Abstract
We report on the development of the first immersion photoacoustic spectrometer (iPAS) for arcing fault detection in power transformers. The spectrometer consists of a detection system and an all-optical photoacoustic sensing head mounted inside a small permeable chamber where dissolved diffuses while the transformer oil is kept out. Our all-optical iPAS sensor can be placed directly inside an oil bath and measure dissolved with the sensitivity and linearity needed for in situ arcing fault detection. Moreover, its fast response time holds great promise for extra-early fault diagnosis.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The failure of a power transformer often leads to discontinuities in the electrical distribution network and gives rise to severe costs for the energy provider. Most early stage thermal and electrical faults of power transformers, however, involve gas production resulting from insulating oil and paper decomposition. Monitoring the concentration of those gases, it is possible to detect the onset of early stage faults and, therefore, initiate the necessary maintenance procedures. It is for this reason that dissolved gas analysis (DGA) has gained worldwide acceptance as a method for the detection of incipient failures in power transformers [1,2]. Among all the key gases that are associated with different types of transformer failures, is the only gas that is specifically released by extremely high-temperature faults (beyond 700°C) [3,4]. While a partial discharge fault may generate traces of , the generation of any amount above a few ppm indicates the occurrence of high-energy arcing. Operating a transformer under sustained high-energy arcing is extremely hazardous and may result in catastrophic failure [5,6]. Hence, the fast response time of a DGA system is essential to provide sufficiently advanced warning of impending arcing failure. A typical DGA process, however, is based on a rather lengthy and inconvenient process where an oil sample is first drawn out of the transformer and then processed to extract the dissolved gases, which are eventually analyzed with gas chromatography or photoacoustic spectroscopy [1]. Even though some of the latest commercial DGA systems can reach a sampling rate of 30 min, the actual response time for fault detection is ultimately limited by the time the dissolved gases take to diffuse from the fault location to the sampling location.
To reduce the response time, it would be ideal to develop DGA systems based on gas sensors that can be directly immersed inside the insulating oil, close to the locations where faults are more likely to happen (e.g., close to the winding). The high electromagnetic field generated by the coils, however, would compromise the working principle of any electronic devices. It is thus not surprising that a number of research groups have proposed to detect dissolved via optical fiber-based Raman or absorption spectroscopy [7,8], or electrochemical sensors [3]. Those approaches showed great potential for real-time arcing fault detection, but seem to lack the sensitivity required by various international guidelines for DGA data interpretation [6,9].
Pushed by the continuous development of miniaturized photoacoustic sensors [10–18], in this Letter we propose the implementation of an immersion photoacoustic spectrometer (iPAS) for DGA. The fiber-coupled sensing head, developed on the basis of a previously tested configuration [17], is fully contained in a small isolation chamber equipped with a permeable membrane that allows dissolved to enter the chamber while keeping liquids outside. We show that our approach guarantees a detection limit of 0.047 ppm (well within standard detection guidelines) over a membrane limited response time of 180 s.
Our iPAS approach is based on wavelength modulation photoacoustic spectroscopy [19]. Here, the wavelength of an excitation laser beam is sinusoidally modulated around one of the rovibrational absorption lines of the gas molecule that has to be detected. The de-excitation process that, after absorption, brings the excited molecules back to their ground state produces a sinusoidal change of the local pressure of the gas (the photoacoustic signal). Measuring the change of local pressure, one can detect the presence and the amount of molecules of that specific gas.
Figures 1(a) and 1(b) show a schematic view of our iPAS system and sensor, respectively. A photo of the inner structure of the sensor is inserted in Fig. 1(b) to indicate the scale. The iPAS sensor, which is similar to the one described in Ref. [17], is mounted inside a gas permeation tube that, upon immersion in oil, protects the sensor while letting gas molecules diffuse through. The gas molecules inside the permeation tube can then further diffuse into the nonresonant cell of the photoacoustic sensor via a small aperture between the inlet of the cell and a flat plate, which is mounted on the free-hanging end of a mechanical beam cantilever. The opposite end of the beam is fixed to an anchor post (anchor fiber in the figure). The plate is everywhere reflective, except for its centermost part, where it is transparent. An optical fiber, aligned with the center of the plate, brings the excitation laser beam in the cell, triggering the photoacoustic signal. Another optical fiber, aligned with the reflective part of the plate, interferometrically detects the vibrations that the photoacoustic signal induces on the beam plate cantilever system, whose amplitude is proportional to the amount of absorbing gas molecules present in the cell.
For this experiment, we used a photoacoustic cell made of a transparent glass capillary tube (length, 13 mm; inner diameter, 600 μm), tapered at the inlet (inner diameter at entrance, 266 μm) and sealed at the opposite end with high-temperature flame. The cantilever was made of a fiberglass filament (diameter of 11 μm as measured on a calibrated microscope, length of 300 μm) equipped with a coated micromirror (, borosilicate glass) suspended at its end. The coating (10/100 nm Cr/Au) at the center of the micromirror (an area of ) was ablated away before mounting to allow transmission of the excitation laser beam. The distance between the micromirror and the cell inlet was set to 15 μm. As for the gas permeation tube, after testing different materials (e.g., platinum-cured silicone tubes and PTFE membranes), we decided to opt for a silicone-coated fiberglass sleeving (out diameter around 3.4 mm; length of 20 mm). The fiberglass supporting braid of the sleeving guarantees a certain structural strength, while the tubular silicone rubber coating layer, which has an average thickness of 115 μm and surface area of , seems to provide sufficient gas permeation for this proof-of-concept experiment. The ends of the fiberglass sleeve were sealed with silicone glue (CAF 4, Elkem Silicones), which may further improve the gas permeation.
Concerning the iPAS excitation system, in this work, we used a near-infrared laser (Oclaro, TL5000VCJ) driven by a current source (Thorlabs, LDC 202C) sinusoidally modulated by the internal function generator of a lock-in amplifier (SRS, SR865). The driving current was set to modulate the excitation wavelength around the absorption line of at 1530.37 nm with a modulation index of 2.4 and frequency of 989 Hz (half the lowest flexural mode frequency of the cantilever). A fiber optic interferometer (Optics11, OP1550) was connected to the readout fiber to detect the cantilever vibration [12], and coupled to the lock-in amplifier to further demodulate the vibration amplitude signal at the cantilever resonance frequency (see [17]). The time constant of the lock-in amplifier was set to 300 ms.
The sensitivity and linearity of the iPAS sensor were calibrated in both gaseous media and oil. For the test in gaseous media, a gas mixture was flushed towards the sensor situated inside a 15 mL centrifuge tube with a flow rate of 100 mL/min [see Fig. 1(c)]. The concentration of was controlled by calibrated mass flow controllers (Bronkhorst EL-FLOW) and varied between 0 and 400 ppm. Excitation laser powers of 24 and 15.8 mW were tested at different concentration steps. For the test in oil, to minimize the diffusion of dissolved gases from the oil sample into the ambient air during the measurements, we inserted the sensor through a hole drilled on the cap of a 15 mL centrifuge tube and sealed the hole with plastic adhesive (3M Scotch-Weld, DP-8005). Then a DGA oil standard (True North) with certificated dissolved concentration was injected into the centrifuge tube (with oil volume slightly over 15 mL). After that, the sensor was inserted in and the cap was tightened as fast as possible [see Fig. 1(d)]. After the lock-in signal was stabilized and enough data were collected, the sensor was taken out, and the above steps were repeated for other concentrations (, , and respectively; ± was defined on a 95% confidence interval for true concentration). The excitation laser power was fixed at 24 mW for all oil sample measurements.
Figure 2 shows the power normalized lock-in R signal (normalized on the photoacoustic signal measured at gaseous concentration of 100 ppm) averaged over 200 raw data points at each concentration step. All the gas concentration values used in the gaseous medium test were converted to equivalent dissolved concentration assuming an equilibrium constant [20] of 0.0816. The coefficient of determination of the linear fit confirms the linear correlation between the photoacoustic signal and the dissolved concentration within the range explored. The bottom and top insets of Fig. 2 show part of the raw lock-in R signal collected as a function of time during the gas medium and oil medium tests, respectively. The bottom inset indicates a membrane limited T90 response time of less than 3 min to step gas concentration change, while the top inset shows a T90 response time as long as 45 min for step-dissolved concentration change, which is limited by insufficient oil circulation around the sensor [21].
The minimum detectable concentration of the sensor for dissolved was determined by wavelength scan [16] when the sensor was immersed in oil sample with dissolved concentration of 10 ppm, as shown in Fig. 3. The peak-to-peak lock-in signal when the excitation laser central wavelength was scanned across the absorption line was equal to 4.49 mV, while the standard deviation of the noise signal collected when the central wavelength fixed away from the absorption line was equal to 21.3 μ V. Hence, a minimum detectable concentration () for dissolved in transformer oil was calculated to be 0.047 ppm. According to IEEE Std C57.104-2008, for a power transformer without dissolved gas history data, dissolved concentration exceeding 2 ppm should prompt additional investigation. Clearly, our iPAS sensor is sufficiently sensitive to detect dissolved with concentration lower than this threshold.
To further validate the ultra early fault diagnosis capability of our iPAS sensor, we tested our system in a custom-made setup designed to simulate a partial discharge fault in a transformer (see Fig. 4). Two pairs of Au-coated electrodes (tip radius of 15 μm) were inserted and sealed into a 15 mL centrifuge tube at two different heights. The tip-to-tip distance of two opposite electrodes was kept as small as possible, while the distance between the two pairs was set to approximately 50 mm. Before the start of the measurement, we poured around 15 mL of new transformer oil (Nynas AB, Nytro Taurus) into the discharge reactor, inserted the iPAS sensor, and closed the cap of the tube to avoid leakage. The vertical distance between the center of the sensor and the bottom electrodes was set to around 14 mm. To simulate the remote sensing scenario in a real transformer, a 15 m long duplex fiber patch cable was used to connect the iPAS sensor to the detection system. To simulate a discharge, we connected the 5 kV output of a negative ion generator to one of the two electrode pairs and turned it on for about 4 s, and then collected the lock-in signal for the following 12 h. The above procedure was repeated multiple times to compare the influence of the sampling distance (the distance between the sensor location and the discharge fault location) on the photoacoustic signal response.
Figure 4 compares the photoacoustic signal response to discharges from the bottom and top electrode pairs. For a better comparison, the lock-in R signal curves were shifted horizontally to align the discharge events at , and vertically to make the value before the discharges equal to 0. From the graphs reported, one can see that the delay time (i.e., the time between the discharge event and the time the sensor starts to detect the dissolved produced by the discharge) for the bottom electrode pair was around 25 min, while for the top electrode pair, it varied from 70–180 min. This experiment not only shows that our iPAS works well in a configuration similar to those that it would encounter in an actual transformer, but also highlights the limits of conventional DGA systems in detecting arcing faults. The distance between an arcing fault location and the oil sampling location typically used in a DGA procedure (e.g., the drain valve) can in fact exceed several meters. From the data reported in Fig. 4, one can see that it could take several days before the produced by an arcing fault diffuses to the sampling location of a traditional DGA system. Although this estimation does not keep into account the oil convection due to the temperature gradient in a real transformer, it still gives a good sense of the kinds of problems that limit standard DGA systems.
In summary, by integrating a miniaturized photoacoustic sensor with gas-permeable material, we extended the applicable domain of photoacoustic spectroscopy from a gas to a liquid medium. Specifically, we demonstrated that our iPAS system can detect dissolved in oil with sufficient sensitivity and linearity for arcing fault monitoring of power transformers. Further evaluation of temperature influence on the photoacoustic signal response is required, as temperature is known to largely vary inside an operating transformer and is expected to affect both the gas absorption coefficient [22] and equilibrium constant of the gas-permeable material [20]. Furthermore, even though the dominant noise emission of a power transformer is below 1 KHz [23,24] (i.e., well below the resonance frequency of the iPAS sensor here described), a thorough characterization of vibration and acoustic noise inside an operating transformer is required before our iPAS approach can be deployed commercially. Yet, it is important to stress that our iPAS system could be well adapted for detection of other transformer fault-related key gases (, , , ), since highly sensitive photoacoustic detection of those gases utilizing fiber-coupled excitation lasers has already been reported previously [17,25,26]. Moreover, we suggest that by placing multiple iPAS sensors inside a transformer at different locations, and connecting them to a single detection system outside the transformer through optical switches, one could realize cost-effective ultraearly arcing fault diagnosis, and it may even be possible to build a triangulation system that could locate the position where the fault has occurred.
Funding
H2020 European Research Council (ERC) (615170); LASERLAB-EUROPE (654148).
Acknowledgment
The authors thank M. Slaman and E. Paardekam for technical support.
Disclosures
DI is cofounder, shareholder, and advisor of Optics11.
REFERENCES
1. N. A. Bakar, A. Abu-Siada, and S. Islam, IEEE Electr. Insul. Mag. 30(3), 39 (2014). [CrossRef]
2. H.-C. Sun, Y.-C. Huang, and C.-M. Huang, Energy Procedia 14, 1220 (2012). [CrossRef]
3. C. Sun, P. R. Ohodnicki, and E. M. Stewart, IEEE Sens. J. 17, 5786 (2017). [CrossRef]
4. I. A. Gray, in Doble Eskom Annual International Conference, Southern Africa (2010).
5. B. H. Chudnovsky, Electrical Power Transmission and Distribution: Aging and Life Extension Techniques (CRC Press, 2012).
6. IEEE, IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed Transformers (IEEE, 2009). [CrossRef]
7. B. Dong, “Fiber optic sensors for on-line, real time power transformer health monitoring,” Ph.D. thesis (Virginia Tech, 2012).
8. T. Somekawa, M. Kasaoka, F. Kawachi, Y. Nagano, M. Fujita, and Y. Izawa, Opt. Lett. 38, 1086 (2013). [CrossRef]
9. G.-M. Ma, S.-J. Zhao, J. Jiang, H.-T. Song, C.-R. Li, Y.-T. Luo, and H. Wu, Sci. Rep. 7, 14961 (2017). [CrossRef]
10. L. Dong, A. A. Kosterev, D. Thomazy, and F. K. Tittel, Appl. Phys. B 100, 627 (2010). [CrossRef]
11. Y. Cao, W. Jin, H. L. Ho, and J. Ma, Opt. Lett. 38, 434 (2013). [CrossRef]
12. G. Gruca, K. Heeck, J. Rector, and D. Iannuzzi, Opt. Lett. 38, 1672 (2013). [CrossRef]
13. J. Rouxel, J.-G. Coutard, S. Gidon, O. Lartigue, S. Nicoletti, B. Parvitte, R. Vallon, V. Zéninari, and A. Gliére, Procedia Eng. 120, 396 (2015). [CrossRef]
14. J. Huber, J. Wöllenstein, S. Kolb, A. Dehé, and F. Jost, Procedia Eng. 2015, 688 (2015). [CrossRef]
15. R. Bauer, G. Stewart, W. Johnstone, E. Boyd, and M. Lengden, Opt. Lett. 39, 4796 (2014). [CrossRef]
16. S. Zhou, M. Slaman, and D. Iannuzzi, Opt. Express 25, 17541 (2017). [CrossRef]
17. S. Zhou and D. Iannuzzi, Rev. Sci. Instrum. 90, 023102 (2019). [CrossRef]
18. J. Peltola, T. Hieta, and M. Vainio, Opt. Lett. 40, 2933 (2015). [CrossRef]
19. S. Schilt and L. Thévenaz, Infrared Phys. Technol. 48, 154 (2006). [CrossRef]
20. L. Tong, G. Wu, J. Sheng, J. Zhang, and L. Zhou, in Conference Record of the 2004 IEEE International Symposium on Electrical Insulation (IEEE, 2004), pp. 97–100.
21. B. Beeskow-Strauch, J. Schicks, and M. Zimmer, Energies 8, 5090 (2015). [CrossRef]
22. L. S. Rothman, I. E. Gordon, Y. Babikov, A. Barbe, D. C. Benner, P. F. Bernath, M. Birk, L. Bizzocchi, V. Boudon, L. R. Brown, A. Campargue, K. Chance, E. A. Cohen, L. H. Coudert, V. M. Devi, B. J. Drouin, A. Fayt, J.-M. Flaud, R. R. Gamache, J. J. Harrison, J.-M. Hartmann, C. Hill, J. T. Hodges, D. Jacquemart, A. Jolly, J. Lamouroux, R. J. Le Roy, G. Li, D. A. Long, O. M. Lyulin, C. J. Mackie, S. T. Massie, S. Mikhailenko, H. S. P. Müller, O. V. Naumenko, A. V. Nikitin, J. Orphal, V. Perevalov, A. Perrin, E. R. Polovtseva, C. Richard, M. A. H. Smith, E. Starikova, K. Sung, S. Tashkun, J. Tennyson, G. C. Toon, V. G. Tyuterev, and G. J. Wagner, J. Quant. Spectrosc. Radiat. Transfer 130, 4 (2013). [CrossRef]
23. Z.-M. Lu, T.-Z. Wang, S.-T. Song, Y. Gang, G.-Q. Liu, C. Zhang, Z.-W. Xia, C.-l. Li, and L. Liang, in International Conference on Electrical Engineering and Automation (ICEEA) (2016).
24. W. M. Zawieska, Int. J. Occup. Saf. Ergon. 13, 381 (2007). [CrossRef]
25. A. Boschetti, D. Bassi, E. Iacob, S. Iannotta, L. Ricci, and M. Scotoni, Appl. Phys. B 74, 273 (2002). [CrossRef]
26. G. Cheng, Y. Cao, K. Liu, G. Zhu, G. Wang, and X. Gao, J. Spectrosc. 2018, 1 (2018). [CrossRef]