考虑多因素的大规模设备分区检修决策

发布时间：2018-03-29 22:02

本文选题：谱聚类　切入点：多因素　出处：《高电压技术》2017年01期

【摘要】：针对大规模设备检修决策问题,提出了基于谱聚类算法的多因素设备分区检修模型。首先考虑设备健康指数、设备故障水平、设备价值、社会影响、设备地位等多方面因素建立了设备检修指标的样本空间;然后通过Euclidean距离构造设备相似度矩阵;以设备相似度矩阵的Laplace矩阵相对特征值差自动确定设备分区检修数目,以Laplace矩阵特征向量为依据,确定设备分区检修方案;最后根据以上检修思路,以江西某市级电网为例进行验证。算例结果表明,所提出的谱聚类算法将Laplace矩阵特征向量分别映射到2维和3维向量空间,当相对特征值差处于峰值时,设备检修分区数目为3,故可将设备划分为3个检修分区,从而达到大规模设备分区检修的效果;此外,非规格化Laplace矩阵和规格化Laplace矩阵的分区相关系数分别为0.842 3和0.842 5,说明规格化Laplace矩阵的检修分区效果更好,验证了该算法的有效性。
[Abstract]:According to the large-scale equipment maintenance decision problem, put forward multi factor model partition maintenance equipment based on spectral clustering algorithm. First consider the health index of equipment, equipment failure, equipment value, social influence, many factors such as the establishment of the status of equipment maintenance of equipment sample space; and then through the Euclidean equipment from the structure of the similarity matrix with Laplace matrix; the relative equipment similarity matrix eigenvalue difference automatically determines the number of partitions to equipment maintenance, Laplace matrix feature vector as the basis, to determine the partition equipment maintenance scheme; finally according to the above ideas to Jiangxi city maintenance, power grid is proposed. Numerical results show that the proposed spectral clustering algorithm based on Laplace matrix feature vector is mapped to 2 and 3 dimensional vector space, when the relative eigenvalue difference at the peak, maintenance the partition number is 3, it can be divided into equipment 3 maintenance division, so as to achieve the large-scale equipment maintenance division effect; in addition, the correlation coefficient partition non normalized Laplace matrix and normalized Laplace matrix were 0.8423 and 0.8425, indicating better repair effect of partition the normalized Laplace matrix, verify the validity of the method.

【作者单位】：南昌大学信息工程学院;国网上饶供电公司;
【分类号】：TM507

【相似文献】