1068

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006

Acousto-optical PD Detection for Transformers

Xiaodong Wang, Baoqing Li, Harry T. Roman, Onofrio L. Russo, Ken Chin, and Kenneth R. Farmer
AbstractPartial discharge (PD) is one of the factors that could lead to failure of power transformers, leading to power outage and expensive repairs. The acoustic wave induced by PD can be measured and used for monitoring, diagnosing, and locating potential failures in the transformers. Fiber optic sensors have been shown to be attractive devices for PD detection because of a number of inherent advantages including small size, high sensitivity, electrical nonconductivity, and immunity to electromagnetic interference. A ber optic sensor based on a Fabry-Perot interferometry is constructed by a simple micromachining process compatible with microelectromechanical system technology. The sensors are used in a transformer to measure PD acoustic waves. The experimental results show the sensor not only has an inherent high signal to noise capability, but is able to accurately localize the PD sources inside the transformer. Index TermsAcoustic sensor, partial discharge, transformer.

basic chemical species. This has been exploited to produce a and thereby to detect PD sources. gas analyzer to detect Chemical testing has some limitations that prevent it from being the only method used for PD detection. Chemical testing does not provide any information about the position of the PD or the extent of the insulation damage. This limits the usefulness of chemical detection and rules it out as a singular solution to PD detection and positioning. B. Electrical Detection Electrical detection focuses on capturing the electrical pulse created by the current streamer in the void [10]. Electrical detection includes two methods: Pulse Current Method and Ultra High Frequency Method (UHF). Pulse Current Method: This method gets the apparent charge by detecting the PD current in detecting impedance, earth line and coil [11][13]. As the only PD detection method of having international standard, it is studied earliest and applied broadest. The test results can be referenced in the insulation on-line monitoring of transformers. This method is easy for quantitative measurement and it has high sensitivity. But there are three main disadvantages [14], [15]. First, because of its high sensitivity in measurement, its false alarm is also high. Secondly, this way puts focus on the ground connection of transformers which will produce electric current if the transformers have partial discharge. The maintenance of equipment can not be done because it is hard to determine the PD location in high-tension coils. Thirdly, this way is not suitable for long-term monitoring of transformers. It costs too much to connect measurement equipment to ground lines. Ultra High Frequency Method (UHF): With recently development of PD detection technology, UHF detection which is based on the detection of electrical resonance at ultrahigh frequencies up to 1.5 GHz caused by PD excitation can be applied to realize not only the PD phenomena but also the location of a PD source [16][19]. UHF detection has its limitations. The primary limitation is its susceptibility to noise. The transformer environment contains high levels of electrical noise, both narrowband and broadband. In some cases, it is extremely difcult to distinguish between noise and a PD because of the short PD pulse width. This problem leads to false detection in online electrical PD systems. The transformer can be taken ofine and connected to an external power source for testing in order to eliminate some of the noise, but taking the transformer ofine can cost hundreds of thousands of dollars a day in lost revenue for the power company [20]. Another problem with electrical detection is that the received pulse characteristics are highly dependent on the geometry of the transformer. Different components within the transformer can distort the pulse shape needed to characterize the

I. INTRODUCTION ARTIAL discharge (PD) is a short abrupt ow or release of current caused by buildup of the electric eld intensity in a limited region. PD in the transformer is a sign of dielectric defects as well as a cause for further degradation of its insulation system and may therefore lead to a failure of the apparatus. PD measurements have been used for many years during acceptance tests of new high-voltage apparatuses for uncovering possible construction problems, such as bad design, defects in insulation materials used, or poor workmanship. They have also been used for detecting the deteriorated parts of the insulation in an apparatus under normal operation [1][5]. II. OVERVIEW OF PD DETECTION Occurrence of PDs in electrical insulation is always associated with emission of several signals: electrical and acoustic pulses and chemical reactions. These signals are a means for detection of PD activities in high-voltage apparatuses. Over the past 40 years, several methods have been developed to detect PDs within transformers. These can be grouped into three categories, based on the PD manifestation that they measure: chemical, electrical, and acoustic detection. Each is briey explained in the following. A. Chemical Detection One of the consequences of PDs is chemical change of material [6][9]. For example, a PD activity decomposes to
Manuscript received April 25, 2005; revised September 11, 2005. This work was supported by the Public Service Electric and Gas Company. Paper no. TPWRD-00244-2005. X. Wang, B. Li, O. L. Russo, K. Chin, and K. R. Farmer are with the Department of Applied Physics, New Jersey Institute of Technology, Newark, NJ 07102 USA (e-mail: xw3@njit.edu). H. T. Roman is with PSE&G Company, Newark, NJ 07102 USA (e-mail: harry.roman@pseg.com). Digital Object Identier 10.1109/TPWRD.2005.861242

0885-8977$20.00 2006 IEEE

WANG et al.: ACOUSTOOPTICAL PD DETECTION FOR TRANSFORMERS

1069

type of PD fault and can again result in erroneous detection. Finally radio frequency interference in the UHF bands may limit utility. C. Acoustic Detection Acoustic detection of PDs is based on the detection of the mechanical waves propagated from the discharge site to the surrounding medium. This signal is created because when the current streamer is formed within the void, the material around the hot streamer is vaporized. This vaporization causes an explosion of mechanical energy, which then propagates through the transformer tank in the form of a pressure eld [21][27]. Acoustic detection has been widely used in diagnostics of transformers. The use of piezoelectric transducers attached to the transformer tank wall has been the most favored approach [28] in transformers. The primary advantage of using acoustic detection over chemical and electrical methods is that position information is readily available from acoustic systems using sensors at multiple locations. This position information can help to identify the type of PD as well as the location and severity of an insulation fault. Acoustic detection has another advantage that observations can be made in the presence of large electromagnetic disturbances. D. Acoustic Detection Sensors PD occurring under oil produces a pressure wave that is transmitted throughout the transformer via the oil medium. One can achieve acoustic PD detection by mounting piezoelectric acoustic sensors externally on the transformer wall. An externally mounted piezoelectric acoustic sensor offers the advantages of easy installation and replacement. However, a piezoelectric sensor may suffer from degeneration of the signal-to-noise ratio caused by environmental noises such as EMI. Another disadvantage associated with an externally mounted piezoelectric sensor is that the multiple paths of the acoustic wave transmission make locating the exact positions of PD difcult. It is thus desirable to have sensors that can function reliably inside a transformer to detect clean PD-induced acoustic signals. Fiber optical acoustic sensors have been shown useful in partial-discharge detection [29], [30]. Early ber optic sensors for acoustic signal detection were based mostly on ber optic intrinsic interferometers such as all-ber Michelson interferometers and Mach Zehnder interferometers. However, they suffer from the fringe fading problems that result from random polarization rotation. They are also unstable because of drift in the source wavelength and temperature-induced path-length changes. More recently, ber optic FabryPerot interferometric sensors have been under development for acoustic-signal detection [31][34]. FabryPerot interferometric sensors are fabricated with a small sensing element known as a FabryPerot cavity formed by two parallel reecting surfaces. Compared to Michelson and MachZehnder ber sensors, FabryPerot interferometric sensors are compact in size and therefore achieve virtually single-point measurement. More importantly, those random polarization rotation and phase changes that are environmentally induced in the ber connecting the optical

Fig. 1. Packaged sensor.

Fig. 2.

Close-up of the spark tank, spark gap and assembly.

source, the sensor head, and the detectors are common mode and therefore do not affect the signal phases. But the measurement sensitivities of these sensors are not high enough for PD detection. III. THE OPTICAL SENSORS IN PD DETECTION This section involves the testing of ber optic acoustic sensors that can detect the occurrence of PDs in power transformers. When a PD occurs inside the transformer, the oil that acts as insulation becomes degraded, and the result can be unexpected catastrophic failure of the power transformer. The optical sensor mounted inside transformers can help warn of the occurrence of PDs, thus helping to prevent transformer failures by measuring PD acoustic signals. An ultra sensitive optical sensor was developed for PD detection inside transformers. Detailed descriptions of sensor design, fabrication and characterization are shown in [35]. A stainless steel housing for the sensor that has been designed and fabricated has the sensor placed at one end as shown in Fig. 1. A 7-in-long tube that can also is attached to a drain plug by a compression tting. The end of the housing opposite the sensor contains a stainless steel tee tting. One tting of the tee is for the optical ber and the other is to control the sensor backside pressure and environment. A. Partial Discharge Generation Fig. 2 shows a close-up of the spark tank, spark gap and assembly. The spark tank consists of a plastic tub lled with transformer oil, two micromanipulators for placing and holding opposing electrodes in close proximity, and two hypodermic needles to serve as sharp electrodes. The needle electrodes have a diameter of 0.1 mm. The gap length between the needles is 0.5

1070

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006

Fig. 3. Upper trace: the acoustic signature of a spark with the sensor roughly 7 mm away from the spark source (the time scale of 20 s and the output voltage scale of 10 v).

Fig. 4. Experimental setup for PD detection.

mm. An air compressor is used to create a source of bubble in the vicinity of the spark gap. B. Sensitivity Experiment A piezoelectric acoustic sensor (Harisonic G-0504 type) was used for PD detection in the transformer oil (shown in Fig. 2) and transfers it to digital capture by the oscilloscope for real time display and storage. The piezoelectric sensor has an acoustic frequency response up to 5 MHz. A typical oscilloscope image is presented in Fig. 3 showing the acoustic signature of a spark (upper trace). In the oscilloscope images, the upper trace is the sensor output curve and the lower trace is measured within the spark generator circuit. Thus the upper trace is the only data of interest. An experimental system is designed to evaluate the optical sensor performance in PD detection. The standard acoustic methods used to detect PD signal employ piezoelectric sensors externally. As shown in Fig. 4, the ber optical sensor is installed in transformer oil at one end of the tank, facing the piezoelectric sensor at the other end of the tank. The PD generation system is inside the tank. This arrangement is chosen because it represents many practical PD test situations in which piezoelectric sensors are attached to the sidewall of a

Fig. 5.

PD detection at the same distance between sound source and sensors.

transformer whereas ber optical sensors can be inserted into the transformer tank. A two channel digital oscilloscope (HP Model 54 616C) is used to display and store the acoustic signals detected by the two sensors. and The experimental results with the time scale of 50 the output voltage scale of 1 V for the sensors are shown in Fig. 5, with the sensors at the same distance (0.4 m) from the PD source. The piezoelectric sensor almost shows no response to the acoustic signals even if the output unit is a full scale of 50 mV, as compared to the full scale of 1 V for the optical sensor. Fig. 5 also shows that the electrical signal which initiates the acoustic signal is picked up by the optical and piezoelectric sensors as unwanted noise almost instantaneously. In

WANG et al.: ACOUSTOOPTICAL PD DETECTION FOR TRANSFORMERS

1071

Fig. 6.

PD detection using piezoelectric and ber optical sensors.

Fig. 8.

2-D localization of PD experiment setup.

Fig. 7. Sensor output of the PD and 5-kHz signals.

addition, the sensors pick up the desired acoustic signal at the same time, although later than the electrical signal. Because of its steel housing, the optical sensor cannot be completely unaffected by the electrical signal. This prototype model using the steel housing can always be changed to some material that has no inuence on electrical signals. Since the piezoelectric sensor cannot detect the signals when it is 0.4 m from the PD source, the piezoelectric sensor is mounted inside the tank and much closer to the source while the optical ber sensor is still 0.4 m from the source. The measurement results are shown in Fig. 6 (the time scale is 50 , output voltage scale for the optical sensor is 2 V and for the piezoelectric sensor 5 mV). Due to the piezoelectric sensor being much closer to the PD source than the optical sensor, and the constant sound velocity in the transformer oil, the acoustic signal arrives at the piezoelectric sensor earlier than the optical sensor. But the detected signal is still very weak compared to the optical sensor. These experiments further illustrate the enhanced sensitivity and superior performance of the optical sensor for PD detection compared to the piezoelectric sensor. C. Detection of PD Signals From Surrounding Noises When the optical sensor works inside a transformer, the sensor should be capable of detecting not only a PD signal but also discriminating the PD signals from surrounding noise. An experiment is designed to test the sensors capability to pick up the PD signal from surrounding noises. A function generator is used to generate 5-kHz signals which are sent to a computer speaker. The optical ber sensor receives both the 5-kHz signal and the PD acoustic signal which are then sent to the digital oscilloscope for demonstration and data storage. The experimental result is shown in Fig. 7 with the time and output voltage scale of 1 v. According to scale of 200 the gure, when the sensor is faced with two different acoustic sources, the sensor clearly discriminates the different sources.

Fig. 9. Output of optical sensor 1 with the time scale of 100 s and output voltage scale of 1 V.

Fig. 10. Output of optical sensor 2 with the time scale of 100 s and output voltage scale of 1 V.

D. Two-Dimensional Localization of PD Signal Localization of a PD source is always the main purpose for PD detection. Since the speed of the electrical wave is much faster than that of acoustic wave in transformer oil, the distance between the sensor and PD location can be calculated using the acoustic speed and the time interval between electrical and acoustic onsets. The two-dimensional (2-D) localization experimental setup is shown in Fig. 8. Two optical sensors are placed in a transformer oil-lled tank at different distances from the PD source. The distance difference between the sensors and PD location is about 6 cm. The acoustic signals detected by the sensors are shown in Figs. 9 and 10. The time for the signal to travel from the PD source to the sensors is about 60 for sensor 1 and 100 for sensor 2. Since the distance difference is 6 cm and

1072

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006

Fig. 11.

Oil-lled transformer used in localization experiments.

Fig. 13.

Experimental results at three positions.

length here is centimeter. The experimental results are close to the actual position. IV. CONCLUSIONS The acoustic wave induced by PD can be measured and used for monitoring, diagnosing, and locating potential failures in power transformers. Optical ber-based sensors have been shown to be attractive devices for PD detection because of a number of inherent advantages, including small size, high sensitivity, electrical nonconductivity, and immunity to EMI. An optical sensor was fabricated for PD acoustic detection. The sensitivity of the optical sensor is much high than that of the piezoelectric acoustic sensor. The experimental results shows that the sensor is not only capable to detect PD signal but also localize the signal REFERENCES
[1] R. Bartnikas, Detection of partial discharges in electrical apparatus, IEEE Trans. Electr. Insul., vol. 25, no. 1, pp. 111124, Feb. 1990. [2] L. Niemeyer, A generalized approach to partial discharge modeling, IEEE Trans. Dielectr. Electr. Insul., vol. 2, no. 4, pp. 510528, Aug. 1995. [3] D. H. Nattras, The early history of partial discharge research, IEEE Electr. Insul. Mag., vol. 9, no. 1, pp. 2731, Jan./Feb. 1993. [4] F. Kreuger, Partial Discharge Detection in High Voltage Equipment. London, U.K.: Butterworth, 1989. [5] R. Bartnikas and E. J. McMahon, Engineering Dielectrics, Corona Measurement and Interpretation. Philadelphia, PA: ASTM, 1979, vol. STP 669. [6] W. A. Pledger and S. Pyke, Gas monitor update: Review of progress in the development of a microelectronic in-situ transformer fault gas analyzer, in EPRI Substation Equipment Diagnostics Conf.Electrical Power Research Institute, Palo Alto, CA, 1994. [7] M. Duval, A review of faults detectable by gas-in-oil analysis in transformers, IEEE Electr. Insul. Mag., vol. 18, no. 3, May/Jun. 2002. [8] , Dissolved gas analysis: It can save your transformer, IEEE Electr. Insul. Mag., vol. 5, no. 1, pp. 2227, Jan./Feb. 1989.

Fig. 12.

3-D Localization experimental setup.

measured velocity of the signal is about 1500 m/s (6 cm/40 ). The result agrees with [21]. Since the time difference between electrical noises which is the onset of PD and acoustic signals can be measured, the distance between the sensor and the source can be used to localize the PD source. E. 3-D Localization of PD Signals The oil-lled transformer used for this experiment is shown in Fig. 11. The transformer has a diameter of 80 cm and height of 1.2 m. A detailed explanation of the experimental setup is shown in Fig. 12. Three sensors are used in the experiment. In order to make sure there is a same static pressure applied in the sensors, the sensors are in three positions at the same height of 5 cm below the oil surface. Fig. 13 shows the sensor outputs in three positions inside the transformer. Using the velocity of 1500 m/s, the calculation results tell us , 6.9, 30.3). The the partial discharge source is located in ( actual position in the experiment is (0, 7.5, 30). The unit for the

WANG et al.: ACOUSTOOPTICAL PD DETECTION FOR TRANSFORMERS

1073

[9] [10] [11] [12] [13]

[14] [15]

[16] [17] [18]

[19] [20] [21] [22] [23] [24] [25] [26] [27]

, New techniques for dissolved gas-in-oil analysis, IEEE Electr. Insul. Mag., vol. 19, no. 2, pp. 615, Mar./Apr. 2003. S. A. Boggs, Partial discharge Part III: Cavity-induced PD in solid dielectrics, IEEE Electr. Insul. Mag., vol. 6, no. 1, pp. 1120, Jan./Feb. 1990. H. Ogihara, Detection and location of coronas in oil-immersed transformer with corona detector, Electr. Eng. Jpn., vol. 84, pp. 1222, 1964. J. E. Timperley, Incipient fault identication through neutral RF monitoring of large rotating machines, IEEE Trans. Power App. Syst., vol. PAS-102, pp. 693698, 1983. A. Kheirmand, M. Leijon, K. Hurtig, and H. Vliviita, A system for continuous on-line monitoring and localization of partial discharges in power generators, in Proc. ISH 97, vol. 4, Montreal, Canada, 1997, p. 315. L. Yaqiao, The research on PD detection using pulse current method and ultrasound method, Power Capac., no. 2, pp. 3135, 1996. G. Zhang, G. Wu, and D.-H. Park, Progress of insulation on-line monitoring and fault diagnosis techniques for high voltage power equipment in China, KIEE Int. Trans. Electrophs. and Applic., vol. 12C, no. 1, pp. 114, 2002. A. Wichmann, P. Grnewald, and J. Weidner, Early fault detection in electrical machines by on-line RF monitoring, in Cigr Symp., Vienna, Austria, 1987, pp. 0587. J. T. Phillipson, Experience with RF techniques in the petrochemical industry, in Proc. 4th Int. Conf. Generator and Motor Partial Discharge Testing, Houston, TX, 1996. M. D. Judd, L. Yang, and I. B. B. Hunter, Partial discharge monitoring for power transformers using UHF sensors. Part 1: Sensors and signal interpretation, IEEE Electr. Insul. Mag., vol. 21, no. 1, p. 5, Mar./Apr. 2005. , Partial discharge monitoring for power transformer using UHF sensors. Part 2: Field experience, IEEE Electr. Insul. Mag., vol. 21, no. 2, Mar./Apr. 2005. H. Kawada, M. Honda, T. Inoue, and T. Amemiya, Partial discharge automatic monitor for oil-lled power transformer, IEEE Trans. Power App. Syst., vol. PAS-104, pp. 422428, 1984. R. T. Harold, Acoustical technology applications in electrical insulation and dielectrics, IEEE Trans. Electr. Insul., vol. 20, no. EI-1, pp. 13, Feb. 1985. L. Lundgaard, Acoustic signal propagation from discharge in GIS, in Annu. Rep. IEEE Conf. Electrical Insulation and Dielectric Phenomena, vol. 91, 1991, pp. 449456. , Partial discharge Part XIV: Acoustic partial discharge detection Practical application, IEEE Electr. Insul. Mag., vol. 8, no. 1, pp. 3443, Jan./Feb. 1992. , Partial discharge Part XIII: Acoustic partial discharge detection Fundamental considerations, IEEE Electr. Insul. Mag., vol. 8, no. 1, pp. 2531, Jan./Feb. 1992. B. R. Varlow, D. W. Auckland, and C. D. Smith, Acoustic emission analysis of high voltage insulation, IEE Proc.-Sci. Meas. Technol., vol. 146, no. 5, 1999. R. T. Harold, Acoustic waveguide for sensing and locating electrical discharges in high voltage power transformers and other apparatus, IEEE Trans. Power App. Syst., vol. PAS-98, no. 2, pp. 449457, 1979. M. Leijon, L. Ming, and T. Bengtsson, PD-source identication in solids, in IEEE Seminarian, Baltimore, MD, 1992, pp. 15418.

[28] A. K. Lazarevich, Partial Discharge Detection and Localization in High Voltage Transformers using an Optical Acoustic Sensor, Master, Virginia Polytechnic Inst. and State Univ., 2003. [29] J. Teunissen, C. Helmig, R. Merte, and D. Peier, Fiber optical on-line monitoring for high-voltage transformers, in SPIE Int. Symp. Environmental and Industrial Sensing: Fiber Optic Sensor Technology II, vol. 4204, Proc. SPIE, M. A. Marcus, B. Culshaw, M. Saad, and J. A. Harrington, Eds., 2000, pp. 198205. [30] Z. Zhao and M. Demokan, Optical ber acoustic sensor for sensing and locating partial discharges in high-voltage oil-lled power transformers, in Proc. 21st Australia Conf. Optical Fiber Technology, Gold Coast, Australia, 1996. [31] K. A. Murphy, M. F. Gunther, A. Wang, and R. O. Claus, Detection of acoustic emission location using optical ber sensors, in Smart Structures and Materials 1994: Smart Sensing, Processing, and Instrumentation, vol. 2191, Proc. SPIE, J. S. Sirkis, Ed., 1994, pp. 282290. [32] A. Wang, M. S. Miller, K. A. Murphy, and R. O. Claus, Split-spectrum intensity-based optical ber sensors for measurement of microdisplacement, strain and pressure, Appl. Opt., vol. 35, pp. 25952601, 1996. [33] J. Wang, W. Zhao, H. Xiao, and A. Wang, Self-calibrated interferometric_intensity-based optical ber sensors, in Sensors and Contr. for Advanced Manufact., vol. 3201, Proc. SPIE, B. O. Nnaji and A. Wang, Eds., 1998, pp. 2026. [34] B. Yu, J. Deng, H. Xiao, and A. Wang, Fiber FabryPerot sensors for detection of partial discharges in power transformers, Appl. Opt., vol. 43, pp. 32413250, 2003. [35] X. Wang, B. Li, Z. Xiao, H. T. Roman, O. L. Russo, K. Chin, and K. R. Farmer, An ultra-sensitive optical MEMS sensor for partial discharge detection, J. Micromech. Microeng., vol. 15, no. 3, pp. 521527, Mar. 2005.

Xiaodong Wang, photograph and biography not available at the time of publication.

Baoqing Li, photograph and biography not available at the time of publication.

Harry T. Roman, photograph and biography not available at the time of publication.

Onofrio L. Russo, photograph and biography not available at the time of publication.

Ken Chin, photograph and biography not available at the time of publication.

Kenneth R. Farmer, photograph and biography not available at the time of publication.