Scholarlink

<p>Accurate and effective prediction of faults in oil-immersed power transformers is essential for maintaining the reliability and stability of power transmission and distribution networks. Among various condition monitoring techniques, dissolved gas analysis (DGA) is widely adopted as a powerful tool for fault diagnosis. The Duval pentagon technique (DPT) is a well-established method for interpreting DGA data and identifying specific fault types. In this study, an intelligent deep learning framework is proposed by leveraging a one-dimensional convolutional neural network (1D-CNN) to classify seven distinct transformer fault categories based on the DPT method. To enhance the model’s performance, the gray wolf optimizer (GWO), a population-based metaheuristic algorithm, is utilized for hyperparameter optimization, resulting in a hybrid CNN-GWO architecture. The proposed model achieves a remarkable classification accuracy of 98.39%, surpassing several state-of-the-art machine learning and deep learning approaches. A comprehensive evaluation is conducted using multiple metrics, including accuracy, precision, F1-score, Matthews correlation coefficient (MCC), recall, confusion matrix, and ROC curve, all of which demonstrate the superior diagnostic capability of the proposed approach. All simulations and validations are carried out in Python using TensorFlow, Keras, and Scikit-learn libraries.</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

Intelligent Fault Diagnosis in Oil-Immersed Transformers: A Deep Learning 1D-CNN Framework with Gray Wolf Optimizer (GWO)

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Elahe Moradi

摘要

Accurate and effective prediction of faults in oil-immersed power transformers is essential for maintaining the reliability and stability of power transmission and distribution networks. Among various condition monitoring techniques, dissolved gas analysis (DGA) is widely adopted as a powerful tool for fault diagnosis. The Duval pentagon technique (DPT) is a well-established method for interpreting DGA data and identifying specific fault types. In this study, an intelligent deep learning framework is proposed by leveraging a one-dimensional convolutional neural network (1D-CNN) to classify seven distinct transformer fault categories based on the DPT method. To enhance the model’s performance, the gray wolf optimizer (GWO), a population-based metaheuristic algorithm, is utilized for hyperparameter optimization, resulting in a hybrid CNN-GWO architecture. The proposed model achieves a remarkable classification accuracy of 98.39%, surpassing several state-of-the-art machine learning and deep learning approaches. A comprehensive evaluation is conducted using multiple metrics, including accuracy, precision, F1-score, Matthews correlation coefficient (MCC), recall, confusion matrix, and ROC curve, all of which demonstrate the superior diagnostic capability of the proposed approach. All simulations and validations are carried out in Python using TensorFlow, Keras, and Scikit-learn libraries.