adding new IEDs. Several factors may have to be considered when making these choices, including transformer asset value, its importance in the production chain (or in energy generation, transmission or distribution systems), among others.
Figure 1 – Monitoring System Architecture
2.2 - Physical communication media
The physical medium used for communication in this case is a copper, shielded twisted pair cable. Even though fiber optic solutions were available and possible, at higher costs, there is a conviction, based on the features of the RS-485 communication standard, that this option could be used with satisfactory results. Among the RS-485 features mentioned above is the fact that the RS-485 operates in differential mode, which associated to the mutual cancellation of interferences in adjacent legs of the twisted pair makes this standard less susceptible to the interferences already expected in substations for this level of voltage.
As expected, the twisted-pair solution has shown to be totally satisfactory in spite of a complication in this kind of layout: the high intensity magnetic fields generated by the high currents used in the aluminum production process.
It is worth highlighting that, as Lavieri et. al. stress: , the fact that the IEDs used are developed specifically for the substation environment is essential for this strategy to be successful. Any equipment originally developed for industrial purposes, when used in this type of application, usually shows fragility and lack of reliability and availability of the serial communication ports when they undergo electromagnetic surges and voltage impulses, in addition to extreme outside temperatures.
2.3 - Information Storage, Availability and Processing
The data supplied by the IEDs located on the transformer, both direct readings and information obtained from data preprocessing, are received by a computer running the monitoring software, in this application located in the plant control room.
The main functions of this software can be divided into two groups: Data Digitization functions, associated to simple data availability and storage, and Monitoring functions, with the objective of transforming simple data items into useful maintenance information. In the monitoring system at ALUMAR, the following functions were created:
- Data Digitization functions:
– Readings, alarms and status displayed online
– Readings, alarms and status stored as history databases.
– Readings, alarms and status queries, stored in history databases as charts or tables.
– System can be accessed remotely or locally.
- Monitoring functions:
– Data processing based on algorithms
– Data processing based on mathematical model
– Transformer current status diagnosis
– Transformer future status prognosis
– Early failure detection.
3.0 - MONITORING FUNCTIONS
In systems in which the purpose is to obtain useful information for transformer maintenance, such as status diagnosis and prognosis, the Monitoring function block is specially important, confirmed in practice in this facility, when a defect in the OLTC was detected early (see item 4.1).
As in the IEDs used in reading data acquisition, the system’s data monitoring functions also have a modular organization, allowing the user to freely choose the functions he desires to install, and also facilitating future expansions, which can be done by simply adding new software modules and their corresponding IEDs. As already explained above, several factors may need to be taken into account in this choice, such as the value of the transformer or its importance in the electrical system, among others.
Following this modular philosophy, the monitoring system in operation at ALUMAR was equipped with the monitoring items described below, considered at that time as the most important for this application, although it is possible to add any other items among those currently available.
3.1 - Insulation Service Life
This monitoring function performs calculates the estimated insulation life lost due to cellulose thermal aging, according to the load and temperature undergone by the transformer. It also calculates the average life reduction rate and the extrapolation of the theoretical remaining service life for the insulation, as described in the next item.
3.1.1 - Cellulose degradation mechanisms
The main component of the different solid insulating materials used in high voltage equipment immersed in liquid, mainly power transformers and reactors, bushings, PTs, CTs, etc., is cellulose. Among solid insulation materials, the most commonly used nowadays is paper.
Cellulose is an organic compound with a molecule formed by a long chain of glucose rings, or monomers. Each new molecule of cellulose has between 1,000 and 1,400 glucose rings, connected as shown in Figure 2. Each cellulose fiber has many monomer chains like this one.
The number of glucose rings connected in the chain is called Level of Molecular Polymerization. Since it is the length of these molecules that imparts to the cellulose-based materials their mechanical resistance, the level of polymerization of the material gives us an indirect indication of its mechanical characteristics, for example: Tensile strength, which can be associated to functionality or service life of the material.
Cellulose degradation is, therefore, caused by the reduction in monomer chain length, as well as by the condition of each chain. Three mechanisms may contribute towards the degradation of these chains in the cellulose used in manufacturing power transformer insulation systems and in similar equipment: Hydrolysis, Oxidation and Pyrolysis . Even though the latter one is related directly to thermal degradation, they all interfere in this aging process, therefore all three mechanisms are interrelated.
Figure 2 – Cellulose molecule
220.127.116.11 - Hydrolysis
Water causes the monomer chains to break, by attacking the oxygen atom that bridges the rings. Two –OH groups are formed, each attached to a monomer. As a result there is a reduction in the level of polymerization and consequent weakening of the cellulose fiber.
Fabre & Pichon  formulated a simple rule for the degradation of cellulose as a function of the water content. They proposed that the thermal aging rate of cellulose is directly proportional to the water content. So, if the results of the thermal aging tests show a given degradation rate for a given level of water content, any equipment operating with double the water content will have a thermal degradation rate for the insulation twice as high as the one measured in the aforementioned test. Data obtained by Shroff e Stannet  confirm this relation, illustrated by the following equation:
PV ∝ QP, where:
PV is the rate of insulation life reduction, and
QP is the water content of the insulation paper.
18.104.22.168 - Oxidation
The carbon atoms of the cellulose molecule are attacked by oxygen, forming aldehydes and acids. Consequently, the union between the rings is weakened, which leads to low polymerization levels. Water, carbon monoxide and carbon dioxide are released. The water released by this process will also contribute to the hydrolysis process described above.
The cellulose is attacked by water, and the oil undergoes oxidation, producing acids, esters and other substances that, in turn, also attack the oil, generating even more oxidation products. These substances also attack cellulose, further degrading it.
The way how oxygen affects cellulose degradation rate has been investigated by several researchers, and the most common procedure is to compare the results of aging rates from samples taken in sealed insulation, free from oxygen, with rates of samples exposed to the atmosphere, like those taken from transformers without oil preservation systems. Some of the researchers who studied this phenomenon were Fabre  and Lampe , who found degradation acceleration factors of 10 and 2.5 times, respectively, for the samples exposed to oxygen in relation to those in sealed samples.
Clearly, the presence of oxygen has an extremely adverse effect on the aging of cellulose, and must definitely be prevented. If the oil preservation system fails, and the oil gets in contact with the air we can expect the aging process to be considerably accelerated.
To prevent this risk, the monitoring system includes a sensor to monitor the rubber membrane that prevents contact between the oil and the atmosphere. If this membrane ruptures, an alarm is issued by the monitoring system.
22.214.171.124 - Pyrolysis
Extreme heat leads to carbonization of the cellulose fibers. On the other hand, moderately intense heat, usually found in transformers, causes rupture of the individual monomers of the cellulose chain, forming solid residue and releasing carbon monoxide, carbon dioxide and water. Again, the level of polymerization is reduced, weakening the mechanical resistance features of cellulose.
Since temperature is not distributed uniformly in transformers, the analysis of cellulose deterioration by heat is done taking in consideration the temperature of the hottest spot, since this will be the site where the highest level of degradation will occur.
3.1.2 - Online Monitoring of Insulation Aging
In compliance with the Brazilian loading standards for power transformers, transformer insulation service life in transformers with insulating oil with characteristics of new oil (neutralization index, content of oxygen dissolved in oil and water content in controlled Law, where the
insulation) is exclusively given by the Arrhenius’s service life logarithm is the inverse function of absolute temperature:
log (life) = A + B / T , where: A and B are constants, and
T is the temperature of the hottest spot. The chart in Figure 3 shows Arrhenius’s law in the format of annual insulation life reduction for different temperature values of the hottest spot, supposing that the temperature will remain constant throughout the period.
Since, in practice, the temperature of the hottest spot varies according to load and surrounding temperature changes, the life insulation reduction is calculated in short time intervals during which the temperature remains virtually constant. The small reductions of insulation service life incurred in these time intervals are accumulated for the entire system operation time, giving as a result the total insulation service life reduction.
Ptotal = Σ Pi , where:
Ptotal is the total aggregate reduction of insulation service life,
Pi are reductions in insulation life for the small incremental time intervals, and
As described above, in item 126.96.36.199, the water content in the insulation also plays a major role in the insulation life rate reduction, by accelerating the insulation degradation proportionally to the existing water content. Thus, thermal service life reduction calculated by the monitoring system is corrected for the water content found in the insulation, which can be the one calculated by the system or an estimated fixed value.
Based on the aggregate percentage insulation life reduction data, it is possible to project the remaining expected life time, by observing the evolution of the life reduction rate for a period in the past that is representative of the average operating conditions for the equipment.
Figure 3 – Annual reduction of insulation service life under constant temperature
3.2 - Final Temperature Gradient Forecast
This monitoring function calculates future values of the oil/winding temperature gradient and issues an alarm when a trend that will lead the coil temperature to reach temperature alarm levels or shut down is detected, in addition to informing the time remaining before alarm and/or shutdown temperatures are reached.
3.2.1 - Winding-Top Oil Temperature Rise
The loads applied to power transformers cause the temperature of the hottest spot on the winding to rise over the top-oil temperature, which is a function of the load applied and the system’s specific reduction and heat exchange features. This principle is also used in determining the temperature of the winding through a process called “thermal imaging”:
∆θEO = f (l,c), where:
∆θEO is the rise in winding temperature over top oil temperature,
l is the load, and
c are the transformer’s own characteristics.
For each load value applied there is a corresponding rise value of the winding temperature over top oil temperature, as in the example shown in figure 4.
In this example, we notice that the rated load (100%) corresponds to a 30ºC rise in the rated temperature, a value obtained in transformer heating tests. However, due to thermal inertia of the copper/insulating material mass, applying a given load does not immediately bring about an instant corresponding temperature rise as shown above. This inertia causes the winding temperature to rise gradually from the current value to the new value (corresponding to the new load) following an exponential curve for a given time constant.
Figure 4 – Rise in winding-oil temperature as a function of the load
3.2.2 - Winding-Top Oil Final Gradient Monitoring
As shown in the chart in Figure 4, when a given load is applied on a transformer, it is possible to forecast winding final temperature after thermal stabilization.
This allows to calculate whether the rise in winding temperature over the top oil temperature will achieve levels that will cause protection systems to trigger alarm signals or even enable shutdown sequences.
If the forecast for the system temperature exceeds the alarm level value, the monitoring system issues the alarm for this condition, also informing the time remaining before the alarm goes off, a value that is reached based on the thermal time constant of the winding.
Figure 5 shows an example of the evolution expected for the winding temperature, the time to reach the alarm value and the final winding temperature after stabilization. In this example, when an overload is applied to the transformer, the monitoring system would
initially calculate that 13 minutes remain before the alarm temperature (or the shut down temperature) is reached, with calculation being continually readjusted.
Likewise, the extrapolation of temperature rises of the winding temperature over top oil temperature can also be applied to the temperature rise of the oil over the surrounding outside temperature, allowing trend monitoring for future rises in temperature with advance warning of hours.
Figure 5 – Evolution of winding temperature along time
3.3 - Gases in Oil
This monitoring function performs online supervision of the concentration of hydrogen dissolved in oil. Since hydrogen is generated in nearly every type of internal defects that might occur in transformers, it is considered a key gas in defect detection.
Therefore, the monitoring system issues alarms based on the ongoing follow up of the hydrogen in oil content, when high levels are reached, such as, for instance, when a rising trend is detected that the hydrogen content will reach these high levels in the near future.
3.4 - Moisture in Oil
As already explained in item 188.8.131.52, the presence of moisture in insulating paper potentializes the effects of insulation thermal degradation proportionally to the water content in the paper.
Keeping water content low in insulation is, therefore, essential. While the transformer is being manufactured, its active part is thoroughly dried, and the same is done with the oil used in filling the transformer’s tank for the first time. The new equipment, if this is done, is guaranteed to have a low water content in its insulation paper.
After this, several different processes can increase the water content in insulation. Among these processes is cellulose degradation, in which water is generated, but the main factor for increasing water content in insulation paper is sealing failure, which allows water from the environment to find its way into the transformer. In this case, water coming from outside will be absorbed first by the oil, from where it migrates to the insulation paper.
The monitoring system will first check oil expansion tank seal integrity by verifying if there was a rupture of the rubber membrane that prevents the contact of the oil with the environment. In addition to this the system also monitors the content of water dissolved in oil.
This monitoring function performs online supervision of the level of water dissolved in oil, issuing alarms for both high content levels reached and for rising trend detected that may in the future result in high levels of water in oil.
3.5 - Forced Ventilation Maintenance Assistant
Adequate transformer cooling is essential for their safe operation without accelerated insulation service life losses when operating under heavy load regimes. In this transformer’s case, this is achieved by using several fans to force air circulation through the radiators (ONAF cooling). Therefore, it is essential that these fans operate perfectly. Failure of one or more fans can cause activation of the protective mechanisms that limit temperature or reduce transformer load, and therefore restrict the availability of the equipment.
For this reason, normal fan wear must be monitored, which is traditionally done offline through the preventive maintenance schedule recommended by manufacturers. These interventions are usually based on the equipment’s operating time, and include replacing components (for example, windings).
The Forced Ventilation Maintenance Assistant allows accurate calculation of fan operating time, thus avoiding these manufacturer-oriented maintenance interventions to happen much before or after the time recommended by manufacturers. This monitoring function also offers much more useful information in order to help in fan maintenance:
- Total fan and pump operating time, from the beginning of operation, and time since last maintenance intervention, with motor start and stop records;
- Average daily fan and pump operating time;
- Time until next and subsequent recommended inspection or maintenance, based on daily average fan and pump operating time;
- Warnings issued with programmable advance for inspection or maintenance of the equipment due to operating time.
3.6 - On Load Tap Changer Maintenance Assistant
All failure statistics for power transformers show that the On Load Tap Changer is one of the main sources of failure, especially due to the moving parts that conduct and interrupt high currents while subjected to high voltages.
This is why this monitoring function helps in supervising regular commuter wear and tear, which is traditionally done offline, via the preventive factory maintenance schedule. These interventions are usually based on the number of tap changes and equipment operating time, and include visual inspections and contact thickness measurement procedures.
This monitoring function provides useful information to help with on load tap changers maintenance:
- Sum of the total current commuted since the changer was first placed in service, yielding the contact wear rate.
- Total number of operations since the changer was placed back in service after last maintenance
- Calculation of current thickness of the arc interruption contacts, through extrapolation based on previous thickness measurements and number of tap operations
- Total tap changer service time and total service time since last maintenance
- Daily average contact wear and daily average tap changes
- Time in which the contacts will reach minimum thickness or number and time until the number of operations or maximum time period before inspection or maintenance.
- Warnings, with programmable advance, about tap changer inspection or maintenance.
3.7 - OLTC Operating Times
On Load Tap Changers is one of the main sources of power transformer failures. The reason for this, as described previously, is the fact that OLTCs are mechanical, based on moving parts. Thus, mechanical failures of the On Load Tap Changer can cause problems of different magnitudes, starting from equipment unavailability until severe dielectric failures.
In this context, the function that monitors commuter operating times supervises the time required to perform the tap change in each operation of the OLTC, issuing an alarm if this time is different from the times observed during regular operation of the equipment. In item 4.1, there is a description of how this function detected an actual failure in one of the OLTCs.
4.0 - EXPERIENCES IN SYSTEM INSTALLATION AND OPERATION
4.1 - Diagnosis of a defect in an On Load Tap Changer
With early detection of failures being one of the main purposes of an online monitoring system, the most interesting event during the operation of the ALUMAR system was observed while still in the commissioning phase, when the system diagnosed a problem on the On Load Tap Changer on the 230kVside.
This diagnosis was made by the diagnosis function of the “Tap Changer Maintenance Assistant”, which, among other parameters, monitors the time spent in each commutation. According to this measurement, a difference was observed in relation to the system history.
When checking the On Load Tap Changer activation panel, we observed the presence of oxidation was on the cam tree of the activation mechanism. The oxidation caused the activation motor to remain in operation longer than required to perform the commutation.
Since this was detected early the defect was quickly repaired, and did not actually cause a disturbance in the regular operation of the transformer. If, however, the transformer and the On Load Tap Changer was not equipped with an online monitoring system, the problem would probably go unnoticed, and become worse and worse. After a certain point, the On Load Tap Changer would either “shoot” up or down, rising or lowering the voltage, when the maximum or minimum tap positions were reached. Depending on the voltage level at the 230kV input, voltage regulation for the aluminum manufacturing process could be severely impaired, with the client running the risk of production loss.
4.2 - Metal Pair Serial Communication
Because ours was the first commercial system to operate in monitoring transformers online in Brazil, one of the points to be checked when it was placed in service, in 2001, was to verify the feasibility of using serial communication RS-485 with copper cables in substations. This objective was fulfilled when this interface operated satisfactorily with copper cables even in very adverse electromagnetic interference conditions found in the facilities, which has been attested by these nearly six years of system operation.
4.3 - Compatibility with Other Existing Systems
Still during system installation, due to space limitation issues, ALUMAR wished to avoid installing an additional PC in the control room, which would have happened if they had chosen a dedicated computer to run the monitoring system. The monitoring system could possibly be installed in the same computer already running the supervisory system, which was confirmed during installation and long-term operation of the system, without any compatibility problems between systems.
5.0 - CONCLUSIONS
Because this monitoring system was, at the time, the first online transformer monitoring system to operate commercially, there were many expectations around the system we discussed in this article. In fact, shortly after it went into operation, facts evidenced the gains achieved with the installation of the system, when it detected a defect in an On Load Tap Changer that, under other conditions, would remain unnoticed and might cause severe future losses.
This fact shows that the savings that can be obtained from using an online monitoring system to avoid severe failures are substantial, and the price, often used as an excuse to not install monitoring systems is actually low.
To our discussion, we must add the fact that monitoring systems based on decentralized architectures can be assembled modularly, according to each client’s needs and budget, allowing the system to be gradually expanded in the future.
6.0 - REFERENCES
 Alves, Marcos, “Sistema de Monitoração On-Line de Transformadores de Potência”, Revista Eletricidade Moderna, Maio/2004.
 Amom, Jorge, Alves, Marcos, Vita, André, Kastrup Filho, Oscar, Ribeiro, Adolfo, et. al., “Sistema de Diagnósticos para o Monitoramento de Subestações de Alta Tensão e o Gerenciamento das Atividades de Manutenção: Integração e Aplicações”, X ERLAC – Encontro Regional Latinoamericano do CIGRÉ, Puerto Iguazu, Argentina, 2003.
 Lavieri Jr., Arthur, Hering, Ricardo, “Novos Conceitos em Sistemas de Energia de Alta Confiabilidade”, Encarte Especial Siemens Energia, http:// mediaibox.siemens.com.br/upfiles/232.pdf, Janeiro/2001.
 McNutt, W. J., “Insulation Thermal Life Considerations for Transformer Loading Guides”, IEEE Transaction on Power Delivery, vol. 7, No. 1, pp. 392-401, January 1992.
 Fabre, J., Pichon, A., “Deteriorating Processes and Products of Paper in Oil. Application to Transformers”, CIGRE Paper 137, 1960.
 Shroff, D. H., Stannet, A. W., “A Review of Paper Aging in Power Transformers”, IEE Proceedings, vol. 132, Pt. C, No. 6, pp. 312-319, November 1985.
 Lampe, W., Spicar, E., Carrander, K., “Continuous Purification and Supervision of Transformer Insulation System in Service”, IEEE Winter Point Meeting, IEEE Paper A 78 111-7, January/February 1978.