Robert C. Breazeal and Alan Sbravati
Senior Technical Supervisor and Global Customers Application Manager
Southern California Edison and Cargill
In 2018 Southern California Edison began utilizing distribution class transformers with natural ester fluid. In addition, they embarked on a program where units removed from service were retro-filled with ester fluids before reinstallation. It subsequently became clear that the dielectric characteristics of ester fluids significantly affected the conventional parameters used in condition assessment of distribution transformers. In response, SCE embarked on a research program where they conducted a series of controlled experiments including aggressive loading conditions. Two of these studies will be discussed at length. Mineral oil and ester units were subjected to one-year overload test periods, accelerated aged at up to 188% of nameplate rating. The results of the first-year cycle were previously published in . The specimens were evaluated at specified intervals to extract dielectric response and insulation resistance data. This allowed SCE to trend changes in insulation behavior over time in response to varying degrees of overload. DGA and fluid quality tests were performed on each unit to correlate specific test parameters with load level and fluid type. These tests aim both to identify the aging markers and to validate the possibility of safe overload conditions.
Utilizing ester fluids in distribution class transformers. More recently, the California State government began targeting the electrical utilities in an attempt to recover the cost of fighting wildfires allegedly caused by transformer failures and downed power lines. Along with covered conductor, advanced protection schemes, and power shut-off policies, natural ester-filled transformers are used to significantly decreased the litigatory exposure for the California utilities.
A. Year 1 – Retro-filled Transformers
Eleven 12 kV transformers, including 7- 37.5 kVA and 4- 25 kVA units, were obtained from SCE new inventory. The mineral oil was drained from six transformers and later, the transformers were re-filled with natural ester liquid, while the other five remained as original. Using a variac and inductors for regulation, the prescribed voltages were applied to the specimens to generate the specified load currents, corresponding to 115, 125, and 145% of rated capacity.
The short-circuit loading conditions were kept continuously for 4 weeks. The loading was interrupted by the end of the day of the Friday of the 4th week and the transformers were let cool down over the weekend. On Monday morning all SCE team performed all measurements and start the next 4 weeks period. This was repeated 12 times, representing a full year.
B. Year 2 – Factory Filled and Retro-filled Transformers
The presence of 10-12% of residual content of mineral oil and the possibility of localized overheating in some of the units, associated with the absence of indication of paper degradation on the natural ester units, motivated a second round of tests.
At this time SCE received transformers already filled with the soybean-based natural ester. During the planning stage, an alternative supplier of insulating liquids requested to participate in the test, which was accepted by SCE. This round of test was performed with 12 kV 25kVA 1p transformers:
a) 2 soybean ester factory filled units (brand 1);
b) 2 soybean ester retro-filled units (brand 1);
c) 2 soybean ester retro-filled units (brand 2);
d) 2 canola ester retro-filled units (brand 2);
e) 2 synthetic ester retro-filled units (brand 2);
As in the first one-year test, the transformers were connected in groups, targeting 175 and 185% of rated capacity. Periods of 4 weeks of accelerated aging were intercalated with the measurements after a weekend de-energized.
The full set of curves and results are being published in an IEEE Transaction paper.
DISSOLVED GAS GENERATION
For the purposes of this topic, the discussion will focus on an analysis of 10 transformers filled with four types of ester fluids from several different suppliers.
Generation of ethane is a normal occurrence with natural ester fluids. This process is classified as a stray gassing process, meaning it does not indicate an abnormal behavior. In this loading protocol, the canola-based (rapeseed) ester units averaged significantly higher than the other test groups. Conversely, the synthetic ester ethane levels were minuscule in comparison to the other groups, matching typical results of mineral oil units. Methane and ethylene values generally run below 10 ppm for all specimens, although the methane values are elevated slightly for the synthetic oil specimens. Carbon monoxide concentrations are fairly consistent for all units with the exception of unit 8.
EFFECTS OF LOADING ON FLUID QUALITY
The main observations from these properties are:
- The prescribed loads did not appear to have affected the viscosity of the fluids, in that there were no differences between loading groups. Thus, no indication of excessive oxidation.
- H2O concentrations were slightly elevated in the 145% loading group, yet much lower than continuous use limits.
- Dielectric strength does not appear to be affected by the applied loading conditions.
- Neutralization number correlates extremely well with load. It must be reiterated that in regard to ester fluids this is not a particularly useful parameter in that the fatty acids are almost exclusively high molecular weight and are therefore benign in nature.
- Power factor correlates well with load. In this temperature range, this is not an indication of degradation but reflects the generation of polar compounds with momentary dipole characteristics.
Factory-filled 25 kVA pole top soybean ester transformers were loaded at up to 188% of nameplate rating. The retro-filled units were loaded at up to 175% of nameplate. Average winding temperatures were measured at up to 140° C. As previously demonstrated in figure 8 the fluid in these specimens were in pristine condition at the termination of the protocol when the units were properly sealed. Although not tested, it appears that having these specimens vacuum filled and provided with a nitrogen blanket would lead to further enhanced performance. Yet, the minor degradation effects identified after one year of almost continuously loading the units to near double of its capacity is a remarkable result.
In an earlier study where retro-filled ester units were loaded at 145% along with mineral oil-filled units, it was shown that the acidity of the deteriorated mineral oil attacked the copper components causing discoloration. In addition, the plastic components were significantly discolored. The windings of all specimens in this study were disassembled with paper samples extracted from the interface between the primary and secondary windings. Furan and Degree of Polarization analysis was performed on each sample at great expense. Unfortunately, because furans/DP are poor indicators of aging in scenarios involving short-term loading, the data, in this case, gave absolutely no indication of aging in the samples. It should be noted however that when the windings were disassembled the paper samples from the mineral oil-filled units were noticeably brittle when compared the paper from the natural ester specimens.
The interfacial polarization processes also affect insulation winding insulation measurements. The migration of the polar components with the application of the DC potential induces measurable current which biases the measurement. The effects of low-frequency polarization losses create a situation where conventional condition assessment tests become ambiguous.
While this presents a problem for distribution class transformers, the bolstered durability of the cellulose winding components more than compensates for this shortcoming. SCE has experienced a noticeable decrease in the number of distribution transformers salvaged due to overload.
The authors want to thank Dr. Kevin Wirtz for the contributions to the discussion on the chemistry and explanations for the differences between the fluids.