By Glenn Swift and Tom Molinski
Introduction
A recent two-day seminar in Toronto entitled the Life Cycle Management of Power typifies the current interest in determining the condition of in-service power transformers, and in
minimizing both the cost and risks of keeping them in service and the risk involved.
One thrust of this effort has been to improve on-line gas-in-oil analysis. Another is the investigation of the effects of a variety of impurities other than dissolved gasses. A task force has generated a report summarizing some of these approaches and other factors that lead to transformer failure.
An approach that relates loading to insulation degradation effects is the ANSI/IEEE Standard Guide for Loading Mineral-Oil-Immersed Power Transformers, that defines a relationship between winding temperature, sometimes called 'hot spot temperature', and rate-of-loss-of-life of a power transformer. While many would argue that the relationship is imperfectly defined, it is nevertheless used by many utilities, as an overloading guideline. In fact, in 1991, the Standard was reviewed and revised slightly to include transformers above 100MVA.
An indication of the validity of the Standard's hot spot temperature calculation equations can be found in a 1995 Canadian Electrical Association study which concluded that "The thermal model provided in the literature [the Standard] for transformers with natural oil circulation is quite adequate to describe the thermal behavior of a transformer subjected to a variable load."
The equations are usually applied over a period of a few hours, or a day, because of their complexity. Here, it is shown how to apply the equations in a rational way over much longer periods: years, or even decades. A year-long temperature variation model was developed to facilitate the calculations. This made it possible to track the lifetime of a transformer back into the past and/or forward into the future. This, combined with economic analysis, allows criteria to be developed regarding the best time at which to replace a power transformer due to load growth, i.e. to minimize the cost without significantly increasing the risk.
The transformer replacement analyses considered in this paper are for power transformers greater than 30 MVA on the Manitoba Hydro power system. These transformers typically provide connections between major system voltage levels or high capacity supplies to sub-transmission systems. In the Manitoba Hydro system, the normal permissible loading levels of these transformers are 100 per cent of their maximum name plate rating for either winter or summer peak. The permissible contingency loading levels are 125 per cent of rating for winter peak and 100 per cent for summer peak, assuming a summer ambient temperature <30¡C and a winter ambient temperature <0¡C. The basis of this is that during normal or emergency loading, the loss of life of the transformer over a complete day must not exceed the normal daily loss of life as defined in the IEEE Standards.
Figure 1 illustrates the fact that a high rate-of-loss-of-life for a short time is not a significant factor. Over the long term, it is the integrated rate-of-loss-of-life that is significant. Over the short term, it is the peak hot-spot (winding) temperature that matters, and whether or not it has caused undesirable contaminating products to form.
The recent trend for many utilities is to reduce capital spending. One area that is being closely scrutinized is spending on power transformers. Most utilities want to make efficient use of the transformers without creating operating or maintenance problems. Load growth causes increased loading on transformers, necessitating the procurement of new transformers. Utilities may opt for increased transformer loading, which could lead to other transformer problems that were undetected at the lower loading levels.
Loss-of-life Analysis
In order to use thermal loss-of-life equations (not given here because of lack of space) one must either know or assume a loading pattern and an ambient temperature pattern. There are both daily and yearly aspects to these as discussed below.
Load Model
The daily load variation for Manitoba Hydro and many other utilities has two common forms: a single hump shape and a double hump shape. A typical double hump shape is shown in Fig. 2. Historical data along with 'load shape' models based on the nature of known customer loads were used in this study.
The yearly load variation shown in Fig. 2 includes (for the case used in this study) a two-week outage of the transformer in parallel with the transformer being studied, at the time of the peak winter loading: a worst-case scenario.
The last part of the load model is the load growth over years. Load growth for each year into the future is estimated from known factors, for example planned industrial installations and geographically-related load patterns.
Ambient Temperature Model
Similar principles apply here, except that the variations throughout a day and throughout a year are based on long term averages published by Environment Canada. Sinusoidal approximations are shown in Fig. 3. Note that these have to be determined for particular regions. In Manitoba there would be two such regions; one for Northern Manitoba and one for Southern Manitoba.
Hot Spot Temperature and Loss of Life
The winding hot spot temperature is usually the principal factor limiting the "loadability" of a power transformer. Higher winding hot spot temperatures cause degradation of the winding insulation material which can result in the formation of gas bubbles that facilitate the dielectric breakdown characteristic of the transformer oil.
Industry standards recommend that during rated load, the temperature of the winding hot spot should not exceed 110¡C or 80¡C rise above ambient (with the ambient daily average temperature of 30¡C). These temperatures (24hr/day) result in what is defined as the normal loss of life for the power transformer, which works out to be 0.0369 per cent per day.
The loss of life is related to the thermal degradation of the insulating paper. For paper insulation, the end of life is defined at the degradation point where the paper has lost half of its mechanical strength. The life of the paper insulation is then only 7.42 years at a continuous winding hot spot temperature of 110¡C which increases to 50 years with a continuous winding hot spot temperature of 92¡C.
For contingency overload conditions (few days), the industry recommendation is not to allow the winding hot spot temperature to exceed 140¡C in order to limit the risk of releasing gas bubbles. However during certain emergency overload conditions, allowing a maximum winding hot spot temperature of 160¡C, or even 180¡C for a short duration (few hours) may be an acceptable risk for very infrequent occurrences.
Since not all utilities have the same transformer loading criteria, the maximum allowable winding hot spot temperature limits tend to vary a great deal. This paper outlines a methodology to determine the maximum winding hot spot temperatures, so that transformer loading criteria can be safely rationalized taking into account that each utility can accept various levels of risk.
Case Study Results
To better illustrate the argument, an actual case study will now be presented: the Manitoba Hydro Minitonas Terminal Station Bank No. 4. Historical records and future load forecasts were used as data for the plot of peak load in Fig. 4. Temperature variations throughout each year were assumed to follow "standard patterns" as described earlier. From this data, peak hot spot temperature and accumulated loss of life were calculated. These results are also shown in Fig. 4.
The present policy is that a transformer will be taken out of service and replaced with a larger unit when the peak load in this case (during winter when ambient temperature is less than or equal to 0¡C) exceeds 1.25 per unit. It can be seen that the peak load would have reached about 1.25 in 1996. In anticipation of this, replacement actually took place in 1995.
When planning for the future, the loading estimates are based on system load forecasts. Winding temperatures and loss-of-life are calculated in the same way.
The accumulated loss of life plot in Fig. 4 up to the replacement year (1995) is very low; less than 1 per cent. From the peak winding temperature plot, notice that keeping the transformer in service another fifteen years, to the year 2011, would take the hot spot temperature to just about 150¡C, the temperature at which it is sometimes assumed that gas bubbles may start to form in the oil. Correspondingly, the accumulated loss of life by the year 2011 is only 4.5 per cent. It is only in the year 2020 that the accumulated loss of life reaches a full 100 per cent.
Economic Analysis
Reducing the life cycle costs of power transformers involves an evaluation of all present and future costs over the expected life of the transformer. Present value analysis is used to convert all future costs to equivalent present costs. The scenario with the lowest present value cost is the lowest life cycle cost of the power transformer. The procedure and formulas used in the present value economic analysis are not given here for purposes of brevity.
The premise of the engineering economic study was to utilize the present value method to examine the cost impacts (savings or debits) caused by delaying the scheduled replacement date of a new transformer, by allowing the existing transformer to remain in service after it has exceeded the existing Manitoba Hydro transformer replacement criteria. These studies were completed on a 50MVA transformer that was replaced in the Manitoba Hydro system in 1995.
The scenarios involve delaying the replacement date of the replacement transformer itself from 1-16 years or from 1996-2011 (2011 is the planning horizon year for a 35 year economic study on a transformer initially purchased and installed in 1976). This is accomplished by estimating the "cashflows" for each scenario from 1995 to 2011. The net cash flows of all costs and residual (salvage) values are discounted for each year to bring them back to present values, which are in 1994 dollars in this study.
Note that the financial end of life (35 years in this study) is usually not the same as the technical end of life.
The latter is the life over which the transformer will be permanently taken out of service (failed, not repairable, etc.) usually 35-50 years in Canadian utilities. The financial end of life is that time over which the transformer's residual value is depreciated to zero, usually less than the technical end of life.
Associated Costs
The study included the following costs and residual values associated with the transformer replacement:
1.Purchase and installation (including Engineering) costs of a new (93.7MVA) year the replacement occurs.
2.Residual value of the existing (50MVA) transformer in the year replacement occurs
3.Residual value of the new transformer at the end of the study horizon, 35 years from the replacement date of the existing transformer .
4.Purchase and installation of a sophisticated on-line dissolved gas analyzer (DGA) and a sophisticated transformer protection and monitoring device (relay) in the first year of deferment of replacing the existing transformer.
5.Salvage value of the DGA and relay when the existing transformer is replaced by the new transformer.
6.Costs associated with the increased load, assuming there are no load losses as a result of retaining the existing transformer in service longer than originally planned.
It should be noted that operating and maintenance (O & M) costs for planned maintenance were assumed to be the same if either the existing or the new larger transformer were placed in service and were therefore not included as a factor in the analysis. The unplanned O & M costs (repairing oil leaks, etc.) are an infrequent occurrence (e.g. one in 20 years) and were also omitted in this analysis.
Failure costs are also assumed to be the same for either the existing or new transformer and were omitted. This is based on the fact that the statistical failure rate for power transformers (loaded < 160 per cent of rated) during the study period is the same for newer or older transformers as shown in Fig. 5. A detailed probability of failure analysis has been performed, confirming this statement, using the method of reference .
When compared with other factors such as transformer loss of life analysis, winding hot spot operating temperatures, risk, and other vital non-monetary factors, the net present value will furnish the utility with the knowledge required to identify the least-cost scenario.
It has been determined by others that failure costs can become increasingly important at load levels above 160 per cent of nameplate rating. However for the most part our studies are at load levels below 160 per cent of nameplate.
Associated Sensitivities
It is recognized that there is an uncertainty regarding future interest, escalation and load growth factors. This study utilized load growth, interest, and escalation rates forecasted by the Manitoba Hydro corporate experts. More present value studies will be completed to study the sensitivity of the cost savings due to various load growth factors, interest and escalation rates, over different economic study periods of 35, 40 and 45 years.
Various methods of calculating depreciation are possible. The sinking fund method was used here, because it most accurately reflects the true market value of the transformer being replaced.
The residual values of the existing and replacement transformers, and the relay and dissolved gas analysis equipment are all taken into account.
Results of Economic Analysis
As anticipated, the results of the present value economic study indicated that significant savings could have been realized by delaying the 1995 replacement date of the new transformer by several years.
The results of the study are shown in Fig. 6, where the curves represent the anticipated total net savings by delaying the replacement date of the new transformer from 1 to 16 years. In general the longer the replacement date delay, the greater the savings. The savings occurring in year 2011 are $490,000 in 1994 present value dollars, even when the cost of a sophisticated monitoring relay and dissolved gas analysis equipment are taken into account.
Although it may not be prudent to attempt to 'significantly and safely' overload a power transformer without the relay and dissolved gas analysis (DGA) devices attached, an option to consider is to allow "moderate" load increases say from 125 per cent to 135 per cent and not include these cost-saving monitoring devices.
Protection And Monitoring System
Even though this analysis indicates that transformers may technically be kept in service longer than is the present policy at Manitoba Hydro, it is not disputed that there is a very small increase in the likelihood of failure of the apparatus. It is therefore recommended that if the longer life policy is adopted, a more elaborate protection and monitoring scheme is justified.
A highly recommended form of monitoring is on-line dissolved gas analysis, and several such devices are commercially available.
A new multi-function transformer protection and monitoring system is yet another "line of defence" that provides not only the usual protection functions (differential protection, etc.) but also continuous monitoring of potentially damaging conditions such as the aforementioned dissolved gas analysis records, high harmonic current content, through-fault current stress, top oil temperature, hot spot temperature, and rate and accumulated thermal loss of life conditions.
In the case studied here, the cost of these protection and monitoring devices is well below the anticipated transformer replacement cost savings, especially considering that such devices can probably be re-used.
Conclusions
This paper demonstrates a methodology to determine the savings that can be realized by keeping power transformers in service longer than is the present practice.
The final recommendation for the particular case studied here was to delay replacement by nine years. This was considered to be a judicious choice since the loss of life and hot spot temperature both start to increase exponentially and the 'savings' curve starts to flatten out.
It is shown that it is more economical to overload existing transformers and accept the penalty of increased loss of life, than to relieve the loading by installing larger or more transformers. It is recognized that such a new policy, if followed, will lead to a greater dependence on the short-term or emergency overload capabilities of existing transformers and might increase the use of mobile transformers in the event of an outage or failure.
Based on loss of life and probability of failure analyses, it was determined that the risk of failure due to overloading for loading up to 160 per cent of rating, is very small.
Sophisticated on-line transformer monitoring and relay systems for dissolved gas analysis and other important transformer parameters can be used as an added information source to safely improve transformer loadability and instill confidence in a longer in-service life policy for large transformers. A second benefit is that these devices are capable of alarming to indicate that a potential problem is developing so that it can be dealt with before serious damage occurs. ET
Glen Swift is with Alpha Power Technologies and Tom Molinski is with Manitoba Hydro