Effect of Magnetic Field on Partial Discharge Initiated by Metallic Particle in Thermally Aged Natural Esters Under AC and Harmonic Voltages

This paper reports the experimental and theoretical investigations of particle levitation voltage on thermally aged ester fluid, under AC and harmonic AC voltages, in the presence of both electric and magnetic fields (130 mT and 160 mT). The results indicate a higher sensitivity to identify partial discharge (PD) initiated due to particle movement in aged ester fluids with an ultra-high frequency (UHF) sensor than the fluorescent fiber technique. The cause for the reduction in sensitivity of PD detection due to the fluorescent fiber technique with thermally aged fluid is analyzed using steady-state fluorescent measurement. The reduction in the levitation voltage noticed under high-frequency AC voltages is much more severe than its impact under the fundamental frequency of AC supply voltage. In addition, the presence of a magnetic field reduces the magnitude of levitation voltage substantially. The UHF signals generated due to particle movement-initiated discharges with aged ester fluids indicate a shift in its dominant frequency of 0.9 GHz under the absence of a magnetic field to around 0.6 GHz with the effect of a magnetic field.

in higher cost-benefit considering its overall insulation 27 The associate editor coordinating the review of this manuscript and approving it for publication was Arpan Kumar Pradhan .
lifetime [3]. Besides, they are biodegradable in nature, mak-28 ing them a desirable fluid for transformer operation. Continu-29 ous research is underway to better understand the behavior of 30 ester fluid insulation. The major focus is on their performance 31 under a variety of electrical and thermal loads, as well as 32 chemical processes such as oxidation and hydrolysis that 33 [18]. Considering the above aspects 101 of particle intrusions into the dielectric fluid and the exis-102 tence of magnetic field inside the transformers, the present 103 work examines the partial discharges with copper particle 104 immersed in natural ester fluid with a steady magnetic field 105 (130 mT, 160 mT). A sphere-plane electrode configuration is 106 adopted to allow the alternating quasi-electric field to interact 107 with the magnetic field orthogonally. Vegetable oils contain 108 large concentrations of vitamin E and luminous fatty acids. 109 Thus, the liquid properties are attributable to the change in 110 chemical composition [19]. Therefore, spectroscopic inves-111 tigations of ester fluids are interesting, thereby selecting a 112 suitable fluorescent fiber for identifying incipient discharges. 113 Considering the above review of the literature on PD activ-114 ity, the following experimental investigation was performed 115 in the present work. (i) Steady-state fluorescence analysis on 116 thermally aged ester fluid, (ii) Particle movement-initiated 117 PD activity on aged ester fluid adopting both fluorescent 118 fiber and UHF sensor with and without a magnetic field, (iii) 119 Characteristics of UHF signals involved during PD activity. 120

121
A. THERMAL AGEING

122
The base fluid used in the current work was natural ester 123 (MIDEL 1215) produced from soybean seeds. The thermal 124 aging was conducted with the copper sheet wrapped around 125 the pressboard substrate (1.5 mm thickness) and then sub-126 merged in a natural ester fluid inside a beaker. The weights 127 of the pressboard, copper sheet, and oil (1:1:10) were kept 128 identical to those in real-time power transformers. Based 129 on the author's previous experience [20], thermal aging was 130 carried out at 160 • C as per IEC 62332-2 standard in a 131 temperature-controlled convection oven in an open beaker 132 with internal air circulation. Both the oil samples and the 133 pressboard were taken out of the reaction mixture at different 134 times maintaining their weight ratios and then inspected for 135 further investigation. The aged samples were sampled at 136 different aging duration including 24, 48, 72, 96, 250 and 137 500 hours in order to have a controlled aging history.  Fig. 1 shows the schematic test setup used for detecting 140 the particle levitation voltage. The different voltage profiles 141 were generated with the use of function generator (Tektronix 142 3051C, 5 GSa/s, 50MHz). The generated signal is then fed 143 to the high voltage Trek amplifier (model 20/20C) with an 144 amplification factor of 4 kV providing a maximum output 145 range limited ±20KV. The spherical electrode (1.5 cm in 146 diameter) as a high voltage and a concave dish type electrode 147 [14] was used as ground to maintain the particle accurately at 148 the gap spacing between the electrodes. The metallic copper 149 particle with a diameter of 0.5 mm is used in the present work. 150 The bottom electrode was made slightly concave in shape, 151 to keep the particle in place.

152
The sphere-plane electrode arrangement is utilized to 153 mimic the partial discharge activity in natural ester fluid. 154 VOLUME 10, 2022 To compare the particle movement initiated partial discharge 155 of natural ester fluid with that of mineral oil from the ear-156 lier work by the authors [21], [22] without the presence of 157 magnetic field, the distance between the high voltage and  region and thus employing UHF sensors for its detection [24]. detected [25]. The fluorescent fiber sensor approach for PD 193 activity during particle movement in ester fluid was studied 194 using a white fluorescent fiber (1 m long and 1 mm diameter) 195 with a silicon photomultiplier module. One end of the fiber 196 is looped circling outside the test cell, orienting towards the 197 gap spacing between the sphere-plane electrode, while the 198 other end of the fiber is fed into the silicon photomultiplier 199 module (biased at 28 V). A sleeve was used to encapsulate 200 the fiber length away from the region of test cell equipment 201 to reduce the external light from interfering with the actual 202 signals. Along with fluorescent fiber, concurrent measure-203 ments are taken using a UHF sensor (3 GHz bandwidth) 204 that is positioned at a distance of 20 cm from the discharge 205 region for better accuracy [26]. Signals from both the sensor 206 (fluorescent fiber and UHF) were recorded using a digital 207 storage oscilloscope (bandwidth of 3.5 GHz and sampling 208 rate of 40 GSa/s) and 50 input impedance.   [19]. The 228 unaged ester fluid has its excitation and emission wavelength 229 at 500 nm and 250 nm respectively. It is observed that 230 increasing the aging time causes a shift in both excitation and 231 emission wavelengths. The rate of change in its wavelength 232 for the lower aging duration was minimal up to 72 hours and 233 a significant shift was seen at higher aging durations of more 234 than 96 hours. The increase in the percentage of excitation 235 and emission wavelengths were respectively around 40% 236 and 168% for 500 hours of aging duration. This prominent 237 Stokes shift observed for the aged ester fluid demonstrates 238 that understanding the characteristic changes in the insulating 239 fluid's EEM spectra may be a useful tool for assessing the 240 fluid's quality in a real-life power transformer. Along with 241 the stroke shift, the decrease in intensity of fluorophore indi-242 cated in the color scale as the fluid gets thermally degraded 243 gives a visual indication of degradation in insulant properties. 244 The absorption and emission peak of the fluorescent fiber 245 utilized for detection of PD activity due to particle levitation 246 movement in ester fluid must be limited to certain spectral 247  [28]. The fluorescence behavior of ester fluids differs 278 from the mineral oil in terms of its excitation and emission 279 wavelengths, which are to be tested with different fluorescent 280 fibers. The type of fluorescent fiber best suited for identifying 281 PD with a higher accuracy during particle movement can be 282 selected from an understanding of the fluid's EEM spectra. 283 Also, steady-state fluorescence spectroscopy can also be used 284 to monitor the insulation status of liquid used inside the power 285 transformer.

287
The current work uses both UHF and fluorescent fiber to 288 study the particle levitation voltage. The voltage at which 289 the first discharge pulse is noticed in the oscilloscope from 290 the UHF sensor/fluorescence fiber is considered as particle 291 levitation voltage. The experimental results of this particle 292 movement are based on an average of 10 readings with a 293 standard deviation of less than 3%.  time varying, the particles will levitate and drops back to 302 the ground electrode or due to the force, the particle hovers 303 over the ground electrode. This process initiates incipient 304 discharges [21]. During this process, the first signal captured 305 using the UHF and fluorescence fiber is recorded as the 306 Particle Levitation Voltage (PLV). Assuming the sphere plane 307 configuration to exhibit a uniform electric field for the smaller 308 gap spacing (5 mm), the charge obtained by the particle 309 during its contact with the electrode is given by [32], shifting of particles to low field zone [22]. thermal aging along with an increased drag force exhibited 336 by the particle is liable for the reduction in its discharges. The 337 levitation voltage measured using fluorescence fiber (Fig. 3b) 338 is higher than the UHF sensor indicating its lower sensitivity 339 towards the detection of partial discharges caused by particle 340 movement in the liquid.

341
In addition, the effect of the magnetic field (130 mT, 342 160 mT) showed a slight reduction in the levitation voltage 343 compared to its influence without the magnetic field (0 mT) 344 and the rate of change in the levitation voltage inferred 345 through the UHF technique under the magnetic field followed 346 a similar trend as that noticed without considering the mag-347 netic field. Although fluorescent fiber-based detection shows 348 a better resistance towards electromagnetic interference [27], 349 its accuracy towards detecting discharges upon the applied 350 magnetic field was very much lesser and inconsistent com-351 pared to the UHF technique. The reduction in the intensity of 352 fluorophore as observed from the steady-state fluorescence 353 of aged ester fluids (Fig. 2) could be responsible for its less 354 sensitivity towards the discharges and thus not suitable for its 355 detection towards the PD activity due to particle movement 356 in thermally aged ester fluids.

2) IMPACT OF HIGH-FREQUENCY AC VOLTAGES
358 Fig. 4 shows the effect of high-frequency AC voltages (3f, 359 5f, 7f) on the particle levitation voltage of thermally aged 360 ester fluids. It is observed that particle starts levitating at 361 lower voltages under high-frequency AC voltages compared 362 to its effect under fundamental frequencies.

363
The amount of charge that the particle has acquired when 364 it is sitting on the ground electrode does not vary with applied 365 frequency. But once the particle is lifted from the ground, 366 it must traverse a specific distance in order to initiate a charge 367 transport process which is a function of the applied voltage 368 frequency.

369
Based on the observations, it is noticed that trajectories 370 of the particle with different frequencies did not have much 371 change significantly, indicating only a minimal variation in 372 its particle levitation voltage. Also, with the impact of a 373 higher magnetic field (130 mT and 160 mT), the particle 374 starts levitating with a smaller gyro radius causing a strong 375 interaction with the magnetic field lines. This phenomenon 376 due to the electromagnetic field causes a partial bridging of 377 the particle between the electrodes. This, results in discharges 378 at a lower voltage (Fig. 4c) than its impact compared without 379 a magnetic field. representing 4% and 40% THD levels.

392
In the absence of a magnetic field (Fig. 5a)

410
The signal pattern obtained from the UHF sensor during 411 partial discharge activity owing to particle movement and its 412 Fast Fourier Transform (FFT) is shown in Fig. 6. The peak-to-413 peak magnitude of UHF signals as noticed in Fig. 6a is higher 414 with the addition of a magnetic field and thus responsible 415 for the reduction in its levitation voltage compared to its 416 influence without a magnetic field. The frequency content 417 of the UHF signal spans from 0.3 GHz to 1 GHz, with 418 0.9 GHz as its dominating frequency without the existence 419 of a magnetic field (0 mT). While considering the impact of 420 discharges measured under the AC voltage profile, the UHF 421 signal obtained with the impact of magnetic field (130 mT, 422 160 mT) has shown a change in its dominant frequency to 423 around 0.5 to 0.6 GHz (Fig. 6b) specifically in comparison to 424 its influence without the presence of the magnetic field, where 425 the frequency is observed in the range of 0.9 GHz to 1 GHz. 426 This shift observed may be attributed to the electromag-427 netic interferences caused by particle movement and the drift 428 produced due to − → E × − → B . Also, there was not much varia-429 tion in the frequency of the UHF signal for the PD activity 430 measured under high-frequency AC voltages and harmonic 431 frequencies with different THDs, whereas the addition of 432 a magnetic field caused a frequency shift similar to that 433 observed under AC voltages. Similar results were observed 434 with higher magnetic field in mineral oil which is due to 435 the spiral gyration (gyrofrequency) experienced by charge 436 carriers [33]. The injected current pulse responsible for the 437 generation of electromagnetic radiation during PD is being 438 distorted by the magnetic field, resulting in a shift in the 439 frequency spectrum of UHF signals. Thus, it has been con-440 firmed that the addition of a magnetic field during the particle 441 movement-initiated discharges in ester fluids can mislead 442 the insulation engineers while monitoring the incipient dis-443 charges occurring inside the power transformers. This motion of the particle leads to a pattern of flow [32] 465 and the force transmitted from the surface charges of the 466 moving particle causes drag in the fluid. The collection of 467 the charged particle behaves in a collective way because of 468 these forces. As explained (1), the particle experiences an 469 electrical force that is proportional to the square of the particle 470 radius, permittivity, and electric field applied. Along with this 471 force, the particle also experiences a viscous force that is 472 given by [32] 473 F d = 6πrηvC (2) 474 where η is the dynamic viscosity (mPa.s), r is the radius of 475 the particle used for study, v is the velocity of the particle 476 and C is the correction factor that attributes to non-linearity 477 in the drag caused by the levitation of particles. The electric 478 force initiates the particle movement thereby absorbing the 479 charges from the electrodes initiating the discharges but upon 480 the addition of a magnetic field, an orthogonal force gets 481 overlapped with the applied electric field. Hence, along with 482 these forces, the charged particles behave in a collective way 483 under the influence of a magnetic field getting affected by 484 Lorentz Force − → F L as, where, q is the particle's electric charge, − → v is the velocity 487 and − → B is the magnetic flux density. Since the direction of 488 force expressed in (3) is the cross product of velocity and 489 magnetic field, Lorentz force always acts perpendicular to 490 the direction of motion which causes particles to move in a 491 circular motion with the covering of orbital motion particles 492 known as gyro-radius. Since this force acts perpendicular 493 to the trajectory of the particle, it experiences a curved path 494 with a gyro radius (R g ) until it forms a complete circle. When 495 the intensity of the magnetic field is higher, it alters the 496 trajectories of the particle which respond differently to the 497 forces that are parallel and perpendicular to the direction of 498 − → B . This is known as magnetized plasma in which plasma 499 tends to have anisotropic properties that react differently to 500 these forces. Therefore, the total force that acts on the particle 501 due to both electric and magnetic fields is given by, When the trajectory of the particle is such that it travels in a 504 direction other than − → B , it moves in a direction perpendicular 505 to it, causing the particle to gyrate. In the presence of a 506 uniform magnetic field, particle covers gyration radius [34], 507 given by where, α is the velocity and is the gyro frequency, the 510 frequency of moving charged particles which is directly pro-511 portional to − → B . Thus, the reduction in the magnitude of the 512 particle levitation voltage as observed from the experimental 513 results can be related to the gyration of the particle as indi-514 cated in (5) which is dependent on the level of a magnetic 515 field. In the range of magnetic field applied to understand 516 the Partial discharge formed due to particle movement, its 517 impact for variation in circular motion of the particle is not 518 observed. In the range of magnetic field applied to understand 519 the Partial discharge formed due to particle movement, its 520 impact for variation in circular motion of the particle is not 521 observed. Admittedly, further work is essential to confirm 522 for any variation in particle movement causing PD variation, 523 under time varying high electric and magnetic fields.

554
• Further, the theoretical studies on the impact of the mag-555 netic field on particle movement-initiated discharges 556 were discussed indicating that the gyration of the particle 557 tends to reduce the particle levitation voltage.