High-Sensitivity Dynamic Detection of Dissolved Acetylene in Transformer Oil Based on High-Power Quartz-Enhanced Photoacoustic Spectroscopy Sensing System

Abstract

To enable the highly sensitive detection of acetylene (C₂H₂) dissolved in transformer oil, a high-power quartz-enhanced photoacoustic spectroscopy (QEPAS) sensing system is proposed. A standard 32.7 kHz quartz tuning fork (QTF) was employed as an acoustic transducer, coupled with an optimized acoustic resonator to enhance the acoustic signal. The laser power was boosted to 150 mW using a C-band erbium-doped fiber amplifier (EDFA), achieving a detection limit of 469 ppb for C₂H₂ with an integration time of 1 s. The headspace degassing method was utilized to extract dissolved gases from the transformer oil, and the equilibrium process for the release of dissolved C₂H₂ was successfully monitored using the developed high-power QEPAS system. This approach provides reliable technical support for the real-time monitoring of the operational safety of power transformers.

1. Introduction

Oil-immersed transformers play a critical role in voltage regulation, energy transmission, and system protection. Their operational integrity directly impacts the safety and stability of the power grid. Statistics show that a single transformer failure can result in an average direct economic loss of USD 125,000 and may trigger cascading failures across the grid [1,2]. Dissolved gas analysis (DGA) technology is used to detect transformer faults by monitoring the concentration and dynamics of characteristic gases dissolved in transformer oil. It has become a cornerstone of transformer condition monitoring systems. Key fault-indicative gases in DGA include carbon monoxide (CO), carbon dioxide (CO₂), methane (CH₄), ethylene (C₂H₄), and acetylene (C₂H₂) [3]. During normal aging, insulating oil primarily generates CO and CO₂. When fault temperatures slightly exceed normal operating levels, CH₄ and C₂H₄ are predominantly produced. In the event of arc discharge—when the arc channel temperature reaches 300 °C and the duration exceeds 5 milliseconds—the concentration of C₂H₂ in the oil can increase tenfold within 24 h. If such high-energy discharges are not promptly detected, they may lead to insulation breakdown and catastrophic failure within 72 h [4,5]. Therefore, real-time monitoring of dissolved C₂H₂ is crucial for ensuring the reliable and safe operation of the power grid.

Various gas analysis techniques have been developed and widely applied for the detection of dissolved gases in transformer oil, including gas chromatography (GC), Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy, tunable diode laser absorption spectroscopy (TDLAS), and photoacoustic spectroscopy (PAS) [6,7,8,9,10,11,12,13]. According to the IEEE standard, GC is regarded as the authoritative method for measuring dissolved gases in transformer oil due to its high sensitivity and capability to simultaneously analyze multiple gas components. However, its frequent calibration requirements, complex operation procedures, and bulky instrumentation hinder its applicability for real-time monitoring and prompt fault diagnosis in transformers. FTIR technology employs a Michelson interferometer and Fourier transform algorithms to analyze the absorption characteristics of dissolved gases at specific infrared wavelengths in the frequency domain [14,15]. This technique enables the simultaneous detection of multiple gas species and simplifies sample pre-processing. Nevertheless, the use of precision optical components increases both equipment and maintenance costs. Raman spectroscopy is capable of detecting characteristic fault gases, but its sensitivity is relatively low compared to other techniques. TDLAS technology offers real-time gas detection with high selectivity; however, its performance depends on laser intensity calibration to maintain measurement accuracy [16,17,18,19,20,21,22]. PAS technology detects weak acoustic signals generated by the interaction between target gases and laser light at specific wavelengths, using an acoustic–electric transducer [23,24,25,26,27,28,29,30]. This approach eliminates the influence of light intensity fluctuations common in traditional absorption spectroscopy. Nonetheless, PAS systems based on broadband response microphones and low-frequency photoacoustic cells are more susceptible to environmental and gas flow noise.

As an alternative approach to PAS, quartz-enhanced photoacoustic spectroscopy (QEPAS) has undergone rapid development since it was first reported in 2002 [31]. In QEPAS, a compact and high-Q-factor quartz tuning fork (QTF) is employed as a sharply resonant acoustic transducer, replacing the conventional microphone [31,32,33,34,35,36,37]. The detection sensitivity of QEPAS is influenced by the acoustic energy concentrated in the gap between the prongs of the tuning fork. To enhance this acoustic energy, Dong et al. optimized the parameters of the acoustic micro-resonator (AmR) and identified the optimal positioning range that yields the highest signal-to-noise ratio [32]. In 2023, Zheng et al. introduced a Helmholtz resonator to QEPAS sensor, utilizing acoustic filtering via Helmholtz resonance to effectively reduce the noise level [33]. Ma et al. demonstrated an H-shaped acoustic AmR that improved laser energy utilization by coupling two off-axis acoustic resonators [34].

Another strategy to increase acoustic energy involves enhancing the excitation laser power. In 2022, Shang et al. employed a high-power multimode laser diode, achieving an order-of-magnitude increase in the QEPAS signal [38]. In 2024, Qiao et al. developed a high-power solid-state laser based on Tm:YAP crystal, which delivered a maximum output power of 150 mW, advancing QEPAS technology through the integration of high-power lasers [39]. Laser power can also be amplified via stimulated emission in optical amplifiers based on doped optical fibers. However, their operational wavelength ranges are inherently limited by the electronic transitions of the dopant ions. Currently, the most widely used Erbium-doped fiber amplifiers (EDFAs) operate within the following bands: the S band (1450–1550 nm), C band (1520–1570 nm), and L band (1565–1610 nm). Therefore, selecting an appropriate fiber amplifier and integrating it with the QEPAS system is an effective strategy for achieving the highly sensitive detection of dissolved C₂H₂ in transformer oil.

In this study, we present a high-power QEPAS sensor based on EDFA for C₂H₂ detection. An EDFA boosted the laser power to 150 mW, enabling a detection sensitivity for acetylene as low as 469 ppb. Dissolved C₂H₂ in transformer oil was extracted using dynamic headspace degassing and monitored in real time with the developed power-enhanced QEPAS system.

2. Materials and Methods

In QEPAS technology, gas molecules absorb laser energy and transition to an excited state, subsequently releasing the energy as heat through the vibration-to-translation (V-T) relaxation process. The thermal power density generated by the gas molecules upon absorbing the laser energy can be expressed as [40]

$$H(r,t)=k_{V\text{-}T}{N}_2(r,t)hv$$

(1)

where k_V-T is the V-T relaxation rate and hv is the energy of the absorbed photon. N₂(r,t) is the number of molecules in the excited state. H(r,t) denotes the thermal power density as a function of the position vector r and time t.

The absorption of light at constant power leads to gas heating but does not produce an acoustic signal. Only temporal variations in the generated thermal power density can induce pressure fluctuations, which manifest as sound waves. Therefore, the fundamental equation of photoacoustic effect can be expressed by the inhomogeneous wave equation

$${\displaystyle \frac{{\partial }^{2}p}{\partial t^2}}-c^2{\nabla }^2(p)=S_{PA}\left(r,t\right)=\left(\gamma -1\right){\displaystyle \frac{\partial H\left(r,t\right)}{\partial t}}$$

(2)

where p is pressure, t is time, c is the speed of sound, and γ is the gas adiabatic coefficient. The source term S_PA(r,t) is proportional to the time derivative of the thermal power density within the medium. Therefore, the generation of a photoacoustic signal requires modulation of the laser, either through intensity modulation (IM) or wavelength modulation (WM), to induce a time-dependent variation in the thermal power density.

Typically, the output wavelength of a diode laser can be modulated by adjusting the injection current. Assuming the laser current I(r,t) follows a periodic modulation, it can be expressed as follows:

$$I(r,t)=I_{DC}(r)+I_{AC}{e}^{jwt}$$

(3)

The source term S_PA(r,t) in the case of WM can be written as

$$S_{PA}(r,t)=j\omega (\gamma -1){\displaystyle \frac{k_{V\text{-}T}}{k_{sp}+k_r+j\omega }}N(r)I_{DC}(r){\displaystyle \frac{\partial \sigma }{\partial \lambda }}{\displaystyle \frac{\partial \lambda }{\partial i}}{I}_{AC}{e}^{j\omega t}$$

(4)

where k_sp represents the rate constant of spontaneous radiation. The effects of all relaxation processes are represented by the rate constant k_r. w represents the modulation frequency of the laser. The acoustic source term is linearly proportional to the concentration N(r) of the absorbing gas molecules, the absorption cross section of the molecular transition, and the intensity of the excitation laser.

3. Results

3.1. Selection of the Laser Wavelength

Currently, commercial fiber amplifier technology in the C-band is well-developed, and is capable of delivering watt-level excitation laser power within a compact footprint. Accordingly, the characteristic absorption lines of C₂H₂ within the range of 1520 nm—1570 nm are selected as the basis for detection. The selection of C₂H₂ absorption line requires consideration of both line strength and potential spectral interference from other gases, such as water vapor (H₂O) and CO₂ in ambient air, as well as CH₄ and CO produced from transformer oil degradation. Based on the HITRAN database, absorption spectrums for 10 ppm C₂H₂, 1% H₂O, 100 ppm CO₂, 100 ppm CO, and 100 ppm CH₄ were simulated at 700 Torr and 296 K, as shown in Figure 1a. The effective optical path length in the simulation is set to 1 cm. The strong C₂H₂ absorption lines were observed in the 1520–1540 nm range; however, this region also exhibited significant H₂O absorption. To minimize interference from water vapor, the spectrum of the 1531–1532 nm range was examined in detail, as shown in Figure 1b. The absorption line at 6529.17 cm⁻¹ demonstrated a strong absorption line strength of 1.165 × 10⁻²⁰ cm/molecule and no overlap with H₂O absorption features. Therefore, this line was ultimately selected as the detection wavelength for dissolved C₂H₂ in transformer oil.

3.2. Experimental Setup

The experimental setup is shown in Figure 2. A distributed feedback (DFB) semiconductor laser with an output wavelength of 1531.6 nm was used as the excitation light source. The output laser power was amplified to 150 mW using an EDFA, which has an output power stability of <2%. The laser driver was operated at 30 °C with a drive current of 90.5 mA. A ramp scanning signal with a frequency of 50 mHz, generated by a function generator, was used to scan the target C₂H₂ absorption line from 6528 cm⁻¹ to 6530 cm⁻¹. A sinusoidal signal with a frequency of 16.384 kHz was generated by a lock-in amplifier (Stanford Research Systems, Model SR830, Sunnyvale, CA, USA), providing high-frequency modulation of the laser. The acoustic detection module (ADM), with a 70-cm³ inner volume, consists of a commercial QTF and an optimized acoustic resonator. The optimal acoustic resonator position is 0.7 mm from the top of the QTF prong. The resonant frequency and quality factor of commercial QTF are 32.76 kHz and 12,035, respectively. The photoacoustic signal generated by the QTF was converted into a voltage signal through a transimpedance amplifier and demodulated by a lock-in amplifier. The integration time of the lock-in amplifier was set to 1 s, with a filter slope of 12 dB/oct.

An oil and gas separation device was designed based on the dynamic headspace degassing method. It comprises an oil sample bottle, a magnetic stirrer, and an internal circulation gas path. A magnetic rotor was placed inside the oil sample bottle and driven by an electromagnetic stirrer located beneath it, enabling continuous agitation of the oil. This promotes the release of dissolved gases into the gas phase above the oil. The separated C₂H₂ gas first passed through a filter to remove oil mist, and then entered the ADM for concentration analysis. The gas flow rate was deliberately maintained at a constant 100 mL/min to minimize flow-induced noise.

3.3. Optimization of the Laser Power

In the high-power QEPAS system, two identical AmRs with an inner diameter of 0.6 mm and an individual length of 4.4 mm were applied to amplify the acoustic wave energy in the form of the standing wave. The commercial QTF was positioned at the acoustic antinode of the AmR, forming an on-beam configuration. The distance between the QTF and the resonator was set to 50 μm to maximize acoustic coupling efficiency. A certified mixture of 500 ppm C₂H₂ in nitrogen (N₂) was filled in the ADM for experimental parameter optimization. Compared to the bare QTF, the addition of the AmR resulted in a 20-fold enhancement, as illustrated in Figure 3. To further enhance the QEPAS signal, a commercial EDFA (BKtel THPOA-SP400ac-FCAPC) with a two-stage amplifier was employed to boost the laser output power from 5 mW to 150 mW. The corresponding spectrum is also shown in Figure 3. As a result, a total gain factor of 200 was achieved in the peak QEPAS signal compared to the bare QTF. This power amplification contributed an additional 10-fold increase relative to the unamplified AmR-QEPAS setup.

3.4. Optimization of the Modulation and Pressure

In QEPAS-based sensors, the signal amplitude is highly dependent on the gas pressure and laser modulation depth. Gas pressure influences the vibration-to-translation (V-T) relaxation rate, the molecular absorption cross section, the absorption line shape, and the molecular number density. The optimal modulation depth depends on the full width at half maximum (FWHM) of the absorption line shape, which broadens with increasing pressure. To maximize the QEPAS signal, both gas pressure and laser current modulation depth were systematically optimized. Figure 4 shows the 2f signal amplitudes for a 500 ppm C₂H₂: N₂ mixture as a function of the modulation depth at different pressures. The experimental results indicate that at pressures below 300 Torr reduced the molecular absorption cross section and slowed V-T relaxation rate limit signal enhancement, even with increased modulation depth. The maximum photoacoustic signal was observed at 600 Torr with a modulation depth of 25mA. Therefore, the atmospheric pressure was selected as the optimal working pressure, with a laser modulation depth of 25 mA for high-power QEPAS sensor applications.

3.5. High-Power QEPAS System Performance Evaluation

The response of the QEPAS signal for various C₂H₂ concentrations is illustrated in Figure 5a. The output wavelength of the DFB laser was fixed at the target absorption line of 1531.6 nm. The C₂H₂ concentration varied from 0 to 1000 ppm, and the signal for each concentration was measured over a 5 min period. After each concentration measurement, the gas chamber was flushed with N₂ for 3 min. The average signal values for each concentration are plotted in Figure 5b. A linear fitting procedure was performed, yielding an R-squared value of 0.999, which confirms the excellent linearity of the high-power QEPAS sensor in response to C₂H₂ concentration. For a 200 ppm C₂H₂/N₂ mixture, a signal-to-noise ratio (SNR) of 426.3 was calculated, based on a signal amplitude of 1.62 mV and a noise level of 3.8 μV. Consequently, the minimum detection limit (MDL) was determined to be 469 ppb.

3.6. Detection of the Dissolved Acetylene Gas in Oil

To evaluate the performance of the high-power QEPAS sensor in detecting C₂H₂ in transformer oil, 900 mL of blank oil sample was spiked with 1% C₂H₂ gas at a flow rate of 100 mL/min for 15 min. Based on the headspace degassing method, the oil sample was continuously stirred on an electromagnetic stirring device. QEPAS signals were recorded at 1 min intervals, as illustrated in Figure 6. The results show that the detected C₂H₂ concentration increased with degassing time, attributed to the higher C₂H₂ concentration in the liquid phase compared to the gas phase. The signal eventually stabilized after approximately 110 min, indicating equilibrium between the liquid and gas phases. The relatively long equilibration time was mainly attributed to the high dissolved C₂H₂ concentration and a low gas flow rate of 100 mL/min in the internal circulation pathway.

4. Discussion

In summary, we presented a high-power QEPAS sensor for dissolved C₂H₂ monitoring in transformer oil. A 1531.6 nm DFB laser and a C-band EDFA were employed to align with the strong C₂H₂ absorption line and deliver up to 150 mW of excitation power. By increasing the laser power and optimizing the gas pressure and modulation depth, a minimum detection limit of 469 ppb was achieved at an integration time of 1 s. The sensor demonstrated effective detection of C₂H₂ concentration ranging from 0 to 1000 ppm. The degassing process of dissolved C₂H₂ in oil was monitored using the headspace degassing method, wherein the blank oil sample was spiked with 1% C₂H₂ gas for 15 min. The liquid–gas phase equilibrium time required was 110 min at an internal circulation flow rate of 100 mL/min. The developed high-power QEPAS sensing system enables the real-time monitoring of dissolved C₂H₂ in transformer oil, offering reliable technical support for fault diagnosis in power transformers. In contrast to conventional GC spectrometers and FTIR spectrometers, which typically occupy volumes in the order of cubic meters, the photoacoustic detection module of the designed high-power QEPAS system features a compact size of only 70 cm³. Additionally, by employing high-power laser excitation, the system enhances detection sensitivity by up to two orders of magnitude compared to traditional photoacoustic spectroscopy techniques. By integrating multiple lasers at different wavelengths with time-division multiplexing, the sensor system can be extended for the simultaneous detection of multiple fault gases, enabling more comprehensive transformer diagnostics.