ABB Australia Pty Limited
Power transformers are critical assets that support the generation, transmission and distribution of electrical energy. With the profile change in power generation and consumption, it is important that the condition of transformers and their utilization are well understood. In the past, most transformers were fitted with basic sensors which at best only provided alarms after the condition of the transformer had already started to deteriorate. Recent years have seen a step change in both the availability of cost effective and reliable sensing solutions together with the data expectations surrounding the knowledge-based economy.
As the power industry embraces the fourth industrial revolution, many stake holders are still uncertain what this will mean for them, but it is clear that within the lifetime of major assets, such as power transformers, there will be a paradigm change in the way they are utilized and their role in the grid intelligence infrastructure.
It is therefore critical that asset and operations managers have not just data, but the right data to provide them actionable information that supports effective operations and maintenance actions. The correct application of sensing and domain knowledge remain key to the value any sensing solutions may bring.
This paper will highlight the steps to be taken to take advantage of the advancements in understanding transformer condition and the role sensing solutions together with domain knowledge will play.
The electrical power landscape is changing and will continue to change at a rate that is difficult to estimate, a recent study by McKinsey , commissioned for the World Economic Forum (WEF), indicates the “Power landscape” will experience more change in the next 10 years than the last 100. There are many reasons for this, not the least of which is the drive towards lower CO2 emissions and the technologies employed to achieve this.
Starting with generation, the uptake of Renewable power generation has gained significant momentum in recent years  and while traditional technologies such as Hydro still have their place, Distributed Energy Resources (DERs) such as wind and solar are seeing the highest growth rates.
The impact of Distributed Energy Resources is both technical and commercial. From a commercial perspective the costs of equipment have reduced dramatically over the last 5 years, while experience has also grown. This combined with what can be much shorter approval times than “traditional” generation sources and fast construction / installation times make even utility scale installations increasingly attractive and more common.
What some utilities may consider less positive, is the impact of “waves of Prosumer level installations”, at the industrial, commercial or residential level, many of which are now driving business models such as Virtual Power stations. The Prosumer model can have a significant negative financial impact for the utility where electricity generated is used locally, but these Prosumers can also contribute to the grid, when generating more power than needed for their own demand. For those not involved in this business area, in the virtual power station model service providers typically combine multiple smaller generation sources to bid on supply contracts, based on the theoretical capacity and forecast variables such as weather. This system is by nature dynamic and data driven.
Energy storage will also play a part in the grid dynamic going forward. Today sources such as pump storage can act as mini Hydro sources, but in the near future battery- based storage looks set to become significant. When sited in the same location as the previously mentioned DERs this could increase their impact. Storage facilities can also become part of the grid load, this will be discussed in the demand side section.
The technical impact of Distributed Energy Resources relates firstly to the intermittent nature of the generation levels and secondly to the ever increasing volume of semi-conductor devices, connected to the grid and the resulting harmonic disturbance.
For the purpose of this paper we will focus on the variable nature of the generation levels, but recommend those involved at the distribution level also consider the harmonic impact .
Renewables include stable sources such as Hydro generation, where for the most part the generation levels are fully under human control, providing of course the locations are not suffering droughts. Solar and in many cases wind generation are classified as Variable Renewable Energy sources (VREs). These are generation sources where only limited control can be applied and nature plays a significant part in dictating the level of power generated, these sources can therefore be subject to dynamic changes in many installations. There can also be cases where the timing of the generation is “out of step” with the demand.
This can further impact the already known demand variable previously identified by the California Independent System Operator CAISO, referred to as the “Duck curve”, a sample of which can be found in Fig 3.
The impact of the above mentioned DERs is further complicated by the incentives for producers to reduce the traditionally stable generation sources such as coal and or nuclear. The net impact of the above is an increase in VREs while at the same time the grid inertia is negatively impacted by a reduction in stable sources. The impact and timing of some DER’s generation are already making their presence felt at the transmission level, with several transmission system operators (TSOs) reporting the “over-fluxing” of transformers where the load side voltage becomes a feed and is exceeding the design parameters of the transformer! In the case of “early adopters” such as California and Denmark, they have focused on a robust transmission system with strong ties with neighboring systems .
Over excitation is of major concern on transformers directly connected to traditional generation sources, in some cases the step-down transformer may benefit from protection relays , however this is not always the case. It is now not unknown for utility customers to include requirements for core temperature measurements via PT100s or even fiber optics in specifications for industrial sites in remote locations, however the service conditions cannot easily be measured during factory acceptance testing (FAT) since high voltage at no load or high current with little core flux are not simultaneous in the factory as per in service.
Note: Transformers have an inherent voltage drop, known as regulation, which is primarily a function of load magnitude, impedance and power factor. Systems need to overcome this regulation to maintain output bus voltages by direct over-excitation and/or use of on-load tap changers. When the tap changer is on the input circuit of the transformer, changing tap ratio also causes further over-excitation of some core sections to the extent = 100/100−regulation %.
The increased DER and power generation dependence on uncontrolled drivers leads to both greater and more rapid variation in bus voltage, power factors and even changed power flow direction. All these factors increase the transformer regulation and hence increase resultant over-excitation. In some cases, insufficient tapping range may result in some transformers resulting in the need to run input bus voltages higher than previously.
It is recommended to review the tapping range requirements for any transformer that could be subject to “Prosumer” type applications.
While generally considered stable in the developed world, the demand in many countries is actually increasing slightly, this is not due to population growth, but the increased demand per capita. As is the case with generation, much of this increase in demand relates to DC power supplies for electronic devices such as televisions and computers or electronically controlled devices such as refrigerators and washing machines. This increase will however pale into insignificance if the demand side is impacted, as expected, by the rise in e-mobility and the adoption of electric vehicles (EVs) charged from the power grid. Several countries have already committed to significantly reducing the number of fossil fuel based vehicles and some have gone so far as to set hard targets to prohibit the sales of vehicles that are not electrically powered, in addition manufactures are forecasting significant uptake in EVs, with even those considered “traditional” expecting more than a quarter of all new cars sold to be EVs by 2025! One of the main obstacles to such policies is the time taken to charge those vehicles and the number of charging stations available. This is set to change, 350kW charging stations are already commercially available which can sufficiently charge a vehicle for a 200kM range in only 8 minutes and Germany, for example, has already mandated that every new home built from 2021 forward must include the cabling for vehicle charging. The EU are also trying to push through legislation. The impact of many cars utilizing fast chargers could be significant, not just for a local micro-grid but, if this were to happen simultaneously in multiple parts of any country, it could impact the grid as a whole. While the final unit size and adoption rate of EV charging points is currently unclear, the oil companies are already getting on board. BP has recently bought Charge Master and advised they will double the current UK deployment rate within the next two years i.e. increasing their new UK installations from 100 to 200 per month. BP is not alone as Shell is also now actively involved with the Ionity Network, who are in turn in a collaboration with a consortium of car makers.
Sticking with the demand side, Data Centers are becoming increasingly important for society. These are also more power hungry than most people realise and it is not uncommon for individual sites to require as much power as a small town; again power electronics based.
Last but not least, we have battery-based storage, which has both the potential to smooth supply and demand peaks or troughs, but would in either case add to the potential for either additional switching pulses or introducing harmonics.
In summary both the supply and demand side are already more dynamic than only a few short years ago. There is every reason to believe that this trend will only increase, that this will have an impact on both how the transformers are being loaded and the associated implications, together with the need for operators to better understand how these key assets can be best utilized.
As described above, there are many things that the human race can now control, however this does not yet include nature. The power infrastructure is often taken for granted, but nature has a habit of reminding us of its own raw power, whether through earth quakes or weather “events” such as superstorm Sandy. In 2017 alone, there were 16 weather events and climate-related disasters with losses exceeding $1 billion each, only in the United States! According to research , 92% of US organizations have experienced an energy failure in the past year, and over a quarter admit that not knowing available options is a roadblock to future-proofing their business.
While it is not always possible to prevent damage to key assets, visibility of the asset status and trends prior to any problems is key to forming a cohesive plan regarding where to focus efforts and resources, when re-establishing power. Good visibility is also required, when advising customers/consumers of both the status and expectations for when power will be returned.
Impact on Power Transformers
Power Transformers have few moving parts, or at least they should have when things are operating as planned. The moving parts they do have i.e. tap changers and pumps or fans are typically designed to allow for maintenance throughout the transformer’s life. It is however worth noting that the potential exists for an increased number of switching operations and the resulting mechanical wear, where transformers are connected to DERs.
Assuming suitable maintenance of these moving parts together with ancillaries such as the bushings, the main life expectancy or indeed end of life of a power transformer relates to the insulation system of paper (cellulose) in combination with mineral oil or a suitable alternative fluid type. The insulating fluid can be considered the life blood of the transformer providing both support for the hard insulation system and the cooling. Continuing with the life blood analogy, the insulating fluid is used as an insight into the transformer’s well being or an early indicator of potential problems. It can also be cleaned or reprocessed at relatively low cost and little inconvenience at the location of the transformer’s installation.
The hard insulation system is much more difficult to either measure, monitor or replace. Chief among the factors that age cellulose is temperature. The temperature and duration that the hard insulation is exposed to such temperature is, therefore, the key factor that dictates the remaining life of the insulation system. Like many other machines or processes, the end of life relates to the weakest link or most critical point, which for power transformers is typically the hottest point in the windings, better known as the “hot spot”.
Power Transformer designers go to great lengths to determine where the hotspot will be and ensure this will remain within the values provided in the customer’s specification. The location of the hotspot cannot simply be assumed to be at the top of the winding , as many factors influence this, not the least of which is the relationship between the individual winding nests, the interaction of any stray flux and the oil flow.
Considerable knowledge, experience and the availability of validation tools developed over many years are key to identifying where the hotspot will be (see fig 6) The ability to simulate the actual conditions inside a working active part requires special tools, however these can vary significantly in mapping the correct hotspot level and location.
Provided the hotspot position can be calculated with sufficient accuracy, then algorithms are available to map the hotspot temperature from known variables .
Many white papers are available that focus on the difference between average oil temperatures, winding temperature and the corresponding hotspot. In order to provide guidance on this topic there are also international standards  and , an example of which can be seen in fig 7.
Figure 7 is generic, but can be used to illustrate the relationship between the oil leaving the cooler and entering the windings, together with the corresponding temperatures in the windings and hotspot. This assumes a sinusoidal waveform and stable loading. However, due to the above mentioned dynamic sources of both generation and demand, it should also be considered that the power transformer could be exposed to rapid changes, and that due to the system hysteresis there can be a significant offset between the rapid increase in winding temperature and the ability of the transformer’s radiators to dissipate the additional heat.
A typical power transformer will only radiate 10% of the heat directly from the tank itself.
In figure 8, conduction is shown in grey, while convection is red and radiation is blue. The impact of this time delay is well illustrated in IEC60076-7  as can be seen in fig 9. where the cooling medium lags the winding in dissipating the additional heat generated in the windings.
In fig 9. K indicates the per unit loading level, where θh is the winding hotspot temperature and θo is the oil temperature. Where K1 is taken as the nominal rating of the transformer it can be seen that any system that controls the cooling based purely on the oil temperature would allow the hotspot temperature to exceed its nominal values for a considerable time.
While there can be commercial or even system safety grounds for the selected overload of power transformers, it must also be considered that the increased hot spot level and its duration are the driving factors in the aging of the paper insulation system and exposure to elevated temperatures will have a disproportionate impact on the remaining life.
In fig 10  we can see an example of the aging impact of exposure to increased temperatures, where attention should be paid to the fact that the y axis is logarithmic i.e. if we take 108°C as the hotspot temperature at nominal rating, then a mere 8°C increase would halve the insulation life for the duration of the exposure. Prolonged exposure or an increase in the temperature in excess of 130°C could also impact epoxy- based components or seals.
In fig 11, we can see the hotspot temperature profile for a collector step-up transformer used on a wind farm. The top oil temperature remained within specification and the wind farm operator was happy with the 1.55 PU average aging factor.
The condition of the paper is typically difficult to sample without taking the transformer out of service, lowering the oil level and typically increasing the risk of introducing foreign bodies. Where it is possible to sample the paper then the depolymerization index (DP value) is used to categorize the paper’s mechanical condition; in fig 12 we have used an example of 1000 for new paper and 181 from a unit that was at the end of its life (any kraft paper DP figure bellow 200 would normally be considered to represent the end of life) . While the DP test was not originally intended for use with power transformers, it is generally considered that paper with a value below 200 will no longer have sufficient strength to withstand a fault or network disturbance and is likely to be the cause of a transformer failure shortly afterwards.
The outlook for operators of power system assets, such as power transformers, looks to be driven by increased pressure on the demands for network reliability and profitability, both of which will see increased pressure to reduce costs and increase efficiencies in the grid itself and the relevant maintenance operations. All of which looks set to coincide with gaps in the workforce and difficulties to replace those with the most experience, many of whom are now of retirement age.
The grid dynamic is set to continue and the expectation is that more “un-regulated” generation sources will need to be accommodated.
In the age of the knowledge-based economy, the solutions technology has to offer the means since the ability to visualize what is happening with key assets will be vital.
Information, not data
The use of electronic monitoring devices on power transformers is not new, but the historical focus has been managing the end of the transformer’s life. Looking forward, the demands on system management are set to increase, as described above. However, typical enterprise systems are operating in the IT environment or at best in corporate metering and protection systems and they will not typically benefit from a mass influx of data points from power transformer condition monitoring systems. Equally, data sources that are not suitably robust to withstand the environment in which the transformer operates are likely to be more of a liability than a benefit.
In the case of power transformers, it is important to take a pragmatic approach to monitoring and first consider what information is of value, what is already available (external to the transformer) and what will be the use of additional data collected. As previously discussed, temperature is key to the health of the insulation system, but today is very difficult to measure directly. It is however possible to utilize proven sensors, already fitted to the transformer, such as the winding temperature indicator and top oil temperature.
Taken together with the design and heat run data, it is possible to calculate the hotspot temperature with sufficient accuracy.
Note: the vast majority of new power transformers are fitted with a current transformer based winding temperature indicator (WTI) and utilize a PT100 fitted in a well (or pocket) for monitoring of the top oil temperature. Where a PT100 is not already requested, this is a simple addition.
The use of electronic temperature monitoring (ETM) is the ideal foundation for any monitoring system as this will typically be able to forecast the consumed life of the paper insulation and also control the cooling system based on the hotspot. This will reduce the risk of unnecessary aging and provide its own self check or watch dog function to highlight any related malfunctions. The more advanced ETMs have the potential to both store trending data and facilitate the generation of useful reports. In addition, it may also be possible to add a simple ambient air sensor, to enable the forecasting of potential overload and associated duration.
The ETM level is the ideal starting position for most new transformers, but considering the long life expectancy, it is also important that this foundation is also the enabler for future requirements, which should facilitate both additional field level devices (sensors) and the ability to connect into higher level systems such as substation SCADA systems or cloud based monitoring. Independent of any connectivity, it is important to start collecting and trending data as soon as possible; dynamic data collection and trending cannot be done retrospectively.
Connectivity is something we now take for granted, to the point where a loss of telephone signal is already considered more than just inconvenient. The perception of connectivity and what is considered safe varies widely from situation to situation. However, most people forget we already trust connectivity, such as software and sensors in the most important aspects of our lives i.e. personal safety and financial security. The brakes on modern cars are almost exclusively fitted with ABS, many throttles are “fly by wire” and the only direct connection that a pilot of a large aircraft has to the plane is the seat that he or she sits in. Likewise, some countries today are using very little cash and almost all of us use some form of credit or debit card.
Fig 15 provides an indication of where different industries are along the digital adoption curve. While most people will not be surprised to see Information, Communication and Telecoms at the forefront, many will be surprised to see how advanced utilities already are, and to learn that some distribution utilities are already employing teams of analysts.
Substation automation is quickly becoming the norm, so any device installed on a power transformer today should be considered as a foundation for the future and, even if not initially connected to any other device, this should be planned for at some point in the transformer’s life time.
Mindful of the threat of cyber-attacks most organizations now have policies in place regarding what devices can be connected to which parts of their IT or OT infrastructure. It is therefore required that any monitoring solution has the ability to store / trend data locally (on premise) but retains the flexibility that such data can be easily converted into “useful information” without the need for customers to connect to the substation, their enterprise system or any other 3rd party internet based system.
Trending of the key variables such as loading and temperature is increasingly important and this should be a priority for all operations professionals. However, the benefits of being able to compare assets requires either time (man hours) or the ability to connect multiple assets to a central point. The degree of connectivity will vary from installation to installation and business to business, but advantages can typically be found by simply comparing the dynamic data from sister transformers, located in the same substation or substation to substation.
Mindful of the importance data will play in the future, it is recommended to work with both open communications platforms such as DNP3, Modbus, IEC61850 and asset management tools that can grow with any requirements, which look set to expand.
The key change in benefits from monitoring and trending of transformer parameters relates to the changing nature of the grid; however the historical elements should not be forgotten. Monitoring was and remains a tool to catch any developing faults at an early stage and prevent potentially catastrophic consequences. While less common than in the past, there can still be some infant mortality (see fig 17 bath tub curve).
In addition to comparing load vs temperature another important “flag waver” is the development of hydrogen. The use of basic dissolved gas analysis (DGA) devices for the detection of hydrogen and moisture is not new, but this type of device has a chequered history with some customers experiencing problems and nuisance tripping. The use of multi-gas sensors is preferred by some, but others find the costs of calibration and servicing “more effort than benefit” on a new transformer. As with many things, technology has advanced in recent years and there are devices now on the market requiring no calibration and little or no maintenance, for 10 years or more. These are now worthy of consideration and in some cases can be moved from transformer to transformer with a minimum of down time.
DGA devices are not the only “sensors” that have evolved over time, however most customers are wary to step into pure “electronic” protection with both feet. They do however see the benefits of being able to monitor the variation of parameters rather than only receiving alarm or trip signals, which only provide a snap-shot from a single point in time.
To meet the demands for both traditional electromechanical instruments and dynamic visibility of process variables it is now possible to utilize devices that retain their electromechanical heart, but are complimented by additional modules, these include Buchholz relays, pressure relief devices, oil level indicators and oil or winding temperature indicators.
In the above examples it can be seen that not only hard-wired alarm or trip options are available, but also the option to detect “flutters” or rate of change information.
All of the above mentioned devices are now available with the option of connectivity via simple two wire industry standard communications. Modbus was originally published in 1979 and is considered the grandfather of communication for industrial field level devices such as sensors. Modbus RTU is an open standard that is vendor independent and royalty free. It sacrifices outright speed for simplicity and robustness and is still today believed to be the most popular form of communication at the industrial sensor level.
Additional sensing options
The range of potential sensors and monitoring solutions covers everything from the ancillaries such as bushings and tap changer through to solutions mounted inside of the tank or windings such as fiber optics and partial discharge sensors. The choice of which of these systems to use typically comes down to experience and the criticality of the transformer in question [11, 12]. However, it should be considered that while bushing or tap changer monitoring can typically be easily retro-fitted, fiber optics must be installed during the production phase and UHF based partial discharge sensors will need suitably positioned valves or windows included in the tank construction.
Fiber optic sensors for measuring the hot spot were first developed in the 1980’s and, as with many new technologies, suffered some teething troubles. These troubles were however mostly related to securing the fiber optic in the correct position. As such, many customers specified 100% redundancy and typically only made use of the measurements during the heat run. The ability to combine the fiber optics with a winding spacer has significantly increased the hit rate these devices provide and today the use of fiber optics is common place.
Note: introducing any foreign body into a power transformer winding brings risks and fiber optics will need an evaluator with the required number of channels. Last but not least, where fiber optics are utilized for operational data then a gland plate will be required.
A Partial Discharge (P.D.) test is part of a power transformer’s factory acceptance testing (FAT). As DGA is also regularly monitored, most P.D. monitoring solutions are retro-fitted sometime later in the transformers life. The three main solutions used are Bushing based which represents the lowest cost and is good for detection, Acoustic which is mostly easily applied and can potentially be used without needing to take the unit offline and UHF, which is currently the state of the art, but requires valves or windows to be located where the sensors will be used. Considering the relatively low additional cost, it would therefore be the good to consider including for these valves / windows when specifying new power transformers for critical installations .
The power grid and the demands placed on power transformers are certain to change over their lifetime, as will their inclusion in “the Internet of Things” / fourth industrial revolution. Systems for calculating the hotspot via proven robust sensors are now well established and Electronic Temperature Monitoring (ETM) should be considered as the entry point in power transformer digitalization. The ETM selected should provide the foundation for easy trending and addition of whatever future monitoring / sensing requirements may be expected over the transformer’s lifetime. Basic DGA is worthy of consideration also on new transformers, but for any sensors introduced into the transformer’s tank, the risks need to be considered as well as the benefits. Where fiber optics are used then it is worth considering their use for operation as well as the factory acceptance testing and also monitoring of the core. Future requirements for PD monitoring should also be planned for at the specification stage.
The ETM should be considered as part of a scalable solution rather than a device in isolation. Future requirements should consider integration into an asset performance management package, which itself is likely to incorporate not only recording of events but suggestions for actions and resource management.
Last but not least, it is never too early to start recording, trending and comparing key parameters, but monitoring is not a substitute for the use of quality materials, components, engineering and production practices.
1 – McKinsey report for WEF. http://www3.weforum.org/docs/WEF_Game_Changers_in_the_Energy_System.pdf
3 – Transformer Aging Due to High Penetrations of PV, EV Charging, and Energy Storage Applications. 2017 Ninth Annual IEEE Green Technologies Conference.
5 – IEEE C37.91-2000 Guide for protective relay applications to Power Transformers.
6 – Greentech Media, “Why energy resilience matters” Webinar July 28th 2018.
7 – Cigre Guide 393 Thermal performance of Transformers.
8 – IEC 60076-7 Loading Guide
9 – IEEE C57.91 loading guide
10 – ABB Transformer Service Handbook
11 – IEEE Std C57.143TM-2012 – Guide for Application for Monitoring Equipment to liquid-immersed transformers and components
12 – Cigre 343 – Recommendations for condition monitoring of Power Transformers.