Optimized Fuzzy Logic Approach with the IEEE Key Gas Method for Diagnosing Power Transformer Faults Using Dissolved Gas Analysis
A fuzzy-logic upgrade to transformer gas diagnosis that separates out carbon monoxide to cut misdiagnoses
The IEEE Key Gas Method (KGM) diagnoses transformer faults from gases dissolved in insulating oil, but its fixed thresholds cause frequent misclassification on ambiguous data. This study built FL-KGM, which combines fuzzy logic with KGM and, crucially, evaluates carbon monoxide (CO) separately from the other gases instead of lumping it in. Tested on real-world data, FL-KGM reached up to 98.6% accuracy, far outperforming the traditional KGM's 62.6% and other fuzzy-based methods.
METAL MEDIA explanatory visual
A fuzzy-logic upgrade to transformer gas diagnosis that separates out carbon monoxide to cut misdiagnoses
- 01The IEEE Key Gas Method classifies transformer fault types (overheating, partial discharge, discharge, etc.) using percentage ratios of seven gases from insulating oil - hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide - but its fixed thresholds lead to up to 50% misclassification in software systems and about 30% error even under expert manual review.
- 02Instead of using all six gases together as before, FL-KGM excludes CO from the main fuzzy calculation and uses five gases (H2, CH4, C2H6, C2H4, C2H2), adding the CO2/CO ratio as a separate auxiliary criterion specifically for judging insulation paper (cellulose) degradation.
- 03The researchers redesigned the membership functions (which define low/medium/high ranges for each gas) and built an 81-rule fuzzy rule set, implementing the system in MATLAB/Simulink using the Mamdani inference method and centroid defuzzification.
- 04On 15 real DGA samples, KGM with CO correctly diagnosed only 7 out of 15 and KGM without CO reached 12 out of 15, while FL-KGM correctly diagnosed all 15; on a larger 150-sample test, FL-KGM achieved 98.6% accuracy compared to traditional KGM's 62.6% and other fuzzy-hybrid models that stayed under 90%.
What they did
- The IEEE Key Gas Method classifies transformer fault types (overheating, partial discharge, discharge, etc.) using percentage ratios of seven gases from insulating oil - hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide - but its fixed thresholds lead to up to 50% misclassification in software systems and about 30% error even under expert manual review.
- Instead of using all six gases together as before, FL-KGM excludes CO from the main fuzzy calculation and uses five gases (H2, CH4, C2H6, C2H4, C2H2), adding the CO2/CO ratio as a separate auxiliary criterion specifically for judging insulation paper (cellulose) degradation.
- The researchers redesigned the membership functions (which define low/medium/high ranges for each gas) and built an 81-rule fuzzy rule set, implementing the system in MATLAB/Simulink using the Mamdani inference method and centroid defuzzification.
- On 15 real DGA samples, KGM with CO correctly diagnosed only 7 out of 15 and KGM without CO reached 12 out of 15, while FL-KGM correctly diagnosed all 15; on a larger 150-sample test, FL-KGM achieved 98.6% accuracy compared to traditional KGM's 62.6% and other fuzzy-hybrid models that stayed under 90%.
Why it matters
Transformers are critical grid components, and early, accurate fault detection directly affects power system stability and maintenance costs. This work offers a data-validated way to reduce the chronic misdiagnosis problem of a widely used industry-standard method, laying groundwork for predictive maintenance systems.
Terms in this paper
- Dissolved Gas Analysis (DGA) · A diagnostic technique that analyzes gases dissolved in transformer insulating oil to detect internal faults
- IEEE Key Gas Method (KGM) · An IEEE standard method that classifies transformer faults based on ratios of specific gases
- Fuzzy logic · A logic system that handles ambiguous or borderline data instead of relying on rigid fixed thresholds
- Membership function · A function that defines how much an input value belongs to categories like low, medium, or high
- Mamdani method / centroid defuzzification · A standard way of converting fuzzy logic outputs into a single definite numeric result
- CO2/CO ratio · An auxiliary indicator used to assess whether insulation paper (cellulose) has degraded
Original abstract (English)
Reliable transformer fault diagnosis is essential for maintaining power system stability. The IEEE Key Gas Method (KGM), a widely utilized approach in Dissolved Gas Analysis (DGA), exhibits limitations in addressing ambiguous data and ensuring high diagnostic accuracy. This study presents An enhanced model combining Fuzzy Logic with the IEEE Key Gas Method (FL-KGM) that introduces refined membership functions, optimized fuzzy rule sets, and a novel separation of CO and CO2 to eliminate diagnostic inconsistencies. By leveraging multidimensional gas ratio analysis and an adaptive classification framework, FL-KGM delivers superior fault identification and classification. Experimental validation utilizing real-world datasets demonstrates that FL-KGM achieves up to 98.6% accuracy, significantly outperforming KGM and other FL-based approaches. These findings elucidate the potential of FL-KGM in advancing transformer monitoring, enabling intelligent fault detection, and enhancing predictive maintenance strategies in modern power systems.
Read on arXivLatest papers
- SWE-bench Science: Can Coding Agents Resolve Engineering Tasks in Science?AI coding agents were tested on fixing real scientific software, and even the best one failed more than half the time
- FlashPrefill V2: Block-Sparse Prefill Attention for Long-Context LLM ServingMaking sparse attention fast enough and accurate enough for real LLM serving, not just papers
- PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM AgentsMaking customer-service AI agents follow the whole procedure, not just avoid one bad action
- EXIMO: VLM Guided Exploration of VLA PoliciesTeaching a robot new chores without human teleoperation, by letting a chatty AI supervise it
- EnvHarness: Awakening Static Worlds for Agent LearningInstead of building new training worlds from scratch, this work adds a plug-in layer that reshapes existing ones around each agent's actual weaknesses
- Bounded Sovereignty and the Control Tax: Pricing AI Oversight When the Deployer Does Not Own the ModelCompanies that rent AI instead of owning it can only do half of AI safety oversight
- PersonalBench: Measuring the Authorship Gap in LLM PersonalizationAI can be prompted to write 'like someone,' but its own voice never fully disappears
- Automated Summarization of Financial News Using Large Language Models and Retrieval-Augmented Generation: An Early Empirical Study (Fall 2023)Testing AI summaries of stock news, the simple approach beat the trendy retrieval-based one