电力电子电路故障评估新指标及基于LSSVM的预测新方法  

An Innovative Metric for Power Electronic Circuit Failure Evaluation and a Novel Prediction Method Based on LSSVM

作　　者：姜媛媛[1,2] 王友仁[1] 罗慧[1] 林华[1] 崔江[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016 [2]安徽理工大学电气与信息工程学院,淮南232001

出　　处：《电工技术学报》2012年第12期43-50,55,共9页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(60871009);航空科学基金(2011ZD52050);中央高校基本科研业务费专项;江苏省普通高校研究生科研创新计划(CXLX11_0183)资助项目

摘　　要：当前电力电子电路故障预测多为元件级,且未考虑电源、负载波动等工作条件影响。针对此问题,提出了一种仅与电路本身故障相关的电路级故障评估新指标——故障特征参数相对变化量,并基于最小二乘支持向量机(LSSVM)算法实现电力电子电路级故障预测。首先,采用LSSVM算法对电路工作条件时间序列、电路参数时间序列预测;其次求解所预测电路未来工作条件下对应的健康电路参数;利用相同工作条件下,健康电路参数与预测的电路参数计算故障评估指标,最终判定未来某时刻电路是否发生故障。以Buck电路为例进行仿真实验验证了方法的可行性和有效性。Currently,fault predictions of power electronic circuit mostly focus on components,and seldom consider the influences of working conditions such as electric network and load fluctuation.An innovative metric only related to the circuit fault for power electronic circuit failure evaluation,named relative shift of fault feature parameter,and a novel method of power electronic circuit fault prediction based on least squares support vector machine（LSSVM） are proposed according to this problem.Working condition time series and circuit parameters time series are predicted based on LSSVM.And then circuit parameters are calculated as the circuit under the predicted working condition is healthy.Thus,the failure evaluation metric is calculated based on the predicted circuit parameters and healthy circuit parameters under same working condition and the future circuit state can be judged finally.Simulation studies on Buck circuit show that the proposed method is feasible and effective.

关 键 词：电力电子电路 故障预测 最小二乘支持向量机(LSSVM) BUCK电路 故障评估指标 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]