Scholarlink

<p>This paper presents the health index (HI) calculation of a power transformer (for a voltage of 69 kV or less) using different neural networks (NNs) structures and practical data for diagnostic tests. Different structures are tested and compared, then the best structure is highlighted where real field data of 30 transformers are employed to test the different defined NN structures. The suggested NN structure is Backpropagation Feed-Forward (BPFF)-based NN with one hidden layer of ten neurons. The calculation is based on six factors which are water content, acidity, Break Down Voltage (BDV), Dissipation Factor (DF), Total Dissolved Combustible Gases (TDCG), and 2-Furaldehyde. To evaluate the calculated HI values and detect the most efficient NN structure in power transformer HI application, the extracted results of the suggested NN structure are compared with other published results such as conventional General Regression NN (GRNN) and feed-forward-based NN structure with two hidden layers. Finally, to show the effectiveness of the suggested approach, its extracted results are compared to the calculated results for same transformers by the experienced electrical Asset Management Company (AMHA).</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

Health Index Calculation of Power Transformer Using Different Neural Network Structures with Efficiency Evaluation of the Calculated Indices

  • 期刊论文

  • 发表日期：

    2025年5月2日

查看全文
  • Haitham Ramadan,
  • Ahmed Hossam-Eldin,
  • Ahmed Elserougi

摘要

This paper presents the health index (HI) calculation of a power transformer (for a voltage of 69 kV or less) using different neural networks (NNs) structures and practical data for diagnostic tests. Different structures are tested and compared, then the best structure is highlighted where real field data of 30 transformers are employed to test the different defined NN structures. The suggested NN structure is Backpropagation Feed-Forward (BPFF)-based NN with one hidden layer of ten neurons. The calculation is based on six factors which are water content, acidity, Break Down Voltage (BDV), Dissipation Factor (DF), Total Dissolved Combustible Gases (TDCG), and 2-Furaldehyde. To evaluate the calculated HI values and detect the most efficient NN structure in power transformer HI application, the extracted results of the suggested NN structure are compared with other published results such as conventional General Regression NN (GRNN) and feed-forward-based NN structure with two hidden layers. Finally, to show the effectiveness of the suggested approach, its extracted results are compared to the calculated results for same transformers by the experienced electrical Asset Management Company (AMHA).