基于最小熵的故障诊断算法-北京邮电大学学报.PDF

2016年6月 北 京 邮 电 大 学 学 报 Jun.2016 第39卷 增刊 Journal of Beijing University of Posts and Telecommunications Vol.39 Sup. 摇 摇 文章编号:1007鄄5321(2016)增鄄0010鄄05 DOI:10.13190/ j.jbupt.2016.s.003 基于最小熵的故障诊断算法 王摇 瑶,摇 刘世栋,摇 郭经红 (国网智能电网研究院信息通信研究所,南京210003) 摘要:针对提升故障诊断的准确度问题,提出一种基于最小熵的故障诊断算法. 首先,依据节点故障概率和各种噪 声的发生概率计算一个可疑节点门限值;其次,计算各个节点对于探测结果的信息增益,与门限值对比,将高于门 限值的节点视为可疑节点,存入可疑节点集合;最后,对可疑节点集合进行排序,将最小熵对应的节点视为故障节 点,并经过仿真实验验证了该算法的有效性. 关摇 键摇 词:电力综合数据网;噪声;熵;故障诊断 中图分类号:TN915郾5摇 摇 摇 摇 文献标志码:A Fault Diagnosis Technology Based on Minimum Entropy WANG Yao,摇 LIU Shi鄄dong,摇 GUOJing鄄hong (Research Institute of Information Telecommunication and Communications,State Grid Smart Grid Research Institute, Nanjing210003,China) Abstract:How to diagnose the fault of network is one of the problems in electric power data network. Currently,one of main factorsthat affect the accuracy of fault diagnosisin power integrated data network is the noise. However,there is almost no mention on thefault diagnosis method in the scheme,because it is difficult to improve the accuracy of fault diagnosis. A new fault diagnosis technique based on mini鄄 mum entropywasproposed. Firstly,theinformationgainof eachnodeiscalculatedaccordingtotheprob鄄 ability of node failure probability and the occurrence probability of each node. Then,the nodes are con鄄 sidered as suspicious nodes. Finally,the nodes are considered as nodes. Key words:electric power data network;noise;entropy;fault diagnosis 摇 摇 随着电力综合数据网的快速发展,网络的可靠 了两种噪声:遗漏噪声(leak),用字母 l 表示,代表 性和稳定性越来越重要. 网络的健康稳定运行离不 探测结果由好变坏的概率;抑制噪声(inhibition),用 开高效、准确的故障诊断. 近年来,网络故障诊断一 字母q 表示,代表探测结果由坏