摘要
通过监测断路器分合闸线圈电流识别断路器状态是断路器故障诊断重要方法.但是,由于断路器动作频率不高,分合闸线圈电流的数据样本较小.为了在数据样本较小的前提下对断路器进行快速准确的故障诊断,提出了一种基于改进的贝叶斯分类算法的断路器故障诊断方法.针对原始的贝叶斯算法只适用于处理离散型变量的分类问题、应用范畴较为局限的特点,利用入侵杂草优化算法合理选取标准状态,并以此为基础引入基于标准状态概率分配的连续变量离散化方法对特征量进行离散化,对原始的贝叶斯算法进行了改进.研究表明,改进的贝叶斯分类算法将贝叶斯的应用范畴扩展至连续变量的分类问题,提高了故障诊断的准确率.通过仿真分析验证改进的贝叶斯分类算法在不同训练样本数量的情况下故障诊断的准确性,并与原始的贝叶斯算法和支持向量机进行比较.仿真结果表明在训练样本数量为10的情况下,原始贝叶斯算法、支持向量机和改进贝叶斯算法的故障诊断准确率分别为45.05%、83.15%、92.25%,改进的贝叶斯算法故障诊断准确率明显高于支持向量机,说明改进的贝叶斯算法诊断效果更好;改进的贝叶斯算法故障诊断准确率明显高于原始贝叶斯算法,说明入侵杂草优化算法的优化及基于标准状态概率分配的连续变量离散化方法在提高小样本状态下故障诊断准确率方面有良好的效果;改进的贝叶斯算法故障诊断准确率最高,这表明本文所提改进贝叶斯算法能够在样本数据较小的前提下快速准确地对断路器进行故障诊断.
To diagnose faults in a circuit breaker,it is important to determine the status of the circuit breaker by monitoring the current through the circuit breaker’s opening and closing coils.However,considering the low operating frequency of circuit breakers,few data samples of the currents of these opening and closing coils are available.To quickly and accurately diagnose faults in a circuit breaker despite the availability of few data samples,we propose an improved Bayesian algorithm.The conventional Bayesian algorithm is only applicable to classification problems about discrete variables,so we introduce an invasive-weeds optimization algorithm to select the standard state,and on this basis,we use a discretization method based on the standard state probability to discretize the continuous variables.The study results reveal that the proposed improved Bayesian algorithm extends the range of categories to which the Bayesian algorithm can be applied to the classification of continuous variables and improves the accuracy of fault diagnosis.We performed a simulation analysis of different training samples to verify the accuracy of the improved Bayesian algorithm for fault diagnosis and compared the results obtained with those obtained using the conventional Bayesian algorithm and a support vector machine(SVM).For 10 training samples,the fault diagnosis accuracies of the conventional Bayesian algorithm,SVM,and the improved Bayesian algorithm were 45.05%,83.15%,and 92.25%,respectively.The accuracy rate of the improved Bayesian algorithm was greater than that of the SVM,which indicates that the diagnostic effect of the improved Bayesian algorithm is better than that of the SVM.The accuracy rate of the improved Bayesian algorithm was also higher than that of the conventional Bayesian algorithm,which indicates that the optimization obtained by the invasive-weed optimization algorithm and the discretization method based on the standard state probability distribution were effective in improving the fault diagnosis accuracy using a small sample size.Because the improved Bayesian algorithm proposed in this paper had the highest fault diagnosis accuracy,we conclude that it can be used to quickly and accurately diagnose faults in circuit breakers using only a small data sample set.
作者
李永丽
吴玲玲
卢扬
孙广宇
Li Yongli;Wu Lingling;Lu Yang;Sun Guangyu(Key laboratory of Smart Grid of Ministry of Education(Tianjin University),Tianjin 300072,China)
出处
《天津大学学报(自然科学与工程技术版)》
EI
CSCD
北大核心
2020年第6期557-564,共8页
Journal of Tianjin University:Science and Technology
基金
国家自然科学基金资助项目(51577128)
国家重点研发计划资助项目(2016YFB0900603)
国家电网公司科技资助项目(52094017000W).
关键词
断路器
故障诊断
贝叶斯分类器
离散化
小样本
circuit breaker
fault diagnosis
Bayesian classifier
discretization
small sample