PTPiREE Conference May 12 14 th 2015, Gdansk Poland Monitoring, Diagnostics, or Changing Condition? Mark Tostrud, PE Brian D. Sparling, SMIEEE Ty A. Foren, MBA Dynamic Ratings Inc. Globally, the fundamental design of a power transformer has not significantly changed over the past 100 years. However, the impact and economic value of losing critical load due to an unexpected transformer failure has increased by many orders of magnitude. In today s economy, the availability and reliability of electricity is fundamental to global commerce and our quality of life. Although our dependency on electricity has increased, the investment in the electric utility infrastructure has not kept pace. Figure 1 shows the yearly installment of large power transformers in the United States over the past 58 years. Although it is difficult to estimate how many of the installed units are still in service, power equipment manufacturers estimate that the average age of large power transformers in the United States is approximately 40 years, with 70 percent of units being 25 years or older. [1] The average life expectancy of a power transformer in the United States is approximately 30-35 years, it would appear that a large portion of the US population will soon exceed their life expectancy. Figure 1 SPX Transformers Estimate of the Number of Large Power Transformers Installed in the US (Ref. 7) While these numbers threaten the continuous supply of electricity, the age of the transformer does not always represent the true condition of the asset. Transformer health is best assessed by reviewing the operating and maintenance history of the asset, and performing a condition assessment based on the results of the condition monitoring tests performed on the transformer.
This paper will review the condition monitoring techniques that are frequently used to assess the condition of a transformer and discuss some of the newer technologies being used in these assessments. It will review how the information from condition monitoring serves as the basis for assessing the condition of the fleet and discuss how continuous on-line monitoring data can be incorporated into these assessments to increase system reliability, make better operating decisions and improve the ability to properly manage utility assets. The Problem While the transformer age is factual, it doesn t accurately capture the true health of a transformer. Transformer age alone provides little information on the maintenance history or operating conditions the transformer has experienced over its life. An example of the how misleading an age based assessment can be, is evident in Figure 2 which shows a Health Index calculation versus Age for a large population of power transformers. In this set of data, the average reduction in transformer health with age is very small, and the scattering of results is very large, indicating that age should not be treated as a critical factor for health assessment. [2] Figure 2 Health Index vs. Age for a Large Population of Power Transformers Further, age provides little value to assess a fleet of transformers, given each asset sees individualized service history, time of use and loading. Condition based ranking methods provide a much better indication of transformer health. This is particularly true with system components such as bushings and On-load Tap Changers (OLTCs). These two components are well documented as the leading causes of unexpected failure of transformers. When assessing the health of a transformer, the type of defect and risk of failure should be considered in the assessment.
What is condition monitoring? Condition monitoring is the process of observing a certain aspect of a machine s performance to detect changes that may indicate a developing fault. By identifying a defect at an early stage, it can be corrected before it leads to failure. http://www.engineeringexchange.com/profiles/blogs/what-is-condition-monitoring While condition based monitoring is not a new idea, utilities are more dependent on the information measured from condition monitoring than ever before. Condition monitoring of transformers can take many forms ranging from online to offline tests. Many of the condition monitoring technologies used today were implemented following a review of transformer failure modes in reliability centered maintenance studies. Many of these tests were implemented to help improve the availability and reliability of the equipment by identifying problems early. Some of the more common condition monitoring tasks performed on transformers are shown in figure 3 below. Avoid relying on off-line monitoring Figure 3 Typical Condition Monitoring Tasks for Transformers The off-line monitoring methodology many are using to assess the condition of their transformers is based on the latest maintenance and test data available on the transformer. This is static, historical data. While this type of assessment is better than doing nothing, the accuracy and relevancy of off-line assessment begins to deteriorate immediately after it has been completed and the condition of an asset can vary greatly between assessment intervals. Unexpected failures are a fact of life with any equipment. Any living asset management program should have the ability to adjust when unexpected events occur. Hence, in many cases, there is little chance of detecting the problem unless some form of continuous on-line monitoring is used. Continuous (on-line) monitoring, data collection and analysis provides the basis for the development of a dynamic transformer health index and a living asset management program.
Only on-line monitoring can drive a condition assessment and an Asset Health Index System Without data an index cannot exist. An index is an indicator, sign, or measure of the item of interest. In a static example, "Exam results may serve as an index of Teacher effectiveness." In a dynamic on-line industry example, On-line asset condition data, combined with the traditional off-line condition assessment data and the substation environmental conditions, serve as the variables necessary to properly index of the transformers Condition, life, and future performance. On-line Monitoring serves as a benchmark of an asset, allowing numeric assignments, ranking and evaluations Enables Condition ranking over time Safety - Eliminates the timed service trips - Environment Considerations Failure Prediction Drives Prevention - Planned Outages - Social / Political Impacts - Increased Uptime - Allows for thorough Root Cause analysis, when necessary Maintenance Optimization - Asset Health could drive proper spare parts inventory - Maintain, that which needs maintaining Asset Renewal Prioritization - Use of actual data to support renewal spending Latest Trends in On-Line Monitoring Many of the technologies employed as part of a condition based monitoring program for transformers have evolved to the point where they are available on-line. Continuous, on-line monitoring is now commonly accepted and has demonstrated it can provide the accurate and reliable data required for the safer and more efficient operations of the electrical power system. Today s monitoring technology feature powerful microprocessors, using well-defined concepts and equations from published standards such as IEC and IEEE. The systems use non-intrusive sensors which continuously measure environmental conditions and most importantly the parameters of the transformer and its five major components being: the Active Part, Oil, Bushings, On Load Tap Changer and the Cooling System. The data from these sensors is securely collected and turned into the useful transformer information network asset owners/operators need to operate their system more safely and make better technical, maintenance, and hence, business decisions. While on-line monitoring was once considered a nice to have, most utilities recognize the role it must play in the grid for the next generation. The younger generation has been trained to expect data and information on demand and readily available on their laptops and smart phones. These expectations are driving the need for more on-line monitoring and better tools to assess the data.
Online DGA monitors have been around for over 20 years and are now specified on the majority of large power transformers in the US. Other technologies like on-line bushing and partial discharge monitoring are not as popular (as a routine test) but are gaining in popularity due to the technologies ability to identify problems at an early stage. On-Line Monitoring Example Figure 4, shows the data from an on-line bushing monitor that was installed on a 69-12.47kV transformer in July of 2010. The off-line power factor test performed at the time of installation showed the bushings were in good condition. Undetected by the off-line system, the on-line bushing monitor recorded imbalance currents which exceeded the warning and alarm set points 15 days after installation alerting the customer to a problem. The increase in imbalance current appears to be intermittent, it is easy to understand why an off-line test performed at the wrong time may not identify a bushing problem. Figure 4 69kV bushing problem identified with on-line monitoring Following the initial alarm, the customer chose to monitor the bushings to see if the imbalance current continued to increase. After the large increase on September 28, the customer contacted Dynamic Ratings to ask what other monitoring could be performed to confirm the alarms being reported by the on-line bushing monitor. A partial discharge (PD) monitor was installed and confirmed the presence of PD in the bushing. Figure 5 shows the data recorded by the monitor which confirmed the presence of a high level of PD immediately after installation and recorded a large increase about 10 days later. Figure 5 PD activity detected in a 69kV bushing using on-line monitoring
Following this analysis, the customer removed the transformer from service and retested the bushings. The B phase bushing power factor was measured at 2.25% confirming the alarms from the bushing and PD monitors. To successfully manage risk, a combination of on-line and offline data should be used to properly assess the condition of an asset. Use of the Data to Understand the Condition of the Fleet Transformer Health Index A Health Index assessment is a practical method used to quantify the current condition of a transformer and allow for comparison across a fleet of transformers. Among the benefits of a Health Index assessment is the resulting classification of transformers, which helps determine capital spending priorities as well as the planning and prioritization of maintenance activities. In addition to looking at the types of failures and how they can be monitored, the probability of failure should also be considered. The purpose of a Health Index assessment is best understood when considering the overall transformer fleet life management process. When a defined population of transformers is considered, the preliminary Health Index assessment is made from test data and operating observations that are readily available. The resulting index allows ranking of the units to identify those that need further testing and inspection in order to make an informed decision on the corrective action that should be taken. The problem with most health index assessments in use today is they were developed based on static data. Ranking is a method often used in condition-based maintenance (CBM) to prioritize the work that needs to be done over the entire system population. The method helps to ensure that valuable O&M dollars are being spent on the maintenance that will provide the most benefit. In the previously published paper Transformers Condition Based Ranking, [3] the following conditions were presented as a way to describe in layman s terms, the impact a defect may have on the health of a transformer: Normal condition, no special actions are required Defective/faulty condition, which affects long-term reliability Defective/faulty condition, which affects long-term reliability; displaying symptoms of fast deterioration Defective/faulty condition, which may affect short-term reliability Defective/faulty condition, which may affect short-term reliability; displaying symptoms of likely catastrophic failure The summary in Figure 6 is a method that has been used to reinforce the condition of the equipment when making decisions on which maintenance tasks to perform or which assets to maintain. [4] Figure 6 Overall Summary of a Health Categorization Each of the five major components of the transformer: the Active Part, Oil, Bushings, On Load Tap Changer and the Cooling System are assessed separately and summed to create a composite score for the transformer. This ranking concept is similar to the concepts proposed in an earlier paper [3] which recommended using quantifiable condition assessments and multi-relational algorithms based on prescribed test ensembles to generate condition codes. The suggested codes are quite applicable for condition-based ranking and often serve the basis for the transformer health index calculation.
On-line and off-line data, drive a Comprehensive, Combined Health Index (CHI). Figures 7 illustrates the various on-line and off-line testing and monitoring methods commonly used to measure, detect, and diagnose the health of a transformer. It can be observed that none of the methods, used individually, can measure or detect all of the various symptoms of a potential failure condition. However, when all methods are used in combination, a Dynamic or living Health Index can be developed allowing near real time evaluation of the equipment health. One question remains, How to weigh one component against the other components in terms of criticality to the systems health? even if only small but critical changes in condition are occurring. Techniques for Failure Detection & Diagnosis at The Transformer CEA Outage Rate by Component Contribution 21.6% 4.4% 29% 0.04% 0.03% 8.70% 7.20% 28.40% Monitoring with Advanced Analytics, Statistical Analysis of ALL data Continuous On-Line Monitoring Periodic In Service Inspections (No Continuous On-Line Monitoring) Transformer Out of Service Transformer Disconnected & Out of Service Transformer In Service Analysis Machine/Complex Rules Analysis Human/Simple rules Diagnostic Method Bushings (including CT's) Windings Figure 7 Use of Offline and Online Information to Determine 'Health' OnLoad TapChanger (OLTC) Core Leads Cooling System Visual Inspection E E X X X X X X (X) (X) X Furans P E P e X X X X X X Lab DGA P E P e X X X X X X X X (X) (X) Lab Oil Physical testing P E P e X X X X X X X X (X) PF & Capacitance testing e e X X X X X SFRA E E X X X X X X (X) FDS E E X X X X X X Excitation test e e X X X X X Winding Resistance test e e X X X X X X TTR e e X X X X X Infra-red camera Inspection (in service) E E X X X X X X X PD Diagnostics (temporary in service as required) E e X (X) X X X X On-Line DGA Monitors, Main tank E e X* X X* X X X (X) On-Line DGA Monitors, OLTC tank E e X* X X* (X) On-Line Data Computations Winding Hot spot e e X* X X X* X X Moisture in paper e e X* X* X* X X* X Moisture in barrier e e X* X* X* X X* X Ageing e e X* X X* X Cooling system efficiency e e X* X X X* X OLTC tap Position Tracking e e X* X X X* X X* OLTC Temp Differential e e X* X X X* X OLTC Motor Torque E e X* X X* X X* Partial Discharge On-Line E e X* X X* X (X) X (X) X PF & Capacitance, Bushing Monitor E e X* X X* X Protection Relays/ Digital Alarms/SCADA E E X X X X X Auxiliary Equipment Other True asset management requires access to continuous information to be able to adjust to the changing system and equipment conditions and electricity demand. A Health Index comprised of historical data combined with continuous on-line monitoring of the critical components provides system operators and asset managers the information they need to properly manage the risk. Building off Figure #7, Figure 8 illustrates a transformer in very good health with a score of 0.1, taking into account the seven major variables that comprise a transformer system using both on-line and off-line measurements techniques. A score of zero signifies perfect health. The scale used in this system is a zero to one scale, with one representing the worst case condition.
Figure 9 considers and illustrates a second transformer with score of 1.0, indicating problems from the thermal and oil points of view. Normally these two items have a low weighting factor, but as the individual score for these two components worsens, the weighting of those components changes and affects the overall health score. Figure Figure 8a - An 8 A Unhealthy healthy unit Unit Figure Figure 9 An 8b unhealthy - A Healthy unit Unit This method draws attention to components driving the assets component and system health. Details of the score, as well as guidance toward the location the problem, are presented to assist in bringing the unit back to normal health. This method provides an easy way for operators / owners to assess whether the transformer health issues are due to a single component that has reached an alarm level or if it is due to a series of components that are approaching the warning level. A dynamic combined health index provides greater opportunity to manage risk by allowing the ability to change the weighting factor as conditions worsen. Weighting factors are often set for parameters when things are operating normally. In the normal course of events, each piece does not contribute much to the overall score. However, when the condition of a piece becomes much worse, (in others words enters the warning or alarm range), its overall contribution to the final number may be swamped by all the rest. There needs to be a way to bring that variable going out of control to a higher more noticeable level. For example, in the early stages of a bushing failure, one may see partial discharge activity but may not see an increase in the bushing capacitance. If the PD activity is low, it may fall below the warning or alarm thresholds and would typically fall below the radar when assessing the health of the equipment. However, if the PD activity reaches the warning or alarm threshold, the use of a sliding or logarithmic scale can raise the awareness to the problem. While transformer health plays an important role in the asset management process, as suggested in the paper [3], criticality should also be used to rank the asset s importance and the impact a failure may have on a system. Criticality Index All transformers do not have the same importance on the network. In many substations, there is sufficient spare capacity to carry the load during a forced or planned transformer outage. However, in several cases, the normal load is such that, the loss of one unit would require the overloading of remaining units or the shedding of some of the load. While the type of load served and system redundancy may have a significant impact of the criticality index score, there are other parameters that should be considered. One example is the replacement of a failed transformer can be quite different depending on the availability of spare units. In this example, spare parts availability become a critical parameter of consideration to the owner / operator. Likewise, the availability of spare parts can also impact whether a transformer repair is even possible or if replacement is required.
Criticality weighting adds an important dimension to asset health indexing, that being a magnitude of risk weighting which can be quite different depending on the transformer location in regard to public safety or workforce safety. Examples of parameters that should be considered and how these parameters may impact the overall health index are given in Figure 10. Parameter Description Weight Spare unit availability N-1 capability, spare on site, spare remote, no spare 20% Energy not generated Applicable for GSU only; score is a function of ease of replacement of missing generation 20% (if applicable) Access for repair In some locations, access might now be difficult 10% Availability of spare parts Some manufacturers have disappeared and replacement is difficult for tap changers and/or bushings. 20% Impact on public safety Proximity of public entails additional risk 15% Impact on environment Possibility of oil spill in river or water resource increases consequence of a failure Figure 10 Criticality Index Parameters 15% Risk Index The overall risk index for each transformer (as a system of it components) is a combination of the Health Index and the Criticality Index. Where: Risk Index = (Health Index) x (Criticality Index) (1) Risk Indexing permits ranking of the whole population being considered. A population of transformers can then be plotted and classified by the distance separating each one from the upper-left corner of the graph as shown in figure 11. Further, the Health Index is 51 percent and the Criticality Index is 33 percent. However, it is easy to understand that this same model of transformer, located on another part of the system could have a much different risk index. Likewise, a sudden change in transformer or bushing health, could also drive the risk index higher. The Opportunity to Do More with the Results Figure 11 Risk Factor as a Function of Health and Criticality
A Risk Factor analysis was performed on a fleet of GSU transformers installed at numerous combined cycle generating stations. Figure 12 presents details of an assessment on this fleet of GSU transformers. This analysis takes into account the condition (using the categories presented in Figure 6), the importance criticality at each station, and the availability of spare transformers. This last rating accounts for spares parts statuses of the company; multiple spares that are available for purchase (open market), and specific spares available for purchase, or the absence of spares. Figure 12 Use of Condition Scores, Spares Available or Not, and Importance to the Plant From this graphical representation of the situation, attention is quickly drawn to units 26 and 27, due to lack of spares, and the data clearly shows that these units are considered to be in very poor condition as compared to the others in the fleet. The data score is a computation of the relative importance of the plant to the company according to their proprietary method. One of the outcomes of the deeper analysis was the fact that contingency plans could be developed based on the availability of spare transformers to back up their fleet in the event of an unexpected failure. When considering the availability of spare transformers in addition to the transformer health index, it is easy to see which generating units in their fleet were most vulnerable. The analysis took into account the number of spares (ratings, and fit for purpose), criticality (in terms of function at the station), and where they were located. In some of these cases, the data score relates to the GSU connected to the steam turbine. In some stations, that GSU is more important because if it fails unexpectedly, the whole station must come down, because running on the gas turbines alone would increase the allowable emissions to unacceptable levels and lead to financial penalties. In other cases, the cost associated with purchasing the lost generation may be significantly higher at one site than the other. Likewise, the risk index could also be considered when selling power on the open market.
Summary Utilities need to recognize that living with and tending to an aging system is the new normal. They need to make the transition to that new normal before they get mired too deeply in the quicksand by developing and institutionalizing the skills, organization, IT, equipment, tools and operational resources required [6] Some have said, the overall condition of a transformer is only as good as its worst performing component. On-line monitoring of components, combined with regular Dissolved Gas Analysis (DGA) and on-line bushing monitoring (or with on-line DGA monitors and on-line Bushing and Partial Discharge (PD) monitoring of the active part), has been demonstrated to be an effective contribution to the Health Index evaluation of the unit. For healthy transformers, the advantage of on-line monitoring is the detailed on-line model calculations that provide the engineer and operator with knowledge about the actual real-time operating condition of a transformer. These models provide the necessary information about moisture content in the insulating system of the transformer (both in the oil and the paper), efficiency of the cooling system, and the rate of aging of the transformer as well as the overloading capability of the units. For unhealthy transformers, when a unit does exhibit symptoms of failure, on-line monitoring provides early warning and should drive the response to carry out further diagnostic testing. Data values, in conjunction with diagnostic models, are used to detect changes from normal behavior. With this capability, catastrophic failures can be considerably reduced and repairs can be scheduled as required, rather than in a panic. The immediate benefits of moving to a diagnostics-based management strategy for critical transformers would include: Better knowledge of the state of the transformer fleet, so as to be able to plan and resource maintenance, servicing and eventual replacement. Better information on developing technical problems, allowing immediate remedial actions to be undertaken hence reducing the risk of catastrophic failure and unplanned network outages. Better management of capital and revenue expenditures, allowing resources to be better focused upon network infrastructure development. Improved predictability of network performance and hence, reduced social and political impact of unplanned network supply interruptions. The benefits above will have a positive impact on the reliability and availability of critical transformers in the Polish Grid. At the same time, continuous monitoring has the promise of reducing the severity of repair costs of a failed transformer due to early detection. Early detection and correct diagnosis of problems as demonstrated by other developed nations, leads to failure predication, outage prevention, maintenance optimization, asset renewal prioritization all steeped in real-time data to ultimately increase safety and lead to continuous commerce and quality of life.
References 1. Office of Electricity Delivery and Energy Reliability, Large Power Transformers and the US Electric Grid, US Department of Energy Study - June 2012 2. A.N. Jahromi, R. Piercy, S. Cress, J.R.R. Service and W. Fan. An Approach to Power Transformer Asset Management Using Health Index, IEEE Electrical Insulation Magazine, March/April 2009, Vol. 25, No 2. 3. Jacques Aubin and Brian Sparling, Determination of Health Index for Aging Transformers in View of Substation Asset Optimization, presented at TechCon North America 2010, Ponte Vedra, FL (March 2010). 4. Victor Sokolov and David Hanson, Transformers Condition Based Ranking, presented at TechCon 2006, Asia- Pacific. 5. Cigre WG A2.44 Draft Guide on Transformer Intelligent Condition Monitoring (TICM) Systems 6. Lee Willis, T&D World, Utility Infrastructure - Old and Getting Older, Aug 20, 2014. 7. EIA: SPX Electrical Products Group Yearly Installment of Large Power Transformers in the United States from 1948 to 2006, May 2010 About the authors Brian D. Sparling, Senior Member of IEEE is a Regional Manager with Dynamic Ratings Inc., serving customers globally from Surrey, BC, Canada. Mark D. Tostrud, PE, is the Technology Officer with Dynamic Ratings Inc. located in Sussex, WI., USA Ty A. Foren is the Global Marketing Manager with Dynamic Ratings Inc. located in Sussex, WI., USA