基于改进K-means算法的指标阈值告警方法研究  

Research on Indicator Threshold Alerting Method Based on Improved K-means Algorithm

作　　者：许健 王琪 唐海荣 韩少聪 张弛 陈梁 倪洋 Xu Jian;Wang Qi;Tang Hairong;Han Shaocong;Zhang Chi;Chen Liang;Ni Yang(Nanjing NARI Information and Communication Technology Co.,Ltd.,Nanjing 211106)

机构地区：[1]南京南瑞信息通信科技有限公司,南京211106

出　　处：《现代计算机》2022年第20期31-36,共6页Modern Computer

基　　金：南京南瑞信息通信科技有限公司科技项目(5246DR220051):智能一体化运维支撑平台运维工具仓库技术研究及应用。

摘　　要：电网企业信息化运维系统需要监测资源指标的异常状态,由人工录入指标告警阈值规则的配置成本和维护成本较高,且欠缺灵活性。设计一种改进K-means算法对指标历史数据按时间划分,分析出各时段的动态阈值降低人工参与度。算法首先将手肘法和Gap Statistic法相结合来确定最佳聚类数,然后基于变步长萤火虫算法确定初始聚类中心点,最后进行K-means迭代分析输出聚类结果。实验结果表明,改进K-means算法能有效聚类分析出指标的阈值区间,提高了阈值规则的灵活性和电网企业运维现场的告警准确性。The information-based operation and maintenance system of power grid enterprises needs to monitor the abnormal status of resource indicators,and the configuration cost and maintenance cost of indicator alert threshold rules entered manually are high and inflexible.An improved K-means algorithm is designed to divide the indicator history data by time and analyze the dynamic thresholds of each time period to reduce the manual involvement.The algorithm first combines the elbow method and Gap Statistic method to determine the optimal number of clusters,then determines the initial clustering centroids based on the variablestep firefly algorithm,and finally performs K-means iterative analysis to output the clustering results.The experimental results show that the improved K-means algorithm can effectively cluster and analyze the threshold interval of indicators,which improves the flexibility of threshold rules and the accuracy of alerts in the operation and maintenance sites of power grid enterprises.

关 键 词：指标阈值 K-MEANS 手肘法 gap statistic法 变步长萤火虫算法 

分 类 号：TM73[电气工程—电力系统及自动化] TP274[自动化与计算机技术—检测技术与自动化装置]