Impact of Oil Velocity on Partial Discharge

Impact of Oil Velocity on Partial Discharge Characteristics Induced by Bubbles in Transformer Oil Ju Tang 1, 2 and Yongze Zhang 1 1

State Key Laboratory of Power Transmission Equipment & System Security and New Technology Chongqing University, Chongqing 400044, China

Cheng Pan 2, Ran Zhuo 3, Dibo Wang 3 and Xingxing Li 1 2

3

School of Electrical Engineering, Wuhan University, Wuhan 430072, China Electric Power Research Institute, China Southern Power Grid, Guangzhou 510080, China

ABSTRACT Bubble is a typical insulation defect in transformer oil, which can easily initiate partial discharge (PD) during the operation of transformer. Moreover, it migrates along with the oil flow and may be deformed due to the forced oil circulation and temperature difference within transformer bulk, leading to the complexity of PD characteristics and PD mechanism. However, the understanding of to this phenomenon is not clear. In order to identify it, finite element method was employed to simulate the movement process and deformation of bubbles, as well as electric field distribution under the conditions with different oil flow rates. In addition, numerous experiments were performed, mainly including the change of PD φ-q and φ-n patterns with oil flow rates. By comparing the simulation and experimental results, it is found that PDs are mainly concentrated in the negative half-cycle. Electric force causes bubbles to stretch along the electric field direction. At the stationary condition, bubbles eventually exhibit horizontal stretching and cause maximum electric field distortion. However, bubbles eventually exhibit vertical stretching in flowing oil. Flowing oil can significantly reduce PD intensity. The PD intensity declines rapidly with the increase of the flow velocity, and then increases slowly. Index Terms — Partial discharge, oil flow rates, bubbles, deformation, transformer oil.

1 INTRODUCTION POWER transformers are one of the most important and expensive equipment in power transmission and distribution. Their safe and reliable operation is vital to the safety of the entire power grid. Transformer oil is the main cooling and insulation medium. Therefore, its internal insulation defects may compromise the safe operation of transformers, causing serious accidents. Bubbles are inevitable in transformer oil due to poor vacuum oil-filling, mechanical vibration [1], aging of seal, and other reasons. In addition, when transformer is involved in overheating [2, 3], PD [4-6], and high-energy discharge failure, the insulating medium is decomposed into gas, which can gather into bubbles. The electric field in bubbles is much higher than that in transformer oil because the dielectric constant of transformer oil is more than twice that of the bubbles. However, the electric strength of bubbles is far lower than that of oil. PD easily takes place in bubbles, thereby generating more gas, which leads to further Manuscript received on 1 July 2017, in final form 1 February 2018, accepted 4 February 2018. Corresponding author: C. Pan.

degradation of the insulation performance of transformer oil [7, 8]. Pompili et al [9-13] used ultra-wide and narrowband detection methods to study the PD development characteristics of transformer oil with the needle-to-plane electrode system. He pointed out that streamer inception and PD occurrence are similar physical events. He discovered that streamer begins in the low-density region and eventually becomes an avalanche that causes discharge. He proposed that the first pulse initiates the formation of cavity and the second pulse occurs in bubbles, eventually forming regularly changing pulse bursts. The research of Denat [14] showed that liquid argon and nitrogen under overheating and discharge can produce microscopic bubbles. The bubbles inside the liquid dielectric are prone to PD, which can reduce the insulation properties of liquid further. He suggested that increasing the applied pressure can diminish the impact of bubbles. Yamashita [15] studied the influence of oil surface pressure on the breakdown process and found that the number of discharge branches increases as the pressure decreases, and the phenomenon suggests the existence of the gaseous phase during discharge. Shiota [16]

studied bubble behaviors in oil and found that PD easily occurs in a large bubble, which produces more bubbles. A small bubble is easier to dissolve in oil than a large bubble. This researcher also observed the elongation of bubble in the electric field direction, which is consistent with the results in the literature [17]. Transformer oil is always flowing in actual operation. The above investigations only focused on stationary bubbles and their impact on the dielectric performance of transformer oil. In [5, 18–19], the discharge characteristics of metal particles were studied. PDs caused by metal particles can also produce bubbles according to these studies, but the PD characteristics of bubbles are not involved in these investigations. Therefore, the PD characteristics of bubbles in flowing transformer oil should be studied further. In this study, the deformation of bubbles and the degree of electric field distortion in stationary and flowing oil were simulated through the finite element method, which can help expound the influence mechanism of PD induced by bubbles in transformer oil with different velocities. Furthermore, numerous experiments were performed at different flow velocities. The impact of different velocities on PD was analyzed on the basis of several measured characteristic parameters, including the φ-q scatter diagram (relation between phase φ and apparent charge q), the φ-n spectrum (relation between phase φ and number n), the average apparent charge, the PD number n, and the apparent charge per unit time. The results can contribute to the design of oil duct and the selection of appropriate oil velocity in engineering practice.

transformer oil, which is shown in Figure 3. The circulation system is made up of main oil ducts, plate electrodes, a net sieve, corrugated pipes, an oil pump, an ultrasonic flow meter, and a temperature control system. The main oil ducts are composed of polymethyl methacrylate (PMMA). The plate electrodes consist of two pieces of flat smooth copper electrodes with rounded edges, and its diameter and thickness are 200 and 10 mm, respectively. The distance between the two electrodes can be adjusted from 0 mm to 50 mm. The net sieve, with 1.5-mm hole radius, is used to control the bubble sizes to a certain extent. The corrugated pipe is a thin cylindrical shell with an extension capability, which can compensate for the changes in oil volume as the temperature changes to avoid oil duct deformation or rupture. An oil pump is used to control the oil velocity between 0 and 0.3 m/s. An ultrasonic flowmeter is used to measure oil velocity. The temperature control system maintains a constant oil temperature at 60 °C.

2 EXPERIMENT SETUP AND METHOD

Figure 1. PD experiment setup and measure system of transformer oil.

R1

T2

T1

Cx

Ck

C1

Zm

C2

AC 220V

R

TVS

Coaxial cable

Oscilloscope

1.6 1.4

|Z|/50

2.1 EXPERIMENT SETUP The experiment setup for studying the PD characteristics induced by bubbles in flowing transformer oil is shown in Figure 1. This platform includes voltage regulator T1, isolation transformer T2, 10 kΩ protection resistance R1, coupling capacitor Ck, test sample Cx, voltage-dividing capacitor consisting of C1and C2，and blocking resistance R2. The blocking resistance is used to prevent the voltagedividing capacitor from bypassing the discharge current. The platform allows the voltage range of 0 kV to 50 kV. PD pulses are synchronously measured by detection impedance Zm (RLC type) [20] and 50 Ω non-inductive resistance R. Figure 2 plots the frequency characteristic of R, which shows that it can respond quickly to discharge current and accurately obtain the PD pulse waveform. Transient Voltage Suppressor (TVS) diode, which is in parallel with non-inductive resistance R, can instantaneously conduct to keep the voltage of the resistance below 5.5 V at the sample breakdown. The leakage current of TVS diode is 0.05 μA at working voltage. The parasitic capacitance of the TVS diode is only 19 pF. These excellent characteristics can protect the oscilloscope while reducing the impact on the measurement signal. An oil circulation device was designed to study the impact of oil velocity on the PD characteristics induced by bubbles in

R2

1.2 1.0 0.8 0.1

1

10

GHz

20

Figure 2. Frequency characteristic of 50 Ω non-inductive resistance R. Temperature Electrode sensor

Net Main sieve oil duct

HV

Flowing direction

Air scoop

Bubble

Corrugated pipe

Oil pump

Thermoelectric Heating pipe cooler

Oil drain valve

Figure 3. Circulation system of transformer oil.

Amplitude(V)

2.2 EXPERIMENT METHOD Several preparations were done before the PD experiment. Initially, the oil circulation device was filled with Karamay No. 25 naphthenic mineral oil, which was processed and reached the engineering application requirements. PD test showed no PD at 35 kV applied voltage. Therefore, the following PD experiments were performed below this voltage value. Thus, the experiment results were free of the disturbances from the experiment devices and high voltage wire. Before injecting bubbles, an air bag filled with dry air was connected to the air scoop. Then, the oil drain valve and the air scoop were opened to release 150 mL of transformer oil, and the same volume of dry air was injected into the oil duct at the same time. After the air scoop and the oil drain valve were closed, the oil pump was started and the air was scattered into bubbles. Finally, bubbles were distributed equally through the net sieve in the circulating oil duct, forming stable cycles after several minutes. The test voltage increased at 1 kV/s from 0 kV to PD inception voltage (PDIV) to obtain the appropriate applied voltage. The applied voltage was set as 1.2 times that of the average PDIV, which not only ensures a relatively stable PD intensity but also avoids sample breakdown. The average PDIV of 5 replicate experiments is 21.5 kV. Therefore, all experiments in this study were carried under 26 kV AC. Figure 4 shows a typical negative PD signal in flowing transformer oil induced by bubbles at an oil velocity of 0.18 m/s under 26 kV AC. 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08

3 BUBBLE MOTION ANALYSIS Bubbles in transformer oil move with oil flow, which is a typical gas–liquid two-phase flow. Many forces act on a bubble, including electric force, gravity, buoyancy, oil drag force, Magnus lift force, added mass force, and Basset force. Bubbles are deformed because these forces with different sizes and directions act on different bubble surface positions and bubbles are not rigid bodies. The deformation of bubbles can change the distribution of electric field and further change the electric force applied on bubbles. The electric force can also influence the deformation of bubbles in return. In addition, the bubbles are generally asymmetrical after deformation. These factors contribute to the difficulty in analyzing the deformation of bubbles in flowing transformer oil based on traditional mechanical method. Therefore, the author used the finite element method to simulate the deformation of bubbles and the degree of electric field distortion in transformer oil with different flow velocities. 3.1 BUBBLE MODELLING The geometric model is shown in Figure 5. The oil duct was 30-mm long and 10-mm wide between two parallel electrodes. A bubble with a radius of 1.5 mm was set in the oil duct. The applied voltage between the two parallel electrodes was 20 kV. The dynamic viscosity of transformer oil was 0.0059 Pa·s, the density was 854 kg/m3, and the relative permittivity was equal to 2.2. The relative permittivity of the bubble was set to one.

Inlet 0

1

2 3 4 5 6 Time(μs) (a) RLC detection impedance

7

Amplitude(V)

Transformers oil Electrodes

Outlet

8

Figure 5. 2D simulation model.

0.1 0 -0.1

-0.2 -0.3 -0.40

Bubble

1

2

3

4 5 6 Time(μs) (b) Non-inductive resistance R

7

8

Figure 4. PD pulse waveforms measured by RLC detection impedance and non-inductive resistance R.

As shown in Figure 4, the PD signal detected by RLC had an obvious oscillation, which can help to capture both positive and negative PD pulses. Therefore, the PD signals detected by RLC were set as the trigger signal for oscilloscope. In contrast, the PD signal detected by R was mainly unipolar and it was closer to the real discharge current waveforms. The PD signals from R and AC signals from the capacitor voltage divider were collected synchronously, and they are used to extract apparent charge, PD number and PD phase. The sampling time and rate were 1 ms (1/20 of one power frequency cycle) and 1 GS/s, respectively. Experiments under each condition were performed five times.

The movement of fluid was mainly determined by coupling electric force, oil drag force, and gravity. The transformer oil and the bubble were the continuous phase and dispersed phase, respectively. Assuming the two phases do not mix and they are incompressible newton fluids, the laws of momentum conservation and mass conservation are written as: u   (u )u    [ pI   (u  (u)T )]  t (1) Fst   g  Fe  u  0 (2) where u is the fluid velocity, ρ is the fluid density, p is the pressure, I is the unit matrix, μ is the dynamic viscosity, and g is the gravity vector. Fe is the electric force, and Fst is the surface tension force. The coupling of the flow and electric fields was established by electric force in this model. In the electrostatic field, the electric force Fe in the dispersed phase bubbles can be expressed by Maxwell stress tensor T, as shown in Equation (3)



Fe    T

(3)

and the The Maxwell stress tensor T is written as: 1 (4) T  EDT  ( E  D)I . 2 To study the deformation behavior of bubbles, the twophase change interface should be tracked. In this model, the phase-field method is used. The diffusion interface is defined as a transition region, which can be expressed by the phase field variable  . The phase field variable varies smoothly from −1 (for bubble) to 1 (for transformer oil) [21]. The Cahn−Hilliard Equation can describe the interface dynamics of two-phase flow, which combines the effect of diffusion fluxes and chemical potential. The Cahn−Hilliard Equation is written as:   u     G (5) t where γ is the mobility and G is the chemical potential. G is defined as   ( 2  1)  G     2  (6)  2   where λ is the mixing energy density and ε is the small thickness of the two-phase interface. These two parameters are related to the surface tension coefficient (σ) through the following Equation:

2 2 (7) . 3  The relation between the surface tension of diffusion interface Fst and chemical potential G is shown in Equation (8):

=

Fst  G . (8) The volume fraction of bubbles and transformer oil are denoted as Vfb and Vfo, respectively. The relation between these two parameters and the phase field variable  can be written as:

Vf b  (1   ) / 2

(9) . Vf o  (1   ) / 2 Therefore, the dielectric constant, density, and viscosity in the interface region can be expressed as:    bVf b  oVf o

  bVf b  oVf o .   bVf b  oVf o

(10)

3.2 SIMULATION OF BUBBLES DEFORMATION IN FLOWING TRANSFORMER OIL Based on the above model, bubble deformations and electric filed were simulated with different velocities at 60 °C. The deformation processes of a 1.5-mm radius bubble at 0.00 m/s (stationary), 0.06 m/s, 0.18 m/s, and 0.30 m/s are provided in Figure 6. Pressure can affect the insulation strength of bubble to a certain extent. Bubble pressure is influenced by oil velocities and bubble position. According to the simulation, the bubble pressure increased by 5 Pa only, while the velocity increased from 0.06 m/s to 0.3 m/s at the same position. In addition, the change in pressure was also beyond 100 Pa due

to different bubble positions. These pressure changes can be ignored compared with the atmospheric pressure of 0.1 MPa. Therefore, the following simulation focused on bubbles deformation and its influence on the electric field distribution. The results of 3 replicate simulations were always the same at the same parameter setting. t=0.07

t=0.07

t=0.04

t=0.04

t=0.01

t=0.01

t=0 (a)0m/s

t=0 (b)0.06m/s

t=0.07

t=0.07

t=0.04

t=0.04

t=0.01

t=0.01

t=0

t=0

(c)0.18m/s

(d)0.30m/s

Figure 6. Deformation process of bubble with different flow velocities.

Figure 6a shows the deformation process of a bubble in stationary oil. Initially, the bubble was stretched slightly along the electric field direction (vertical direction). The deformation in the vertical direction gradually weakened during the subsequent rise process. After reaching the upper electrode, the bubble eventually showed a horizontal elongation. The electric force was the main force applied on the bubble at the first elongation in the vertical direction. In the subsequent rise process, the bubble needed to push some transformer oil along the trail. Therefore, a counterforce and the added mass force were applied on the rising bubble. In addition, due to the effect of gravity and buoyancy, pressure differences existed between the upper and lower sides of the bubble. The lower part of the bubble produced a vortex ring and a jet in the middle. These effects surpassed the vertical elongation from electric force, and the bubble showed a horizontal elongation shape. The deformation process in flowing oil showed different characteristics from that in the stationary oil. Figure 6b shows that the bubble deformation at the velocity of 0.06 m/s was similar to that in the stationary state. The bubble was a smooth ellipsoid at t = 0.01 s (transient deformation). It was still stretched horizontally when reached the upper electrode at t =

3.3 INFLUENCE OF BUBBLE DEFORMATION ON ELECTRIC FIELD DISTRIBUTION The different shapes of the bubble can affect the electric field in the bubble. Figure 7 is the electric field distribution that corresponds to the bubble deformation in Figure 6. The electric field strength in a standard spherical bubble with a radius of 1.5 mm is approximately 1.35 times the electric field strength in the transformer oil. The electric field strength in the bubble decreases significantly when the bubble is stretched in the electric field direction. On the contrary, it increases significantly when the bubble is stretched horizontally. Oil velocity can change the electric field in bubbles by changing the bubble shapes. As shown in Figure 7a, the electric field in the bubble was the minimum because of the vertical elongation of the bubble at the initial simulation stage in stationary oil. When the bubble reached the upper electrode, the electric field in the bubble reached the maximum (3.11 MV/m) due to the horizontal elongation. In most cases, bubbles keep mainly in steady shape and move in oil ducts, so it is more meaningful to pay attention to the electric field in steady bubble. Figure 8 reveals the relationship between the maximum electric field in the steady bubble and oil velocity, which clearly shows the impact of oil velocity on electric field. In flowing oil, the electric field strength in the bubble was lower than that in the stationary state. The maximum field strength in the steady bubble was 2.86 MV/m at the flow rate of 0.18 m/s, whereas the maximum field strength in the steady bubble was 2.76 MV/m at the flow rate of 0.30 m/s. The electric field strength decreased with the increase in flow velocity. When the velocity was 0.3 m/s, the electric field in the maximum curve region reached 2.95 MV/m due to the serious transient deformation at t = 0.01 s. The transient distortion of electric field may increase the probability of PD.

3.11E6

t=0.07

2.84E6 t=0.04

t=0.01

2.68E6

2.95E6

2.86E6

2.67E6

t=0.01

t=0

2.76E6 t=0.07

t=0.07

2.85E6 t=0.04

2.76E6

t=0.04

(b)0.06m/s

2.86E6

2.70E6

t=0.07

2.76E6

t=0 2.69E6 (a)0m/s

t=0.01

2.80E6

2.95E6

t=0.04

t=0.01

t=0 2.76E6 (d)0.30m/s

t=0

(c)0.18m/s

Figure 7. Electric field with different flow velocities.

Maximum fo electric field (MV/m)

0.07 s (steady deformation). However, the degree of stretching decreased. With the increase of oil velocity, the transient deformation of bubble was gradually getting irregular. Bubble had a slight distortion (Figure 6c, t = 0.01s) at the velocity of 0.18 m/s, but it had a serious distortion (Figure 6d, t = 0.01 s). With the increase of oil velocity, the steady deformation of bubble was gradually getting regular. The horizontal stretch of bubble gradually weakened, and bubble eventually became a shape of slight vertical elongation, as shown in these frames in Figure 6 at t = 0.07s. This phenomenon was mainly due to the friction loss and local resistance from the rough wall, the variable cross-section, or the oil duct bend. These resistances resulted in a pressure gradient along the flow direction, leading to bubble compression in the flowing direction and a horizontal compression. The higher the oil velocity, the bigger the horizontal pressure gradient was. The bubble exhibited vertical elongation with sufficiently high velocity when the effect of vertical pressure gradient and electric field stretching was greater than the effect of buoyancy and vertical added mass force.

3.1

3.0

2.9

2.8

2.7 0.0

0.1

0.2

0.3

Velocity (m/s)

Figure 8. Maximum of electric field with different velocities.

4 RESULTS AND DISCUSSION 4.1 PD CHARACTERISTIC ANALYSIS BY Φ-Q SCATTER DIAGRAM AND Φ-N SPECTRUM PD signals were collected continuously for 30 minutes to analyze the PD trend induced by bubbles in transformer oil with different oil velocities correctly. MATLAB was used to extract the phase φ, apparent charge q, and number n of the PD pulses and plot the φ-q scatter diagram and the φ-n spectrum, as shown in Figure 9.

60

300

Number

0

-300

20

-600

600

0 100 200 300 0 Phase (degree) (a) 0m/s 5

0 -300 0 600

100 200 300 Phase (degree)

4 Number

0 -300

3 2 1

-600 0 600

0 100 200 300 0 Phase (degree) (c) 0.12m/s 5


4

300

Number

Apparent charge (pC)

2

0 100 200 300 0 Phase (degree) (b) 0.06/s 5

300

0

-300

3 2

1

-600 0


600

0

0

(d) 0.18m/s 5


4

300

Number


3

1

-600

0 -300

3

2

4.2 AVERAGE APPARENT CHARGE The average apparent charge is the average of many apparent discharge quantities of PD, which can reveal PD intensity to some extent. In this study, the apparent charges of PD pulses in 30 minutes were collected at different velocities. The PD in the negative half-cycle is the negative polarity. The following negative apparent charges were replaced by their absolute values for analysis and comparison. 240

1

-600 0


600

0

0

(e) 0.24m/s 5


4

300 Number



4

300 Number


0


40

and are compressed vertically due to buoyancy and forming ellipsoid shapes, which lead to the increase in electric field distortion (shown in Figure 7a). Therefore, the electric field in the bubble is higher in stationary oil at the same applied voltage, and thus PD is more likely to occur at the peak of the power frequency cycle. In addition, bubbles stay between the two electrodes in stationary oil, thus the condition for PD occurrence is easy to satisfy. Several similarities and distinctions exist between the PD phases in stationary and flowing oil. PDs also mainly occur in the negative half-cycle, and the PD numbers of positive and negative polarities differ significantly in flowing oil, which are similar to the PDs in stationary state. Some bubbles, especially some large bubbles, tended to flow in contact with the upper electrode, so PD may be easy to occur due to the charge injection process at the metal-oil interface in negative half-cycle. It may be the reason for prevailing negative PDs. PDs mainly distribute in 70°to 109°and 230°to 306°. The PD phase interval in flowing oil is smaller than that in stationary oil. Moreover, the number of PD decreases significantly. Most of the apparent charges are between 45 pC to 550 pC. PD intensity decreases significantly in comparison with that in the stationary condition because bubbles in flowing oil are stretched vertically due to the horizontal pressure gradient and the significant decrease of the electric field intensity in bubbles.

0 -300

3 2

1

-600 0


0

0

(f) 0.30m/s


Figure 9. PD φ-q scatter diagram and φ-n spectrum.

Figure 9a shows that in the stationary state, the PDs were mainly distributed in 46°to 133°and 227°to 308°. As many as 77 PDs occurred at 270°, which was far more than the number in the other phase. The PD number in positive halfcycle was less than that in the negative half-cycle, and most PDs occurred in the negative half cycle. Most of the apparent charges of PD were between 35 pC to 600 pC. Bubbles gather on the lower surface of the upper electrode

Average apparent charge (pC)


80

600

Negative Positive Positive and negative

220 200 180 160 140 120 100 0.0

0.1

0.2

0.3

Velocity (m/s)

Figure 10. Average discharge quantities with different velocities.

As shown in Figure 10, the average apparent charge of positive PDs was always smaller than the negative PDs whether in stationary or flowing state, which suggests that PD induced by bubbles may tend to occur at a negative voltage. The average apparent charge at the velocity of 0.06 was slightly higher than that in stationary oil. Numerous small-

4.3 PD NUMBER PD number is mainly used to characterize the frequency of PD. In this study, the PD numbers in 30 minutes were counted at different velocities. As illustrated in Figure 11, the PD number in flowing oil decreased greatly compared with that in the stationary state. This phenomenon can be attributed to several factors. In stationary oil, small bubbles may converge and form large bubbles. These large bubbles eventually gather on the lower surface of the upper electrode and are stretched horizontally. The electric field in the horizontal ellipsoid is extremely enhanced according to the simulation in Figure 7a. In addition, Figure 9a shows that many low-amplitude PDs exist. Therefore, the condition for the occurrence of PD induced by bubbles in stationary oil is easy to satisfy, and thus leads to the significant number of PDs in stationary oil. 1000 Negative Positive Positive and negative

500 300

PD number

100 50 30 10 5 3

sufficiently high. Therefore, the electric field in the distortional region with max curvature increases to some extent. For example, the electric filed strength in Figure 7a at t = 0.01 s was larger than at the initial state. The first effect played a dominant role when the oil velocity was low, therefore the PD number declined sharply. However, the second effect gradually played a dominant role when the oil velocity was sufficiently high, therefore some bubbles exhibited irregular distortion, and then the PD number increased slightly. 4.4 APPARENT CHARGE PER UNIT TIME Although the two parameters, average apparent charge and PD number, can indicate PD intensity to some extent, the average apparent charge is not involved in the time parameter and the PD number does not cover the discharge amplitude. To reveal PD intensity clearly, the average sum of discharge quantities in one minute was calculated based on the apparent charge and PD number in 30 minutes. Figure 12 plots the curve of the apparent charge per unit time with different velocities. The apparent charge per unit time of the all PDs was up to 5210.17 pC in the stationary state, and only 84.53 pC at the velocity of 0.18 m/s. The latter is only 1.6% of the former. The apparent charge per unit time in flowing oil declines significantly compared with that in the stationary state. This is mainly due to the coincident decrease of PD number and apparent charge in the flowing state. The possible reasons are clarified in sections 3.4 and 3.5. 10000 Apparent charge per unit time (pC)

amplitude PDs in stationary state were responsible for its smaller apparent charge, as shown in Figure 9a. However, the number of PDs with small amplitude reduced sharply in the flowing state, increasing the average value in turn at the velocity of 0.06 m/s. As the velocity increased further, the average apparent charge of positive polarity, negative polarity and all of PDs showed a clear downtrend. The decrease mainly resulted from the increase in the horizontal pressure gradient with the increase in velocity. The increase in horizontal pressure gradient stretched the bubbles vertically, and hence decreased the maximum of electric field in bubbles (shown in Figure 8). Therefore, the average apparent charge decreases with the further increase in oil velocity.

5000 3000

Negative Positive Positive and negative

1000 500 300 100 50 30 10 0.0

0.1

0.2

0.3

Velocity (m/s)

1 0.0

0.1

0.2

0.3

Figure 12. Unit time discharge quantities with different velocities.

Velocity (m/s)

Figure 11. PD number with different velocities.

In flowing oil, the PD number shows a tendency to decline first and then rise. When the velocity was between 0.06 m/s and 0.18 m/s, the PD number declined rapidly with the increase in velocity. When the velocity was more than 0.18 m/s, the PD number increased slightly. The phenomenon is mainly caused by the two effects of oil velocity. The first effect is the vertical elongation of steady bubbles. The higher the flow velocity is, the larger the vertical elongation and the smaller the electric field in the bubble are. The second effect is the transient irregular distortion of bubble. Some bubbles exhibit transient irregular distortion when the velocity is

5 CONCLUSIONS (1) In stationary state, bubbles eventually show a steady horizontal elongation shape, and the degree of electric field distortion is the maximum. In flowing state, bubbles eventually show a steady vertical elongation shape, and the degree of electric field distortion is small. With the increase in oil velocity, the vertical elongation of bubbles enhances, and the degree of electric field distortion reduces gradually. The transient deformation of bubbles significantly enhances at high velocity, and the degree of electric field distortion increases. (2) The intensity of PDs induced by bubbles in flowing

transformer oil significantly decreases, compared with the stationary state. Moreover, the PD intensity decreases rapidly and then increases slowly with the increase in velocity. Therefore, an appropriate oil velocity exists. PD intensity will be very low at the appropriate middle velocity. (3) The PDs induced by bubbles in transformer oil are mainly concentrated on the negative half-cycle. Apparent charge of negative PDs are generally more than that of positive PDs in both flowing oil and stationary oil. This study referred to two horizontal electrodes, whereas it might not be the most common situation. Oil moving in the vertical direction is the most common. The flowing direction may affect the final deformation of bubbles and further affect PD characteristics. Therefore, further study on vertical electrodes is really necessary.

ACKNOWLEDGMENT The authors acknowledge the National Natural Science Foundation of China (51377181).

REFERENCES [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

S. M. Korobeynikov, A. L. Bychkov, A. Yu. Ryzhkina, M. V. Sviridenko, L. A. Darian and A. V. Melekhov, “Microbubbling in transformer oil due to vibration,” IEEE Trans. Dielectr. Electr. Insul., vol. 20, pp. 675–677, 2013. T. V. Oommen and S. R. Lindgren, “Bubble evolution from transformer overload,” in Proceedings of the Transmission and Distribution Conference and Exposition, 2001, pp. 137–142. F. W. Heinrichs, “Bubble formation in power transformer windings at overload temperatures,” IEEE Trans. Power Appar. Syst., vol. 98, pp. 1576-1582, 1979. D. J. Swaffield, P. L. Lewin, G. Chen and S. G. Swingler, “Partial discharge characterization of streamers in liquid nitrogen under applied AC voltages,” IEEE Trans. Dielectr. Electr. Insul., vol. 15, pp. 635–646, 2008. X. Li, J. Tang, S. Ma and Q. Yao, “The impact of temperature on the partial discharge characteristics of moving charged metal particles in transformer oil,” in Proceedings of the IEEE International Conference on High Voltage Engineering and Applications, 2016, pp. 1–4. M. Tsuchie, M. Kozako, M. Hikita and E. Sasaki, “Modeling of early stage partial discharge and overheating degradation of paper-oil insulation,” IEEE Trans. Dielectr. Electr. Insul., vol. 21, pp. 1342–1349, 2014. T. Ishikura, H. Muto, K. Wada and N. Hosokawa, “Distinction of the partial discharge source in oil by two frequency correlation method,” in Int’l Conf. on Condition Monitoring and Diagnosis, pp. 42–45, 2008. H. Janani, N. D. Jacob and B. Kordi, “Automated recognition of partial discharge in oil-immersed insulation,” in Proceedings of the IEEE Electrical Insulation Conference, 2015, pp. 467–470. M. Pompili, C. Mazzetti and R. Bartnikas, “PD pulse burst characteristics of transformer oils,” IEEE Trans. Power Del., vol. 21, pp. 689–698, 2006. M. Pompili, C. Mazzetti and R. Bartnikas, “Phase relationship of PD pulses in dielectric liquids under ac conditions,” IEEE Trans. Dielectr. Electr. Insul., vol. 7, pp. 113–117, 2000. M. Pompili, C. Mazzetti and R. Bartnikas, “Simultaneous ultrawide and narrowband detection of PD pulses in dielectric liquids,” IEEE Trans. Dielectr. Electr. Insul., vol. 5, pp. 402–407, 1998. M. Pompili, C. Mazzetti and E. O. Forster, “Partial discharge distributions in liquid dielectrics,” IEEE Trans. Electr. Insul., vol. 27, pp. 99–105, 1992. C. Mazzetti, M. Pompili, H. Yamashita and E. O. Forster, “A comparison of streamer and partial discharge inception voltages in liquid dielectrics,” in Proceedings of the IEEE 6th International Conference on Dielectric Materials, Measurements and Applications, 1992, pp. 93–95.

[14] A. Denat, F. Jomni, F. Aitken and N. Bonifaci, “Thermally and electrically induced bubbles in liquid argon and nitrogen,” IEEE Trans. Dielectr. Electr. Insul., vol. 9, pp. 17–22, 2002. [15] H. Yamashita and H. Amano, “Prebreakdown density change, current and light emission in transformer oil under non-uniform field,” J. Electrost., vol. 12, pp. 253–263, 1982. [16] H. Shiota, H. Muto, H. Fujii and N. Hosokawa, “Diagnosis for oilimmersed insulation using partial discharge due to bubbles in oil,” in Proceedings of the 7th International Conference on Properties and Applications of Dielectric Materials, 2003, pp. 1120–1123. [17] W. Dong, R. Li, H. Yu and Y. Yan, “An investigation of behaviours of a single bubble in a uniform electric field,” Exp. Therm. Fluid Sci., vol. 30, pp. 579-586, 2006. [18] J. Tang, S. Ma, X. Zhang, M. Zhang, Z. Liu, X. Li and F. Zeng, “Investigation of partial discharge between moving charged metal particles and electrodes in insulating oil under flow state and AC condition,” IEEE Trans. Dielectr. Electr. Insul., vol. 23, pp. 1099–1105, 2016. [19] J. Tang, S. Ma, X. L, Y. Zhang, C. Pan and J. Su, “Impact of velocity on partial discharge characteristics of moving metal particles in transformer oil using UHF technique,” IEEE Trans. Dielectr. Electr. Insul., vol. 23, pp. 2207–2212, 2016. [20] R. Bartnikas, “A commentary on partial discharge measurement and detection,” IEEE Trans. Electr. Insul., vol. 22, pp. 629–653, 1987. [21] P. Yue, J. Feng, C. Liu and J. Shen, “A diffuse-interface method for simulating two-phase flows of complex fluids,” J. Fluid Mech., vol. 515, pp. 293–317, 2004. Ju Tang was born in Pengxi, Sichuan Province, China in 1960. He received his bachelor’s degree at Xi’an Jiaotong University and master’s and doctorate degrees at Chongqing University. Dr. Tang is a professor at the State Key Laboratory of Power Transmission Equipment and System Security, Chongqing University and a chief scientist of the National Basic Research Program of China (973 Program) (2009CB724500). He is currently involved in high-voltage electric equipment insulation online monitoring and fault diagnosis. He is the corresponding author in this paper. Yongze Zhang was born in Taian, Shandong Province, China, in 1988. He received his bachelor’s degree from Southwest Jiaotong University. He is currently pursuing his Ph.D. at the State Key Laboratory of Power Transmission Equipment and System Security, Chongqing University, China. He is involved in high-voltage electric equipment insulation online monitoring and fault diagnosis.

Cheng Pan was born in China in 1986. He received the B.S. and Ph.D. degrees in electrical engineering from Xi'an Jiaotong University, China in 2008 and 2014. Now he is a post-doctor of Wuhan University, China. His research involves partial discharge mechanism and surface charge measurement.

Ran Zhuo was born in Guiyang, Guizhou Province, China, in 1986. He received the Bachelor, Master and Ph.D. degrees in electrical engineering from Chongqing University, Chongqing, China. And now he is an engineer in Electric Power Research Institute of China Southern Power Grid.

Dibo Wang was born in Liuzhou, Guangxi Province, China, in 1988. He received the Bachelor, Master and Ph.D. degrees in electrical engineering from Chongqing University, Chongqing, China. And now he is an engineer in Electric Power Research Institute of China Southern Power Grid.

Xingxing Li was born in Chongqing Municipality, China in 1992. He received his bachelor’s degree at Chongqing University. He is currently pursuing his master's degree at the State Key Laboratory of Power Transmission Equipment and System Security, Chongqing University, China. She is involved in high-voltage electric equipment insulation online monitoring and fault diagnosis