Figure 1 Schematic of Transformer T2
Table 1 SCADA
Abstract
The object of this paper is to analyze the activation of two mechanical depressurization devices, the TRANSFORMER PROTECTOR (TP) and the Pressure Relief Valve (PRV), during an internal arc on a transformer installed at the JSC RusHydro Boguchanskaya Hydro Power Plant, located in Krasnoyarsk Krai, Russia. The incident occurred on phase B of the transformer T2, on May 3, 2013.
T2 is a threephase transformer manufactured on June 26, 2008, and has been in operation since May 11, 2012. T2 has a nominal capacity of 400 MVA, but was operating at 360 MVA prior to the fault. The investigations show that the transformer differential protections, the PRV and the TP, were activated. This paper analyzes the data presented by the Boguchanskaya Hydro Power Plant and the information gathered on site, including SCADA records, dissolved gas analysis before and after activation, voltage and current measurements. Based on this information, a series of Computational Fluid Dynamics (CFD) simulations were performed to study the dynamic pressure evolution and static pressure build up inside the tank, and a series of FluidStructure Interaction (FSI) simulations were performed to predict the consequent stresses and deformations on the tank. This article also details the simulation results of transformer tank protections while discussing implications for transformers subject to arcing events. The analysis of this incident demonstrates that the first dynamic pressure peak due to the arc quickly activates the TP, while a sustained pressure for a duration roughly 14 times longer is necessary to open the PRV, which therefore activates with static pressure only. Finally, the analysis suggests that a sealed tank is highly likely to rupture when subjected to a similar arcing event.
Introduction
On May 3, 2013, a fault occurred on a transformer installed at the JSC RusHydro Boguchanskaya Hydro Power Plant, located in Krasnoyarsk Krai, Russia. The incident occurred on phase B of the transformer T2, on May 3, 2013. T2 is a threephase transformer manufactured on June 26, 2008, and in operation since May 11, 2012.
T2 has a nominal capacity of 400 MVA, but was operating at 360 MVA. A schematic of the transformer can be observed in Figure 1. Transformer T2 was equipped with a TP and a PRV. During the investigation, it was observed that the transformer differential protection, the Buchholz relay, the PRV, and the TP were activated.
Analysis of the Event
A) SCADA
Дата 
Время 
Агрегат 
PLC 
Группа 
Устройство 
Событие 
Event 
03.05.13 
19:08:57.083 
ГА2 
2СИС(1) 
Дискретные входы 
ПУ SERGI ТБ. Срабатывание разрывного диска трансформатора (7.04) 
Приход сигнала 
TRANSFORMER PROTECTOR ACTIVATION 
03.05.13 
19:08:57.095 
ГА2 
2СДТТ 
Предупреждения 
Неисправен ввод №1 (~220В) 
Приход сообщения 

03.05.13 
19:08:57.110 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 1. Cработ. дифф. 
Приход 
Differential 
защита ТБ 87T 
сообщения 
Protection Signal 

03.05.13 
19:08:57.110 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 2. Cработ. дифф. защита ТБ 87T 
Приход сообщения 

03.05.13 
19:08:57.112 
ГА2 
2СИС(1) 
Дискретные входы 
Шк. защ. ТБ и ТСН. С2. Газов. защита ТБ 2 ст. Сигн. 63Т2 (13.07) 
Приход сигнала 
First Buchholz Relay Activation 
03.05.13 
19:08:57.125 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 2. Откл. газ. защиты ТБ 2 ступень 63T2 
Приход сообщения 

03.05.13 
19:08:57.125 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 1. Откл. газ. защиты ТБ 2 ступень 63T2 
Приход сообщения 

03.05.13 
19:08:57.139 
ГА2 
2СДТТ 
Предупреждения 
Неисправен ввод №1 (~220В) 
Уход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Аварии 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA006. Перегрев 
Приход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Аварии 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA004. Перегрев 
Приход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Предупреждения 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA013. Перегрев 
Приход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Аварии 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA005. Перегрев 
Приход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Предупреждения 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA014. Перегрев 
Приход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Предупреждения 
Шк. защ. ТБ 1. Сигнал. газ. защиты ТБ 1 ступень 63T1 
Приход сообщения 
Second Buchholz Relay Activation 
03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 2. Cработ. дифф. защита ТБ 87T 
Уход сообщения 

03.05.13 
19:08:57.140 
ГА2 
2СДТТ 
Аварии 
Шк. защ. ТБ 1. Cработ. дифф. защита ТБ 87T 
Уход сообщения 

03.05.13 
19:08:57.149 
ГА2 
2СДТТ 
Дискретные входы 
Шк. соед. ТБ. Сраб. предохр. клап. 2. P в баке > доп. (инв.) (2.07) 
Уход сигнала 
PRV Activation 
03.05.13 
19:08:57.149 
ГА2 
2СИС(1) 
Дискретные входы 
Шк. защ. ТБ и ТСН. С2. Работа предохр. клапана (13.15) 
Приход сигнала 

03.05.13 
19:08:57.153 
ГА2 
2АУГ(1) 
Коммутационные аппараты 
Автомат гашения поля 
Уход из положения 'включено' 
Circuit Breaker Activation 
03.05.13 
19:08:57.167 
ГА2 
2СДТТ 
Аварии 
Шк. соед. ТБ. Сраб. предохр. 
Приход 
клап. 2. P в баке > доп. (инв.) 
сообщения 

03.05.13 
19:08:57.167 
ГА2 
2СДТТ 
Предупреждения 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA014. Перегрев 
Уход сообщения 

03.05.13 
19:08:57.167 
ГА2 
2СДТТ 
Аварии 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA005. Перегрев 
Уход сообщения 

03.05.13 
19:08:57.167 
ГА2 
2СДТТ 
Аварии 
ШУиК вод. охл. ТБ. Заклин. диск. затв. AA004. Перегрев 
Уход сообщения 

03.05.13 
19:08:57.175 
ГА2 
2АУГ(1) 
Коммутационные аппараты 
Автомат гашения поля 
Положение 'отключено' 
Circuit Breaker Fully Open 
According to the SCADA data, the Transformer Differential Protection registered a signal at 19:08:57.110 through a warning associated with its 220 kV windings. However, the TRANSFORMER PROTECTOR activation signal was first registered 27 ms prior. Because the TP activation must succeed the fault, the fault is estimated to be at time 19:08:57.078, roughly 5 ms prior to the TP Activation. Therefore, the Transformer Differential Protection signal was registered 32 ms after the estimate fault origin. The PRV activation signal was detected 71 ms after the estimated fault origin. Finally, the circuit breaker fully open signal was detected 97 ms after the estimated fault origin.
Due to some contradictory data between the oscillograph voltage and current measurements and the SCADA, described in the Short Circuit Energy section of this paper, we have doubts regarding the ability of the SCADA to timely follow all events.
Table 2 SCADA Summary
Time 
Events 
Pressure Calibration 
Time after estimated short circuit origin (milliseconds) 
19:08:57.078 
Estimated Short Circuit Origin (Currently Under Investigation) 
0 

19:08:57.083 
TRANSFORMER PROTECTOR ACTIVATION 
1.2 bar Atmospheric, 17.63 psi 
5 
19:08:57.110 
Transformer Differential Protection 
32 

19:08:57.149 
Pressure Relief Valve Operation 
0.8 bar Atmospheric, 11.75 psi 
71 
19:08:57.175 
Circuit Breaker Fully Open 
97 
B) Short Circuit Location
The short circuit location was identified, among other factors, by burnt cardboard insulation (Figure 2) as being associated with the B phase of the high voltage windings (Figure 3). The arc length is estimated by the transformer manufacturer to be 1 m long by locating burnt sections of the windings.
Figure 2  Burnt Cardboard Insulation
Dissolved Gas Analysis
Figure 3
Location of Short Circuit
Table 3
Dissolved Gas Analysis for Transformer T2
In Table 3, we see the Dissolved Gas Analysis for the Transformer T2. Using the data associated with the date 03.05.13, we may characterize the fault. A general outline is provided in Figure 4 (12).
Figure 4
Gas Generation Scaling with Temperature
One option for classifying the fault is through the Duval triangle, as in Figure 5.
%CH_{4} = 22.82% %C_{2}H_{4} = 42.11% %C_{2}H_{2} = 35.06%
Figure 5
Duval Triangle for Dissolved Gas Analysis (5)
The DGA suggests that the arc may be classified as D2, which corresponds to a high energy arcing event. Because the lines do not strictly intersect, further validation would be useful. An alternative classification system, known as the Rogers Ratio, is defined in an IEEE standard, shown in Table 4 (16)
Table 4
IEEE Classifications for Dissolved Gas Analysis
Based on Table 3 and Table 4, we determine the following gas ratios:
Using Table 4, we may characterize the fault as a high energy arcing event. The IEEE Standard defines arcing temperatures as being between 700 K and 1800 K. The confirmation from the Duval algorithm suggests that the temperature is near, or in excess of, 1800 K.
Short Circuit Energy
The arc energy is defined in terms of the voltage (V), current (I), and time (T) as follows:
In Figure 6, we see the electrical measurements taken 2 ms prior to the short circuit. This is
information will be used as it is the closest set of measurements acquired to the fault. From this figure, the high voltage phase B current peak, Ib_BH, is 4.526 kA. From information provided by the transformer manufacturer, the maximum current on phase B is 4.492 kA. For the purposes of this paper, the short circuit current will be 4.5 kA. However, we are also interested in the voltage across the short circuit. These measurements include information the low voltage side, Ub_HH, but not on the high voltage side.
Figure 6
Electrical Measurements 2ms Prior to Short Circuit
Figure 7
High Voltage Winding Oscillograph Data
One measured potential difference related to phase B of the high voltage windings is 31 kV. It is uncertain across which two points this potential is measured, but let us assume that the two points correspond to the two terminals of the short circuit. Two possible situations are that it represents an RMS value, or a maximum amplitude. If we assume that it is a maximum amplitude, the arc voltage is roughly 31 kV.
If this value is instead an RMS value, it can be supposed that the voltage across the short circuit is 31√2 kV = 43.84 kV. Another set of measured potential differences is in Figure 7, where a line voltage associated with phase B has a maximum amplitude of 45.37 kV. This is consistent
with the interpretation of the value being an RMS voltage. Let us therefore average these two
values: 44.6 kV.
In this paper, we will consider both values, 31 kV and 44.6 kV.
We remark that an empirical measure for the arc voltage, in terms of arc length has been proposed (7). This is shown in Figure 8.
Figure 8]
Relationship between Arc voltage and Arc length
Kawamura uses the upper bound as the more appropriate estimate for the arc voltage. For an arc clearance of 1 m, this implies a voltage of up to 16.66 kV. This is within the same order of magnitude of our values, however more consistent with the 31 kV value.
The SCADA is the electrical output detailing diagnostic events for the transformer. In Table 1, we see a subset of the SCADA data pertaining to critical events. Using this data, we identify the arc duration, the difference between the circuit breaker fully open signal and the estimated fault origin, as roughly 97 ms long. However, the oscillograph of the voltage and current measurements indicate that the arc duration is 65 ms. Because the sampling frequency of the SCADA will be of much lower resolution than the sampling frequency of the oscillograph data, the oscillograph data is considered much more reliable. Therefore, we will use the 65 ms figure to represent the duration of the arc.
We will assume that the stated values for the peak amplitudes of the arc current and voltage are constant across this arc duration, and that the voltage and current oscillate at a 50 Hz frequency (u).
The AC current and voltage can be described as proportional to sin (2nut + $), where $ is the phase. It is difficult to determine the precise phase of the voltage and current at the beginning of the arc. This paper assumes that both the current and voltage start at a phase of 0.1 radians (therefore, approximately 10% of its maximum value). Using equation 1:
The arc energy may be somewhat larger or smaller than these values, depending on the phase of the current and voltage. We will use the higher energy of 6.586 MJ for all presented simulations, as it is the worst case scenario, and therefore the most problematic.
Generated Gas Volume
One paper uses a simplified set of reactions: as oil breaks down, H atoms and CH_{3} radicals recombine to produce gases such as H_{2}, C_{2}H_{2}, CH_{4}, and C_{2}H_{4}, given a gas temperature T (4). The simplified model is shown in equations 57.
An experimental test campaign, at the CEPEL laboratory, was performed on a series of transformers subject to internal arcs (11). These experiments have determined the following dependence for the relationship between the arc energy and generated gas volume:
Specifically, a 6.586 MJ arc produces a gas volume of 3.11 m3 at standard temperature and pressure.
Rupture Disc Aperture
Figure 9
Rupture Disc Aperture
The top layer of the rupture disc is open at roughly 90% the maximum cross section. This result is consistent with a strong arc.
CFD Simulation Background
We are interested in modeling the propagation of pressure waves in transformer oil, when subject to internal arcing events. Such phenomena are modeled as a 3D compressible twophase flow, using a set of partial differential equations based on a 5 equation model developed in (6) and described in equations 6ae. These equations represent the conservation of mass (q), momentum (qu), and energy (E), as well as the advection of the volume fraction (α) for each phase. Source terms relating to gravity (g), viscosity (µ), and heat conduction (T) are added in the conservation equations to adhere to physical constraints.
This model was selected to accurately depict the pressure wave propagation inside liquids and gases. A finite volume algorithm is adopted to transform the system of differential equations into algebraic equations. The fluxes across cell boundaries are determined by the Godunov Riemann solver. The volumes are defined by an unstructured 3D mesh, to allow a precise description of complex geometries such as transformer tanks.
The experimental test by CEPEL was simulated in order to verify the mathematical model developed for arcinduced dynamic pressure peak within transformers (2).
Originally developed and presented, HYCTEP (HYdrodynamic Code for Transformer Explosion Prevention) is implemented as a hydrodynamic numerical tool for computational fluid simulations (3).
The mesh used to discretize the transformer geometry has up to 139,794 tetrahedral elements, and is shown in Figure 10.
Figure 10
Transformer Mesh
The transformer oil and its vapor are represented as a stiffened gas fitted to the mineral oil dodecane (Table 5).
Table 5
Fluid Parameters (14)
Included on the geometry are a TP and a PRV. The TP DC has a 300 mm diameter, and the PRV has a 150 mm diameter. The PRV is set to open at 0.8 bars above atmospheric pressure, and the TP RD opens at 1.2 bars above atmospheric pressure.
The arc parameters used in the simulation are listed in Table 6, and energy is injected using HYCTEP’s arc model 4, which guarantees that the total power input, throughout the arc region, is determined by the product of the voltage and current.
Table 6 Arc Parameters
Max Current 
Max Voltage 
Duration 
Phase 
4.5 kA 
44.6 kV 
65 ms 
0.1 
The simulations were run for up to 900 ms, with a time step of 10–6 s. Four cases were run:
 The actual case where Transformer T2 has both a TP and a
 T2 has only a
 T2 has only a
 T2 is completely
CFD Simulation Results
Figure 11
Average Tank Pressure, Arc Energy = 6. 586 MJ
In Figure 11, the average tank pressure is visualized for the four simulated cases in the case of arc energy of 6.586 MJ. It can be observed that for both cases with a TP, the tank is rapidly depressurized. For the case of T2 with both a TP and a PRV, the average tank pressure first drops below below the approximate static withstand limit of the tank (2.2 bars) after 125 ms.
In contrast, for the case with only a PRV, the average tank pressure does not drop below the static withstand limit until 461 ms, a duration more than three times as long as the case including the TP.
With neither a TP nor PRV, the average tank pressure approaches a steadystate of roughly 15.6 bars, more than seven times the static withstand limit.
In Figure 12, we see the pressure evolution under the three most distinct cases. This figure reinforces our observations for Figure 11. The transformer with a TP only is safely below the static withstand limit by 150 ms, in contrast to the case of the transformer with only a PRV, and particularly the sealed tank.
Figure 12
Pressure evolution of Transformer T2, E = 6. 586 MJ
FSI Simulation
In this section, we simulate the arcing event with a FSI based model. This entails coupling the fluid solver described earlier in this paper to an open source structural solver developed by the French utility EDF, Code ASTER. This structural solver was validated in reference (1). Results from a one way coupling was first described in (9). The algorithm was later refined to allow for a twoway coupling (8). The twoway coupling algorithm was further optimized and experimentally validated in reference (10), and will be used in this paper.
As with the fluid case, we consider cases including a sealed tank and the TP. The PRV has yet to be integrated into our FSI software, so we do not consider this case. The mesh for the tank with the TP is shown in Figure 13.
Figure 13 FSI Mesh
The main tank is 1 cm thick, the vertical beams are 4 cm thick, and two bolted 5 cm thick horizontal beams are modeled as a single 10 cm thick beam. We note that the top of the tank is typically significantly thicker than 1 cm, so stresses and deformations will be exaggerated in this region. The material used is ASTM Steel A345 with a plastic constitutive equation. The material parameters are specified in Table 7.
Table 7 Structural Parameters
Figure 14
Stress Evolution of Transformer T2, E = 6. 586 MJ
It is apparent that the case with the TP is largely within the elastic domain by 60 ms, while large parts of the sealed tank are still deforming plastically at this time.
In particular, the areas near the high voltage bushings are susceptible to rupture. We may conclude from these simulations that protections near the bushings are recommended to mitigate higher arc energy events.
Finally, we note that the top of the tank where there are relatively few reinforcement beams encounters high stresses, but this is expected given the lower tank thickness of this region.
Average pressures for both cases are shown in Figure 15.
Figure 15
Average Tank Stress for Transformer T2, E = 6. 586 MJ
By the end of the sealed tank simulation, the average stress in the sealed tank is 3.435 times as large as for the tank with a TP. It is reasonable to expect that this average stress will continue to increase as energy within the mineral oil is absorbed by the tank structure.
Figure 16
Maximum Tank Strain for Transformer T2, E = 6. 586 MJ
In Figure 16, we show the average tank strain over time. At the end of simulation, the maximum strains in the sealed case are 2.985 times larger than the tank with a TP, indicating that rupture is likely.
Conclusions
A fault was identified in Transformer T2, with a 400 MW capacity, at the Boguchanskaya HPP, equipped with a TP. No permanent damage to the tank was observed.
Due to the dissolved gas analysis, the fault was identified as a high energy arcing event through the insulation. From observations of the current and voltage data, the energy of the short circuit is approximately 6.586 MJ.
Using this knowledge, we attempted to model the sequence of events through a CFD simulation, followed by an FSI simulation. The CFD simulation tool is designed to model pressure wave propagation in a two phase compressible media, while the FSI simulation tool is designed to predict stresses and deformations on the tank. Observing burnt areas of the insulation allowed us to approximate the spatial extent of the arc. Using a schematic of the transformer, a mesh was generated to discretize the geometry.
Four simulation CFD cases were run: Transformer T2 with both a TP and a PRV, T2 with only a TP, T2 with only a PRV, and a completely sealed T2. The two cases including a TP behaved broadly similar, though the tank with both the TP and PRV depressurized below the static withstand limit by 125 ms. In contrast, the case with only a PRV didn’t depressurize below the static withstand limit until 461 ms. The sealed tank reaches a steadystate of 15.6 bars, likely leading to rupture.
We observe that the first dynamic pressure peak due to the arc quickly activates the TP, while a sustained pressure for a duration roughly 14 times longer is necessary to open the PRV, which therefore activates with static pressure only.
Two FSI cases were run: Transformer T2 with only a TP and a completely sealed tank. The sealed tank simulation was only run to 68 ms, so a clear comparison can only be used for this period. By 68 ms, the tank with the TP was largely below the yield stress, though brief and local exceptions occur subsequent to this time. In contrast, the average stress in the sealed tank is
3.435 times higher than the case for the TP. Similarly, the maximum strain the sealed tank is 2.985 times higher than the case for the TP. These conditions reinforce the observation from the CFD simulations, that there is a high probability of rupture for the case of the sealed tank.
From the FSI simulations, we observe that the largest stresses and strains are concentrated near the high voltage bushings. This observation suggests that to effectively mitigate higher energy short circuits, protections for the Oil Bushing Cable Boxes (OBCB) are recommended.
We may conclude that the inclusion of the TP allowed the tank to depressurize very quickly, saving the transformer from explosion. This conclusion has been attested by RusHydro through a TP Successful Activation Certificate (13).
References
 ASTER Official Website, 2013. Presentation, Documentation, and Downloads. [Online] Available at: codeaster.org
 Brady, R. et al., ”Prevention of Transformer Tank Explosion, Part 2: Development and Application of a Numerical Simulation Tool,” ASME PVP, Chicago, IL (2008).
 Brady, R. et al., “Prevention of Transformer Tank Explosion, Part 3: Design of Efficient Protections using Numerical Simulations,” ASME PVP, Prague, Czech Republic (2009).
 Cuk, N., “Oil Tank Explosion Resistance”: Canadian Electrical Association,
 Duval, M., “A Review of Faults Detectable by GasinOil Analysis in Transformers.” DEIS (2002).
 Guillard, H. & Murrone, A., “A five equation reduced Model for compressible two phase flow problems,”, INRIA, Volume 4778 (2003).
 Kawamura, T. et al., “Prevention of Tank Rupture due to Internal Fault of Oil Filled Transformers,” CIGRE, Volume 1202 (1988).
 Landis, B., Brady, R., PetrovanBoiarciuc, M. & Perigaud, G., “Investigation of an Internal Arcing Event in an On Load Tap Changer Using Fluid Structure Interaction”, ASME PVP, Bellevue, WA (2010).
 Landis, B. et al., “Prevention of Transformer Tank Explosion, Part 4: Development of a Fluid Structure Interaction Numerical Tool”, ASME PVP, Prague, Czech Republic (2009).
 Landis, B. et al., “Development of a TwoWay Fluid Structure Coupling for Studying Power Transformers Subjected to Internal Dynamic Overpressures” ASME PVP, Paris, France, (2013).
 Muller, S. et al., “Prevention of Transformer Tank Explosion, Part 1: Experimental Tests on Large Transformers”, ASME PVP, Chicago, IL (2008).
 Ramu, , “Diagnostic Testing and Life Estimation of Power Equipment”. Kent: New Academic Science, 2012.
 RusHydro, “Successful Activation Certificate”, ArkbTP130503RUS59A5824br,
 Saurel, R., Petitpas, F. & Abgrall, R., “Modelling phase transition in metastable liquids: application to cavitating and flashing flows”. J. Fluid Mech., Volume 607 (2008).
 Shrivastava, K., “Data Mining Approach With IEC Based Dissolved Gas Analysis for Fault Diagnosis of Power Transformer”, International Journal of Engineering Research & Technology (2013).
 Transformers Committee of the IEEE Power Engineering Society, “IEEE PC57.104 Guide for the Interpretation of Gases in Oil Immersed Transformers”: IEEE Standards Activities Department,
Biography
Omar Ahmed is currently working at Transformer Protector Corporation as a Research Engineer, where he is modeling the physics associated with transformer explosion and depressurization strategies. He specializes in optimizing the computational fluid dynamics algorithms to model energy transfer from the arcing event to the transformer oil, and the subsequent pressure wave propagation.
Omar completed his undergraduate degrees in Mathematics and Physics at the University of Texas, Austin in 2006, and completed his Masters in Geophysical Fluid Dynamics at Rice University in 2009.
Anne Goj is currently working at Transformer Protector Corporation as a Research Engineer where she spends so much time calculating quantities with physics that she occasionally wonders how her degrees all have “chemistry” written on them.
Anne studied theoretical and computational physical chemistry at Cornell University before relocating to Texas in 2007.