Source-linked AI summary
Optimized Fuzzy Logic Approach with the IEEE Key Gas Method for Diagnosing Power Transformer Faults Using Dissolved Gas Analysis
Kim-Anh Nguyen, Huy Hoang Le, Ba Tu Phung
TL;DR
The IEEE Key Gas Method can misclassify ambiguous or complex transformer faults. This paper develops an enhanced fuzzy-logic version that separates CO from CO2 and achieves up to 98.6% accuracy, outperforming KGM and other fuzzy-based approaches.
Problem
The IEEE Key Gas Method can produce misclassifications or inconclusive diagnoses in ambiguous, borderline, and complex transformer-fault cases.
Method
FL-KGM combines redesigned membership functions and optimized fuzzy rules with five-gas analysis, CO exclusion, and a CO2/CO criterion for thermal paper faults.
Results
Up to 98.6% accuracy was achieved, surpassing KGM and several other fuzzy-based approaches in transformer fault classification.
Takeaways & Limitations
FL-KGM shows practical potential for transformer condition monitoring by maintaining stable diagnosis across varied gas concentrations and complex failure patterns.
Takeaways & Limitations
The IEEE Key Gas Method’s predefined gas categories may not reflect real-world failure profiles, where overlapping signatures can produce diagnostic inconsistencies.
Abstract
from arXiv · showhide
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.
I. INTRODUCTION (HEADING 1)
The study addresses the need for early, accurate transformer-fault diagnosis by introducing FL-KGM, which independently evaluates CO and CO₂ and redesigns fuzzy logic analysis using real-world datasets. The model is validated through Matlab/Simulink simulations for adaptive fault classification.
- Motivation: Transformers are critical to energy transmission and distribution, while failures cause economic losses and threaten power-system stability and reliability.These consequences make early fault detection and accurate diagnosis essential.
- Methodological contribution: The proposed method independently evaluates CO and CO₂ instead of considering them collectively.This adjustment prevents CO from disproportionately influencing diagnosis, particularly for insulation degradation and thermal-condition assessment.
- Methodological contribution: A completely redesigned fuzzy-logic system for key-gas analysis uses real-world datasets to support data-driven and adaptive fault classification.The redesigned FL system forms part of the proposed FL-KGM approach.
- Validation: FL-KGM is validated through simulations in Matlab/Simulink.The passage identifies simulation-based validation but does not specify additional validation metrics here.
II. PROPOSED FUZZY LOGIC-BASED IEEE KEY GAS MODEL
The study develops an effective fault diagnosis model by examining traditional key gas methods and enhancing the fuzzy-logic-based IEEE Key Gas Method (FL-KGM). It focuses on gas-generation percentage ratios and fuzzy logic fundamentals.
- Model development: The study focuses on developing an effective transformer fault diagnosis model.This development is framed around two core issues addressed by the proposed approach.
- Traditional key gas methods: It examines traditional key gas diagnosis methods using percentage ratios of gases generated during transformer operation.The analysis evaluates gas-generation percentages as the basis of traditional key gas fault diagnosis.
- Fuzzy logic enhancement: It explores fuzzy logic fundamentals while enhancing the FL-KGM approach.The passage identifies fuzzy logic exploration and FL-KGM enhancement as the second core issue.
A. IEEE Key Gas Method and Limitations
The IEEE Key Gas Method is a widely adopted, IEEE-recognized DGA approach that relates generated-gas ratios to transformer fault types. However, it can produce substantial misclassification or inconclusive results, especially in complex or borderline cases.
- DGA foundation: DGA diagnoses potential transformer faults by analyzing gases dissolved in insulating oil.Fault-related gases include H₂, CH₄, C₂H₆, C₂H₄, C₂H₂, CO, and CO₂.
- Key gas analysis: Key gas analysis correlates specific gas formations with distinct transformer fault types and includes the IEEE Key Gas method among its established approaches.Other cited approaches include LCIE, CSUS, TDCG, and the model proposed by Muller et al.
- IEEE Key Gas Method: The IEEE Key Gas Method, developed by Doble Engineering Company in 1973 and refined by David Pugh in 1974, became an industry standard recognized in IEEE Std C57.104.It analyzes gas ratios produced under thermal and electrical stresses, focusing primarily on hydrogen, hydrocarbons, and carbon monoxide.
- Method limitations: Up to 50% of software-based diagnoses using the IEEE Key Gas Method may be misclassified or inconclusive, while expert manual interpretation retains an approximately 30% error rate.These limitations highlight difficulty handling complex or borderline transformer-fault cases.
B. Proposed FL-KGM Model
The proposed FL-KGM model combines fuzzy logic with IEEE key-gas rules to handle ambiguous transformer DGA data. It redesigns gas membership functions, optimizes fuzzy rules, and adds CO₂/CO assessment to improve diagnostic reliability.
- Fuzzy logic foundation: Fuzzy logic uses fuzzification, fuzzy computation, and defuzzification to infer transformer fault conditions from ambiguous gas data.The approach is motivated by fuzzy logic’s ability to address uncertainty in dissolved gas analysis.
- Limitations of traditional KGM: The traditional IEEE KGM uses percentage ratios of H₂, CH₄, C₂H₂, C₂H₄, C₂H₆, and CO but can suffer ambiguity and CO-related diagnostic distortion.These limitations may produce suboptimal accuracy, particularly in exceptional cases.
- Membership-function optimization: FL-KGM redesigns membership functions for H₂, CH₄, C₂H₂, C₂H₄, and C₂H₆ by rationally redividing value ranges and adjusting their shapes and slopes.The redesigned functions are based on the ranges in Table II and are presented in Table III and Fig. 3.
- Rule-set optimization: The optimized fuzzy rules address overlapping gas-ratio thresholds by integrating multiple ratios, reducing uncertainty when compositions fall near or outside predefined fault limits.This framework is intended to reduce inconclusive outcomes and misdiagnoses across diverse operating conditions.
- CO₂/CO assessment: For F1 thermal faults, FL-KGM evaluates the CO₂/CO ratio; values below 3 indicate intact insulation, while higher values signal possible insulation degradation or abnormal temperature conditions.The diagnostic framework first checks transformer operating status, then analyzes dissolved-gas ratios when abnormalities are detected.
III. SIMULATION AND EVALUATION OF THE FL-KGM MODEL
The improved FL system was evaluated through MATLAB/Simulink simulations using a real-world DGA dataset, comparing it with the traditional KGM and highlighting improved diagnostic accuracy.
- Simulation setup: The KGM and improved FL system were implemented and tested through MATLAB/Simulink simulations.The evaluation focused on the improved FL system’s effectiveness.
- Evaluation dataset: A real-world DGA dataset was used to assess fault identification capabilities and compare the traditional method with the enhanced fuzzy logic system.The comparison examined diagnostic performance using simulation data derived from real-world DGA records.
- Results: The simulation results highlighted improved diagnostic accuracy for the proposed FL system.The results were used to demonstrate the extent of the proposed system’s diagnostic improvement.
A. MATLAB/Simulink-Based Simulation Model
The MATLAB/Simulink model integrates verification, preprocessing, fuzzy-logic, CO₂/CO ratio analysis, and result-display blocks to process DGA data and diagnose abnormalities.
- Model Components and Data Processing: The model includes normal-state verification, data preprocessing, FL, CO₂/CO ratio analysis, and result-display blocks for DGA-based fault diagnosis.Preprocessing extracts H₂, CH₄, C₂H₆, C₂H₄, C₂H₂, CO, and CO₂ concentrations and calculates percentage ratios using KGM before fuzzy inference.
B. Simulation outcomes and assessment
Simulation assessment across selected and comprehensive DGA datasets shows that FL-enhanced diagnosis improves fault-identification accuracy over traditional KGM. The proposed FL-KGM also provides stable, adaptive performance across varying gas concentrations and complex fault patterns.
- Selected-sample assessment: 12 out of 15 samples were accurately diagnosed by KGM when CO gas percentages were excluded, compared with 7 out of 15 when CO was included.The samples covered normal operation, PD, T, and D fault conditions.
- Comprehensive evaluation: 150 DGA samples were used to evaluate diagnostic reliability and generalizability across various fault categories.Accuracy was compared across conventional KGM and FL-enhanced IEC, RRM, and DRM methodologies.
- Comprehensive evaluation: Traditional KGM remained below 85% accuracy in several fault scenarios, indicating vulnerability to misclassification when fault signatures overlap.The observation was derived from results on a comprehensive dataset spanning different fault types.
- Comprehensive evaluation: All four FL-enhanced methods exhibited an advantage over traditional KGM in the comprehensive assessment.The passage contrasts the distinct advantage of FL-enhanced methods with KGM’s variable performance across fault types.
- Robustness and adaptability: FL-KGM dynamically adapts to diverse fault profiles, maintaining stable performance despite gas-concentration variations and complex failure patterns.This adaptability avoids the manual threshold adjustments frequently required by conventional techniques for different transformer operating conditions.
IV. CONCLUSION
The study integrates an enhanced fuzzy-logic model with the traditional IEEE Key Gas Method, using optimized membership functions, expanded fuzzy rules, and auxiliary criteria to improve transformer fault classification. Experimental results indicate that the improved model achieves up to 98.6% accuracy.
- Model advancement: The enhanced FL model integrates with the traditional KGM to advance transformer fault diagnosis.The integration is presented as a significant advancement in the study.
- Model enhancements: Optimized membership functions and expanded fuzzy rule sets improve processing of complex data.The model also incorporates auxiliary criteria, including the CO₂/CO ratio.
- Experimental outcome: 98.6% accuracy is achieved by the improved FL model in experimental results.The supplied passage reports this as an upper-bound result for the improved model.