Transformer fault diagnosis using machine learning: a method combining SHAP feature selection and intelligent optimization of LGBM

Q2 Energy
Cheng Liu, Weiming Yang
{"title":"Transformer fault diagnosis using machine learning: a method combining SHAP feature selection and intelligent optimization of LGBM","authors":"Cheng Liu,&nbsp;Weiming Yang","doi":"10.1186/s42162-025-00519-3","DOIUrl":null,"url":null,"abstract":"<div><p>This paper proposes a novel approach for transformer fault diagnosis. Initially, a high-dimensional feature set comprising 19 features related to five gas concentrations is constructed to reflect the gas-fault relationship. Subsequently, the Shapley Additive Explanations (SHAP) method is employed to evaluate feature importance and select a subset that significantly influences model predictions, thereby simplifying the model and enhancing its interpretability. Following this, the bald eagle search (BES) intelligent optimization algorithm is utilized to optimize the hyperparameters of the light gradient boosting machine (LGBM) model, further improving its predictive capability. Comparative experiments with various traditional machine learning models validate the effectiveness of the proposed method. The SHAP-BES-LGBM model achieves the highest accuracy of 0.9509 and an f1 score of 0.9606 on the test set, with only 11 samples misclassified, demonstrating superior classification performance and underscoring the advantages of this integrated approach in transformer fault diagnosis.</p></div>","PeriodicalId":538,"journal":{"name":"Energy Informatics","volume":"8 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://energyinformatics.springeropen.com/counter/pdf/10.1186/s42162-025-00519-3","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Energy Informatics","FirstCategoryId":"1085","ListUrlMain":"https://link.springer.com/article/10.1186/s42162-025-00519-3","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"Energy","Score":null,"Total":0}
引用次数: 0

Abstract

This paper proposes a novel approach for transformer fault diagnosis. Initially, a high-dimensional feature set comprising 19 features related to five gas concentrations is constructed to reflect the gas-fault relationship. Subsequently, the Shapley Additive Explanations (SHAP) method is employed to evaluate feature importance and select a subset that significantly influences model predictions, thereby simplifying the model and enhancing its interpretability. Following this, the bald eagle search (BES) intelligent optimization algorithm is utilized to optimize the hyperparameters of the light gradient boosting machine (LGBM) model, further improving its predictive capability. Comparative experiments with various traditional machine learning models validate the effectiveness of the proposed method. The SHAP-BES-LGBM model achieves the highest accuracy of 0.9509 and an f1 score of 0.9606 on the test set, with only 11 samples misclassified, demonstrating superior classification performance and underscoring the advantages of this integrated approach in transformer fault diagnosis.

利用机器学习诊断变压器故障:SHAP 特征选择与 LGBM 智能优化相结合的方法
提出了一种变压器故障诊断的新方法。首先,构建由5种气体浓度相关的19个特征组成的高维特征集来反映气-断层关系。随后,采用Shapley加性解释(SHAP)方法评估特征重要性,选择显著影响模型预测的子集,从而简化模型,增强其可解释性。随后,利用秃鹰搜索(BES)智能优化算法对光梯度增强机(LGBM)模型的超参数进行优化,进一步提高其预测能力。通过与各种传统机器学习模型的对比实验,验证了该方法的有效性。SHAP-BES-LGBM模型在测试集上的准确率最高,为0.9509,f1得分为0.9606,只有11个样本被误分类,显示了较好的分类性能,也凸显了该综合方法在变压器故障诊断中的优势。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
Energy Informatics
Energy Informatics Computer Science-Computer Networks and Communications
CiteScore
5.50
自引率
0.00%
发文量
34
审稿时长
5 weeks
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信