融合频域介电谱及支持向量机的变压器油浸纸绝缘老化状态评估  

Aging Evaluation of Transformer Oil-Immersed Insulation Combining Frequency Domain Spectroscopy and Support Vector Machine

作　　者：范贤浩 刘捷丰 张镱议[1] 王子潇 Fan Xianhao;Liu Jiefeng;Zhang Yiyi;Wang Zixiao(School of Electrical Engineering Guangxi University,Nanning,530004)

机构地区：[1]广西大学电气工程学院,南宁530004

出　　处：《电工技术学报》2021年第10期2161-2168,共8页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(51867003);广西自然科学基金(2018JJB160064)资助项目。

摘　　要：变压器油纸绝缘老化-受潮协同效应能够显著影响其频域介电谱(FDS)数据,因此,在绝缘严重受潮情况下难以利用FDS技术实现纸绝缘老化状态的准确评估。鉴于此,该文融合FDS以及支持向量机(SVM)提出了一种考虑水分-老化协同效应的变压器油浸纸绝缘老化状态评估方法。在当前的研究中,首先,制备具有不同老化及受潮状态的油浸纸绝缘样本,并基于FDS测试提取用于表征绝缘劣化状态的特征参量。其次,基于多元回归分析构建用于训练老化状态分类模型的拟合介电特征参量数据库。最后,利用拟合介电特征参量数据库和SVM构建用于老化状态评估的分类模型。研究结果证明,构建的分类模型可以准确评估不同受潮状态的变压器油浸纸绝缘样本的老化状态。The aging-moisture synergistic effect of transformer oil-paper insulation can significantly affect its frequency-domain spectroscopy(FDS)data.It is difficult to use the traditional FDS technique to accurately assess the aging state of paper insulation under serious damp conditions.Therefore,combined with the FDS and support vector machine(SVM),a novel method of aging evaluation of transformer oil-immersed paper insulation considering the aging-moisture synergistic effect was proposed.Firstly,the oil-immersed paper insulation samples with different aging and moisture states were prepared,and the feature parameters used to characterize the deterioration state were extracted based on the FDS test.Then,a fitting database of dielectric feature parameters for training classification model of aging state is constructed based on multiple regression analysis.Finally,a classification model for aging state evaluation is constructed based on SVM and fitting database.The research findings prove that the constructed classification model can accurately evaluate the aging state of transformer oil-immersed paper insulation samples with different damp conditions.

关 键 词：变压器 油浸纸绝缘 老化状态评估 频域介电谱 支持向量机 

分 类 号：TM855[电气工程—高电压与绝缘技术]