Scholarlink

<p>Transformers are critical assets in electrical power systems, and their early fault detection is vital for operational reliability. This study presents an intelligent diagnostic framework based on dissolved gas analysis (DGA), using over 5800 validated transformer cases. Key dissolved gases namely H<sub>2</sub>, CH<sub>4</sub>, C<sub>2</sub>H<sub>6</sub>, C<sub>2</sub>H<sub>4</sub>, C<sub>2</sub>H<sub>2</sub>, CO, and CO<sub>2</sub> along with gas ratios such as Rogers and International Electrotechnical Commission (IEC), are employed to detect insulation degradation and incipient faults. A novel parallel convolutional neural network (CNN) and Support Vector Machine (SVM) named as Parallel Support Vector Machine (PSVM) model is proposed, in which a CNN extracts nonlinear gas features and SVM performs fault classification. The model is optimized using hyperparameters and trained to distinguish thermal faults, arcing, and partial discharge events. For comparative analysis, the PSVM model is benchmarked against traditional diagnostic methods like the Rogers Ratio Method (RRM), Duval Triangle Method (DTM), and modern classifiers including artificial neural networks (ANN), and standalone SVM. The proposed PSVM framework achieves a classification accuracy of 90.9%, outperforming existing models. This work include the development of a hybrid diagnostic model, comprehensive benchmarking with legacy and AI approaches, and the delivery of a scalable, interpretable tool suitable for real-time transformer health monitoring.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Parallel support vector machine classification framework for diagnosis of transformer faults

  • 期刊论文

  • 发表日期：

    2026年1月9日

查看全文
  • Rakeshkumar A. Patel,
  • Jigneshkumar P. Desai,
  • Ravi C. Bhavsar,
  • Satyam M. Joshi

摘要

Transformers are critical assets in electrical power systems, and their early fault detection is vital for operational reliability. This study presents an intelligent diagnostic framework based on dissolved gas analysis (DGA), using over 5800 validated transformer cases. Key dissolved gases namely H2, CH4, C2H6, C2H4, C2H2, CO, and CO2 along with gas ratios such as Rogers and International Electrotechnical Commission (IEC), are employed to detect insulation degradation and incipient faults. A novel parallel convolutional neural network (CNN) and Support Vector Machine (SVM) named as Parallel Support Vector Machine (PSVM) model is proposed, in which a CNN extracts nonlinear gas features and SVM performs fault classification. The model is optimized using hyperparameters and trained to distinguish thermal faults, arcing, and partial discharge events. For comparative analysis, the PSVM model is benchmarked against traditional diagnostic methods like the Rogers Ratio Method (RRM), Duval Triangle Method (DTM), and modern classifiers including artificial neural networks (ANN), and standalone SVM. The proposed PSVM framework achieves a classification accuracy of 90.9%, outperforming existing models. This work include the development of a hybrid diagnostic model, comprehensive benchmarking with legacy and AI approaches, and the delivery of a scalable, interpretable tool suitable for real-time transformer health monitoring.