基于微型气体传感阵列的空气绝缘设备放电故障识别  

Miniature Gas-Sensing Array Employed for the Discharge Fault Diagnosis of Air-Insulated Equipment

作　　者：王琼苑 褚继峰 李秋霖 杨爱军[1] 袁欢[1] 荣命哲[1] 王小华[1] Wang Qiongyuan;Chu Jifeng;Li Qiulin;Yang Aijun;Yuan Huan;Rong Mingzhe;Wang Xiaohua(State Key Laboratory of Electrical Insulation and Power Equipment Xi'an Jiaotong University ,Xi'an 710049 China)

机构地区：[1]电工材料电气绝缘全国重点实验室(西安交通大学),西安710049

出　　处：《电工技术学报》2023年第23期6494-6502,共9页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(U2166214,52207170);中国博士后科学基金(2022M712510,2022TQ0252);陕西省自然科学基础研究计划(2023-JC-JQ-41);陕西省重点研发计划(2022GXLH-01-11,2022GY-273);电力设备电气绝缘国家重点实验室(EIPE23408,EIPE23314)资助项目。

摘　　要：空气作为一种天然的绝缘气体,在电力设备中(开关柜、环网柜等)被广泛应用。研究表明,当电力设备发生放电故障时,空气绝缘介质会产生以NO_(2)为代表的特征分解产物。放电分解产物的组分及含量能够反映放电故障的严重程度,因此,气体分解产物检测对电力系统安全稳定运行具有重要意义。该文在不同的电压等级、持续时间下,分别模拟了包括电晕放电、火花放电及电弧放电在内的15种空气放电故障,并发现NO_(2)气体的含量在不同故障条件下存在显著差异。对此,设计了一款装载有四种对NO_(2)气体具有高度选择性气敏材料的微型气体传感阵列。经多次实验测试,传感阵列对15种放电故障气体表现出差异性响应信号,构成丰富的样本数据集。分别采用四种机器学习算法(极限树、决策树、K邻近和随机森林)实现基于传感信号的空气放电故障识别,其平均准确率最高可达84.88%、81.82%、76.86%和81.32%。其中,传感阵列对局部放电和电弧放电的识别能力略高于火花放电,可以归因于多次火花放电过程中存在一定的NO_(2)饱和现象。该文提出的基于微型气体传感阵列的检测方法具有操作简单、识别准确率高的显著优势,在空气放电故障诊断领域具有广阔的应用前景。As a natural insulating gas,air is widely used in power equipment(switchgear,ring cabinets,etc.).It will produce characteristic decomposition products represented by NO_(2) when the power equipment occurs internal faults.The content of composition reflects the severity of the discharge fault.Therefore,the detection of gas decomposition products is helpful to the stable operation of power equipment.The cross-sensitivity of the sensor materials towards NO_(2) and CO have seriously restricted the accuracy of fault diagnosis.To address these issues,this paper develops a miniature gas-sensing array with high sensitivity and great reliability,which aimed for the discharge fault diagnosis of air-insulated equipment.This paper has simulated 15 air discharge faults including corona discharge,spark discharge,and arc discharge at different voltage levels and durations.After that,Fourier infrared tests were performed on the 15 gas samples.However,there has no apparent absorption peak of CO and O3 in the infrared absorption spectrum.The absorption peak in 1650~1550 cm-1 band corresponds to the antisymmetric stretching vibration of O=N=O chemical bond.This indicates the presence of NO_(2) in the decomposition products of air.Besides,discharge with different voltages and times shows significant differences in the content of NO_(2).To discriminate fault characteristic gases,this paper has integrated a micro sensor array loaded with four gas-sensitive nanomaterials(10%WO3-10%SnCl2-In2O3,5%NiO-10%SnCl2-In2O3,10%TiO_(2)-10%SnCl2-In2O3 and 5%SnO_(2)-10%SnCl2-In2O3).Then the paper obtains the repeatable gas response recovery curves under different discharge faults,which lays the foundation for the construction of gas identification model.The results demonstrate that the higher voltage and the longer discharge time means the larger response value of sensor array.Accordingly,it is further proved that the higher air-discharge voltages and times generates more characteristic decomposition gas.In addition,t-SNE dimensionality reduction ha

关 键 词：空气绝缘 放电故障模拟 NO_(2)检测 微型传感阵列 故障诊断 

分 类 号：TM930[电气工程—电力电子与电力传动]