摘要
针对智能物联塑壳断路器存在的多类隐性故障,采用二维特征提取与线性分类方法相结合的方式,对断路器的故障进行分析和诊断。考虑断路器线圈的工作电流波形,获得包含时间、电流幅值以及电流能量在内的多类断路器典型工况特征,深入分析因故障引发的工作电流波形畸变以及对特征造成的谐波干扰,采用核主成分分析(KPCA)方法对故障特征进行约简。同时,采用粒子群优化(PSO)算法对LSSVM方法进行优化,最终构建基于KPCA-LSSVM的断路器故障诊断模型。经过试验验证,该模型解决了因样本冗余以及特征相关造成的故障识别率较低、模型泛化能力不足以及收敛性弱的问题,可有效判断出智能物联塑壳断路器的多类故障隐患发展情况。其在高噪声影响下的故障诊断准确率高达89%以上,具有很强的工程应用价值。
Aiming at the many types of hidden faults in smart Internet of things(IOT)molded case circuit breakers,the analysis and diagnosis are carried out for circuit breakers faults by means of 2D feature extraction combined with linear classification method.The working current waveform of the circuit breaker is studied to obtain several typical working condition features of the circuit breaker,such as time,current amplitude and energy.The working current waveform distortion caused by faults and corresponding harmonic interference for features are deeply analyzed,and the fault features are reduced with kernel principle component analysis(KPCA)method.At the same time,the particle swarm optimization(PSO)method is used to optimize the least square support vector machine(LSSVM)method,and finally,the circuit breaker fault diagnosis model based on KPCA-LSSVM is constructed.Through experimental verification,the model is proved to solve the problem of low fault identification rate and low model generalization caused by sample redundancy and feature correlation,and weak astringency.The hidden trouble developing of many faults is effectively identified for smart IOT molded case circuit breakers.The fault detection rate is as high as 89%under the influence of strong noise.Its engineering application value is superior.
作者
陈颢
王伏亮
李澄
葛永高
包正君
CHEN Hao;WANG Fuliang;LI Cheng;GE Yonggao;BAO Zhengjun(Jiangsu Frontier Electric Technology Co.,Ltd.,Nanjing 211102,China)
出处
《自动化仪表》
CAS
2021年第12期19-22,32,共5页
Process Automation Instrumentation
基金
江苏方天电力技术有限公司一般基金资助项目(YF201903)。
关键词
断路器
故障诊断
核主成分分析
粒子群优化
最小二乘支持向量机
Circuit breaker
Fault diagnosis
Kernel principal component analysis(KPCA)
Particle swarm optimization(PSO)
Least square support vector machine(LSSVM)