Transformers are among the most critical and expensive assets in any power system. Their operational reliability directly affects the safety and stability of the entire grid. Dissolved Gas Analysis (DGA) is widely recognized as the most effective method for detecting incipient transformer faults – truly the “blood test” for transformers.
After decades of development, online DGA monitoring has evolved from simple single‑gas hydrogen detection to a landscape with multiple coexisting technologies. This article systematically reviews the principles, characteristics, and application scenarios of the main DGA technologies – gas chromatography, infrared spectroscopy, photoacoustic spectroscopy, and others – to help readers select the most suitable technical solution for their needs.
Single‑gas hydrogen (H₂) monitoring was the first technology applied to online transformer monitoring. Hydrogen offers high sensitivity and fast response, making it capable of detecting incipient internal faults and reflecting most electrical defects.
Advantages:
Low cost and mature technology
Hydrogen is a common byproduct of most early‑stage faults
Fast response
Limitations:
Monitors only one gas – cannot identify fault type
Residual hydrogen trapped in transformer steel structures slowly releases over time, creating background interference
Some faults (e.g., local overheating) may generate methane before hydrogen, leading to missed detection
Delayed fault diagnosis and difficulty in pinpointing the specific problem
Because of these inherent limitations of single‑gas monitoring, multi‑gas monitoring technologies have gradually become mainstream.
Today, DGA multi‑gas monitoring technologies fall into four main categories: gas chromatography, infrared spectroscopy, photoacoustic spectroscopy, and sensor arrays. Among these, gas chromatography is the traditional laboratory standard, while infrared spectroscopy and photoacoustic spectroscopy have become research hotspots for online monitoring in recent years.
Gas chromatography is the classic method for DGA and is recognized as a reference technique by international standards (IEC 60567, IEEE C57.104). Its working principle consists of three steps:
Step 1: Gas extraction
An oil sample is collected from the transformer, and dissolved gases are separated from the oil using vacuum degassing or headspace extraction.
Step 2: Chromatographic separation
The extracted gas mixture is carried by an inert carrier gas (such as helium, nitrogen, or argon) through a chromatographic column. The column is packed with a stationary phase. Different gas molecules interact differently with the stationary phase, so they travel through the column at different speeds, achieving separation.
Step 3: Detection and quantification
The separated gases enter a detector (commonly a Thermal Conductivity Detector, TCD, or a Flame Ionization Detector, FID). The detector output produces a chromatogram – peak position identifies the gas species, and peak area reflects gas concentration.
Advantages:
High sensitivity: Down to 0.1 ppm; detection limit for key gases like acetylene ≤ 0.5 ppm
High accuracy and repeatability: As a laboratory standard, results are reliable
Comprehensive gas coverage: Can measure nine or more fault gases simultaneously
Mature technology: Well‑established international standards and extensive application experience
Limitations:
Requires carrier gas: Consumes high‑purity inert gases, increasing operating costs
Requires regular calibration: Needs standard gas mixtures for calibration
Discontinuous monitoring: Traditional GC is batch‑based – cannot provide real‑time continuous monitoring
Complex maintenance: Columns, valves, and other parts require regular servicing
Time delay: From sampling to results takes hours to days
Laboratory offline analysis: Used as a benchmark method for fault confirmation
Periodic inspections: Regular sampling for non‑critical transformers
Online monitoring systems: Miniaturized GC systems exist, but they still consume carrier gas
Studies show that GC and photoacoustic spectroscopy (PAS) exhibit high consistency in measuring dissolved gas concentrations in oil; both technologies effectively determine gas composition and concentration.
Infrared spectroscopy detects gases based on their absorption of specific infrared wavelengths, following the Lambert‑Beer law. Different gas molecules have unique infrared absorption spectra – their “fingerprint” spectra.
Principle:
FTIR uses a Michelson interferometer to generate interference light. After passing through the gas sample, the light undergoes Fourier transformation to produce an infrared absorption spectrum. The position of absorption peaks identifies gas species, and the peak intensity corresponds to gas concentration.
Characteristics:
Can measure multiple gases simultaneously
High spectral resolution and good selectivity
Relatively complex optical path system
Limited sensitivity for trace gases (typical detection limit ~0.5 ppm)
Principle:
TDLAS uses a tunable semiconductor laser whose emission wavelength is precisely aligned with the characteristic absorption line of the target gas. By scanning the laser wavelength and measuring the absorption peak intensity, gas concentration is quantified.
Characteristics:
Can measure only one or a few gases per measurement
High sensitivity, down to ppb levels
Strong anti‑interference capability and good selectivity
Requires multiple lasers or wavelength scanning for multi‑component measurement
Characteristics:
Relatively simple structure and lower cost
Moderate selectivity – susceptible to interference from other gases
Detection limit ~0.5 ppm
Commonly used for single‑ or two‑component gas detection
Photoacoustic spectroscopy is a rapidly developing optical gas detection technology. Its principle is based on the photoacoustic effect: when gas molecules absorb infrared light at a specific wavelength, they transition from the ground state to an excited state. Through collisional relaxation, the absorbed energy is converted into heat, causing a transient temperature rise. When the light source is modulated, the periodic heating generates pressure waves – i.e., sound waves. A high‑sensitivity microphone detects the sound wave intensity, which is directly proportional to gas concentration.
The key formula for PAS can be expressed as:
S = k · α · P · C
Where:
S = photoacoustic signal intensity
k = instrument constant
α = gas absorption coefficient
P = optical power
C = gas concentration
Advantages:
No carrier gas required: Direct gas measurement – no consumable carrier gas
High sensitivity: Detection limit for key gases like acetylene reaches 0.1–0.5 ppm
Fast response: Suitable for real‑time continuous monitoring
No moving parts: When using MEMS electronic modulation, reliability is high
Low maintenance: No consumables
Limitations:
Traditional PAS uses a mechanical chopper to modulate the light source, introducing vibration noise
Environmental vibration may affect measurement accuracy
High demands on microphone sensitivity
Modern enhanced PAS technology overcomes the limitations of traditional PAS through several innovations:
MEMS infrared light source: Electronic modulation replaces the mechanical chopper, eliminating vibration noise
Dual‑chamber enhanced gas cell: Gold‑coated absorption cavities with resonance enhancement technology improve detection sensitivity
Vacuum degassing: Temperature‑controlled vacuum degassing compatible with multiple oil types
Enhanced PAS achieves consumable‑free, maintenance‑free, full‑gas‑coverage online monitoring capability.
Studies have demonstrated high consistency between PAS and GC in measuring dissolved gas concentrations in various oil samples. One comparative experiment using 30 oil samples from transformers in service for more than three years confirmed that both technologies effectively determine gas composition and concentration.
Differences in PAS measurements observed in the experiment were mainly attributed to air introduced during sampling, which caused slightly elevated CO₂ and CH₄ concentrations. This highlights the importance of proper sampling procedures rather than any inherent technology deficiency.
Raman spectroscopy is based on the Raman scattering effect. It analyzes gas composition by measuring the scattering spectrum produced when laser light interacts with gas molecules.
Characteristics:
No sample preparation required – can measure directly in oil
A single laser can simultaneously detect multiple gases
Relatively low sensitivity – requires enhancement techniques
Higher system cost
Current status:
With advances in laser and fiber optic technologies, RS is gradually expanding its role in DGA. Enhanced RS systems have reduced detection limits for gases such as C₂H₂ and CH₄ to tens of ppm, sufficient for online monitoring applications.
Sensor arrays consist of multiple gas‑sensitive sensors. Gas composition is analyzed using pattern recognition algorithms.
Characteristics:
Relatively low cost
Small size – easy to integrate
Selectivity and stability need further improvement
Significantly affected by ambient temperature and humidity
(A detailed comparison table would be placed here. Below is a textual summary.)
| Technology | Principle | Detection Limit (C₂H₂) | Carrier Gas | Maintenance | Best For |
|---|---|---|---|---|---|
| GC | Separation + detection | ≤0.5 ppm | Required | High | Laboratory / reference |
| PAS | Photoacoustic effect | 0.1-0.5 ppm | Not required | Low | Online continuous |
| FTIR | Infrared absorption | ~0.5 ppm | Not required | Medium | Multi‑component online |
| TDLAS | Laser absorption | <0.1 ppm | Not required | Low | Single/dual gas high precision |
| NDIR | Infrared absorption | ~0.5 ppm | Not required | Low | Single gas |
| Raman | Raman scattering | 10-50 ppm | Not required | Medium | Direct in‑oil measurement |
Sources: Multiple peer‑reviewed studies and technical datasheets.
Traditional GC requires carrier gas, increasing operating cost and complexity. Optical technologies (PAS, infrared spectroscopy, etc.) require no carrier gas and represent the mainstream direction for future online monitoring.
Moving from single‑gas to multi‑component monitoring, combined with moisture, temperature, and other parameters, enables comprehensive transformer condition assessment.
Different technologies have different strengths: GC offers high accuracy, PAS provides maintenance‑free operation, and TDLAS delivers exceptional sensitivity. Future trends favor integrating multiple technologies to achieve complementary advantages.
Combining artificial intelligence (AI) and machine learning algorithms enables automatic fault type identification and severity assessment – overcoming the dependency on expert experience and inconsistent diagnostic results of traditional DGA methods. Studies show that artificial neural networks (ANN) can achieve 76.8% accuracy in DGA fault diagnosis, significantly higher than traditional methods (Dornenburg 55%, Duval triangle 40%, Roger 38.4%, IEC 31.8%).
| Application Scenario | Recommended Technology | Rationale |
|---|---|---|
| Laboratory benchmark analysis | GC | Highest accuracy, standard‑approved |
| Critical transformer online monitoring | Enhanced PAS | Maintenance‑free, full gas coverage |
| Medium‑voltage transformers | 3‑gas PAS or NDIR | Moderate cost, covers main faults |
| Fast leak localization | TDLAS | High sensitivity, fast response |
| Distribution transformer early warning | Single‑gas hydrogen monitor | Low cost, meets basic needs |
| Research & in‑depth diagnostics | Raman or FTIR | Multi‑dimensional information |
Dissolved gas analysis for transformer oil has evolved from single‑component to multi‑component, from offline to online, and from electrical to optical methods.
Gas chromatography, as the traditional standard method, remains irreplaceable in laboratory applications due to its high accuracy and comprehensive coverage.
Photoacoustic spectroscopy, with its maintenance‑free and consumable‑free characteristics, has become the ideal choice for online continuous monitoring.
For power utilities, selecting the right DGA technology requires balancing asset criticality, budget constraints, maintenance capabilities, and diagnostic needs. As next‑generation technologies like enhanced PAS mature, maintenance‑free, full‑gas‑coverage, intelligent diagnostic online monitoring is becoming a reality – providing strong support for the transition from “periodic maintenance” to “predictive maintenance” of power equipment.
HERTZINNO’s online DGA systems (DGA900, DGA500, DGA300) use enhanced MEMS‑based photoacoustic spectroscopy to deliver consumable‑free, maintenance‑free continuous monitoring, with flexible configurations ranging from single‑gas to 9+1 (nine gases + moisture). Learn more →
