Historically, monitoring substation equipment has been based on routine maintenance, which was performed at preset intervals every three to seven years. Outages were planned during those times to enable access to the equipment. With demand for service growing, utilities can no longer afford the luxury of such outages. Therefore, the utilities must employ new methods to determine equipment status without the necessity of scheduling outages. In this respect, to ensure the health of its equipment, Xcel Energy (Minneapolis, Minnesota, U.S.) is migrating toward condition-based maintenance instead of time-based maintenance.
Oil circuit breakers (OCB), consisting of moving and stationary contacts and ancillary components involved in making and breaking the circuit, can wear out and lose the ability to adequately perform their intended functions. This may happen because of misalignment, wear, poor contact surfaces and improper timing of contact movement. Dielectric failure may occur from excessive localized moisture or excessive amounts of conductive particles in the oil. Thermal runaway causes carbonization and byproduct polymeric films to form on conductors, increasing the surface resistance of the contacts, which can result in overheating to the point of failure. Even stationary components such as the interrupter shell and nozzle grid materials can break down, which results in inadequate arc quenching and carbon buildup, compounding the overheating condition.
Various approaches can be used to develop diagnostic tools to detect incipient faults in OCBs. One approach is to identify problems by visual inspection, electrical tests and oil tests to see which ones provide results exhibiting significant changes from comparison tests on sister units. Another way is to study working populations of equipment and use statistical techniques to determine confidence intervals for defining “normal” values for the tests. Using these methods, Doble Engineering (Watertown, Massachusetts, U.S.) developed a set of oil tests to detect OCB problems that can then be further identified by electrical and mechanical tests.
A breaker program determines which oil tests would best evaluate breaker status. An initial group of oil tests in the program, chosen to diagnose the condition of an OCB, revealed some tests were not necessary or did not supply quality or detailed information that would aid in a diagnosis. As a result of that determination, some tests were abandoned and other tests were incorporated into the program, including dissolved gases in oil (DGA), particle count and size, oil quality and presence of metals.
Dissolved Gases in Oil
The DGA test is an important diagnostic tool for detecting localized overheating or excessive arcing. The test also can provide information on other abnormalities. Localized overheating, which may lead to a thermal runaway, is a common failure mode in OCBs. The detection of hydrocarbon gases dissolved in the oil and their relative ratios aid in identifying the condition. Hydrocarbon gases of interest include methane, ethane, ethylene and acetylene. Studies have shown that, as an overheating event develops into a thermal runaway condition, the ratio of the hydrocarbons changes, making this parameter an important element in making the diagnosis. Regardless of some built-in complications inherent in such an analysis, DGA has become an invaluable tool in detecting and identifying problems in OCBs.
Particle Count and Size
The total number of particles by size groupings is used to detect abnormal quantities of byproduct and wear materials. The ratio(s) of the size groupings provides information on the progression of a detrimental condition. Larger particles are especially important because the dielectric strength of the insulating oil can be adversely affected by these particles even if they are not wet.
In OCBs, particles are formed from wear, arcing and overheating. The size, morphology and types of wear particles depend on the severity of the pressure applied to the surface and the angle at which the surfaces intercept. Arcing can also form metal particles, although these particles are different in morphology and topography because they are not the product of mechanical wear. Arc-produced metal particles are formed from molten metal being quenched by the cooler surrounding oil, resulting in a teardrop-shaped particle. Arcing will also produce organic particles such as carbon fines and larger conglomerations from the breakdown of the oil. Overheating increases the rate of decay of other materials and induces the formation of byproducts such as polymerized oil films. Particle typing is performed by passing a prescribed amount of oil through a filter of given porosity. The filter is then washed to remove the oil residue for examination under a microscope. The examination of the filter provides a qualitative identification of the types of particles that exist in the breaker. Particles, in addition to those mentioned above, include fibers, paint and insulating varnishes. The examination is also used to determine the type and details about the particles to include type of wear, type of coating on the surface (when present) and depth of surface coating. In the process of particle typing, there is an attempt to relate the particles to specific materials of construction (Fig. 1). In addition to the microscopic examination of the particles trapped on the filter, a Doble carbon coding process has been introduced to quantify the carbon loading in the sample.
Three oil-quality tests were selected to provide the essential information necessary to diagnose equipment status without being too complicated. The dielectric breakdown voltage test provides information on the insulating capability of the oil. Determining the water content reveals how wet or dry the system is and whether free water exists. The neutralization number, which measures the acidity level, provides information on the extent of the degradation of the oil. The acidity is a factor in determining the severity of a condition because high concentrations of organic acids can exacerbate an already deteriorated condition.
The metals test, consisting of both particulate metals and those dissolved in oil, gives an indication of the amount of material that has been worn from the contacts and is now present in the oil. It also provides a quantitative analysis regarding the composition of the metals found in the oil, such as copper, silver and tungsten.
The Breaker Analysis Program
The purpose of the program is to provide a consistent, reliable, analytical technique to detect OCB problems and to provide a ranking or relative health index, which results in a high confidence level for finding a problem when inspecting a breaker. An additional goal is to eliminate the reporting of false positives, where the analysis might indicate that a problem exists, but when inspected, no defect could be found. The technique was not intended to be overly conservative where problem breakers were overlooked. A statistical analysis, performed on the oil results from specific families of breakers, was based on results from the DGA, oil quality, particle count and metal tests. These data were averaged, the standard deviation was calculated and confidence levels were determined.
To provide a condition assessment for each OCB tank, a numerical ranking was determined based on rankings from the analytical data collected from tests on the DGA, oil quality and particle count. The rankings from the three groups were summed to arrive at the overall ranking, where the highest ranking that could be achieved was 15. The ranking system was further reduced to a Condition Code from which specific maintenance functions were recommended. For example, a point range of 12 to 15 was assigned a condition code of 1, which resulted in a recommendation to remove from service immediately. A point range of 9 to 11.5 had a Condition Code of 2, which required immediate investigation; a point range of 5.5 to 8.5 had a Condition Code of 3, which required resampling in three months; a point range of 2 to 5 had a Condition Code of 4, which required resampling in 12 to 15 months and a point range of 0 to 1.5 had a Condition Code of 5, which required no further attention for at least three years.
A large emphasis was placed on the ethylene to acetylene ratios, which clearly distinguishes the severity of the overheating above certain levels. In general, oil quality, particle count and metal results were not given as much weight as were the gassing results and the ratios because these latter two parameters detect a whole range of problems in the earlier stages and are the most reliable indicators.
Not all of the analytical data were used in the ranking and condition coding schemes, even though they are used in the overall final condition assessment. For example, hydrogen and carbon monoxide were not used in the DGA ranking because they are easily lost to the atmosphere and sampling close to a fault operation may influence changes in their concentration. Oxygen, nitrogen and carbon dioxide represent atmospheric gases and were not included. However, in the final breaker assessment, oxygen and hydrogen levels were monitored to determine if the breaker breathing mechanism was plugged or some other unusual circumstance was present.
The scheme for particles involves the use of a weighted ranking based on size of particles and ratios among different size classes. More emphasis was placed on the larger particles and their ratios to other classes, as they were more likely to affect the operation of the breaker and as an indicator that breaker materials were disintegrating.
Although much emphasis was placed on the statistical analysis, the total point rankings and condition codes do not provide a complete assessment based on the analytical data. In the final analysis, a complete assessment is based on the following criteria:
Calculated breaker condition code
Any single gas, where oxygen has a concentration less than 3000 ppm and hydrocarbon gases are over a specific limit
Presence of free water
Particle counts where 50 to 1000 micron-sized particles are greater than a specific limit, and the >100 micron-sized particles are greater than a specific limit.
Comparison between each of the three tanks in a three-phase system
Visual indication during filter examination of unusual conditions or particles from fibers, masses of fibers, metal fines, large metal pieces, polymerized oil, powdery material and crystalline materials
Other poor test results from electrical tests involving contact resistance or infrared thermography.
Strengthening the program
If a problem is detected when the oil analysis is used, electrical and mechanical tests should be used to obtain further information regarding the condition and its severity. In this context, operational and fault counts can be used for obtaining additional information. For example, if two OCBs have the same total point ranking, but one has 100 fault operations and the other only five, then we must conclude that the one with five sustains more damage per fault than the first and is probably in a more critical condition. Even here, using these breaker operation counts is not as reliable as we might have expected because defective counters, unset counters and lack of knowledge of when the last reset was performed make the counts a potentially unreliable process.
Sampling of breakers is critical for undertaking a diagnostic program because poor sampling will yield nonrepresentative results and skew the interpretation. It has been found that one quart of flush oil must be removed for breakers with less than 300 gal (1136 l) of oil and that are equipped with 1-inch piping. After the flush oil has been removed, an additional quart is removed and a syringe sample is taken for the DGA analysis. For larger breakers, 1 to 1.5 gal (3.8 to 5.7 l) of flush oil is removed, then the quart of oil for the oil-quality test followed by the syringe sample for DGA testing.
Special consideration must be given to breakers that have been maintained and the oil processed, replaced or just returned to the breaker without any remediation. Returning the same oil to the breaker will likely reduce some of the gas concentrations to a small degree and will not affect any of the other tests. In this scenario, the breaker will be in good condition, but the results from subsequent oil analysis may suggest a problem. Filtration of oil before it is returned to the breaker may affect gas levels only to a certain degree. It will also increase the dielectric strength but will not alter the neutralization number by a significant amount. Depending on the condition of the filters, water content can actually be increased instead of lowered, as might be expected. Flushing the breaker with new oil and then replacing the breaker volume with new oil will substantially reduce all levels of gases, metals, acids and particles. Regardless of the oil processing employed after breaker maintenance, the breakers should be sampled to establish baseline levels from which to draw future conclusions.
A Case Study
An Allis Chalmers 23-kV, three-phase breaker, originally placed in service in 1951, was subjected to oil tests to determine its status. The combustible gas content, as detected by the DGA analysis was found to be 2605 ppm. The water content was 9 ppm, and the particle count showed total particles per 10 ml to be 154,036 for a range of particles from 5 to 100 microns. A total point ranking of 10 out of 15 was assigned, which translated into a Condition Code of 2, requiring an immediate investigation.
Based on the particle and DGA data, out-of-service testing was required. Ductor tests did not indicate any contact resistance above normal limits of about 90-110 µΩ. Contact wipe and speed tests did not yield any data that would suggest that the arcing or primary contacts warranted replacement.
Power factor testing, using the Doble M4000 instrument, indicated that the overall to ground watts loss on Phase 3 was high. A result of 0.313 W was obtained using the open breaker test, whereas typical results would be in the range of 0.100 to 0.200 W. At this point, an internal inspection was initiated, requiring the removal of the arc interrupters and arcing contacts. In the process of removing the arcing contacts, it was observed that the copper braids that connect the arcing contacts to the main contacts were discolored (Fig. 2) and that the braid retaining bolts were loose. The gases being generated in the OCB can be directly attributed to the loose braid bolts. It was also discovered that the large amount of carbon particles in the oil caused tracking along the bushing surface to the ground plane, which explained the high watts loss. In this respect, the evidence shows why the contact resistance, contact wipe and Doble tank loss index were in the normal range, while the loss to ground and power factor were both abnormally high.
In economic terms, condition-based maintenance of OCBs makes practical sense because it focuses resources on preventing failure to save substantial amounts of money in terms of replacement and installations costs, as well as to avoid lost revenue. The cost of the program is much less than continuing to service breakers on a prescribed maintenance schedule. The condition assessment program for breakers offers an analytical approach using DGA, particle counting and typing, oil quality and metals analyses. This diagnostic approach helps to quickly identify breakers that are in poor condition, and then focuses on those breakers in a timely fashion. A priority hierarchy can be established that will permit maintenance activities to be developed and managed in an efficient manner, which will save time and money, while preventing outages due to failures.
Wendell Bordash is a licensed Colorado master electrician. He attended Front Range Community College and Red Rocks Community College. Bordash joined Public Service Co. (Xcel) in 1982. He has 15 years experience as a substation maintenance electrician and six years experience as a tester. He also spent three years developing and implementing a circuit breaker condition assessment program.
Lance Lewand is the laboratory manager for the Doble Materials Laboratory and is also the product manager for the Doble DOMINO, which is a moisture-in-oil sensor. Since joining Doble in 1992, Lewand has published several technical papers pertaining to testing and sampling of electrical insulating materials and laboratory diagnostics. He was formerly manager of Transformer Fluid Test Laboratory, and PCB and Oil Field Services at MET Electrical Testing Co. (Baltimore, Maryland, U.S.). Lewand received the BS degree from St. Mary's College of Maryland. He is actively involved in professional organizations such as ASTM D-27 and is a subcommittee chair. He is also secretary of the Doble Committee on Insulating Materials.