基于机器学习的变压器超疏水薄膜滤油效果识别  

Judgment of oil-filteration effect for superhydrophobic film based on machine learning

作　　者：张增辉 冯德旺[2] 张云霄 张天峰 池正南 林滔 ZHANG Zenghui;FENG Dewang;ZHANG Yunxiao;ZHANG Tianfeng;CHI Zhengnan;LIN Tao(School of Mechanical and Electrical Engineering,Fujian Agriculture and Forestry University,Fujian 35000,China;School of Computer and Information,Fujian Agriculture and Forestry University,Fujian 35000,China;School of Electrical Engineering and Automation,Fuzhou University,Fujian 350108,China;Super High Voltage Branch of State Grid Fujian Electric Power Co.,Ltd.,Fujian 350013,China)

机构地区：[1]福建农林大学机电工程学院,福建福州350002 [2]福建农林大学计算机与信息学院,福建福州350002 [3]福州大学电气工程与自动化学院,福建福州350108 [4]国网福建电力有限公司超高压分公司,福建福州350013

出　　处：《绝缘材料》2023年第12期98-103,共6页Insulating Materials

基　　金：国家自然科学基金资助项目(51907101);国网福建省电力有限公司科技项目(52130A22000H)。

摘　　要：在长期运行过程中,由于海上风电变压器油老化时易产生水分等杂质,引起变压器绝缘失效故障,从而造成经济损失和安全事故,亟需有效的变压器油滤油和判断方法,以改善变压器油性能与评估变压器油老化状态。本文通过对滤油薄膜进行超疏水改性,探究了滤油次数、薄膜种类和超疏水改性等对变压器油过滤前后性能的影响;通过支持向量机算法,构建了变压器油健康分类模型;提出了一种基于机器学习的超疏水薄膜滤油效果判断方法,对超疏水薄膜滤油效果进行评估。结果表明:经超疏水处理后薄膜的滤油性能得到大幅提升,经3次改性有机膜过滤后,变压器油的综合性能显著提升,符合变压器油的适用标准。对比多个算法发现,利用支持向量机算法构建的模型对变压器油健康分类的精确度最高,达到84.8%。In the long-term operation process,due to the ageing of transformer oil for offshore wind power,the moisture and other impurities will be produced,which may cause transformer insulation failure,resulting in economic losses and safety accidents.Thus,it is urgent to propose an effective method of transformer oil filteration and judgment to improve the transformer oil performance and evaluate its ageing status.In this paper,the oil filter film was conducted superhydrophobic modification,and the effects of the number of oil filtration,the type of film,and the superhydrophobic modification on the properties of oil before and after filtration were investigated.The health classification model of transformer oil was established by support vector machine algorithm.In addition,a new method based on machine learning was proposed to evaluate the oil filtering effect of superhydrophobic film.The results show that the oil filtering performance of the film after superhydrophobic treatment is improved greatly.Especially after three times of modified organic film filtration,the comprehensive performance of transformer oil has significantly improved,meeting the applicable standard of transformer oil.Compared with several algorithms,the model built by support vector algorithm has the highest accuracy of 84.8%.

关 键 词：变压器油 超疏水 电气性能 支持向量机 健康评估 

分 类 号：TM214[一般工业技术—材料科学与工程]