基于密度聚类算法的电力通信监测分析

Analysis of Power Communication Monitoring Based on Density Clustering Algorithm

为解决传统基于密度的噪声应用空间聚类(DBSCAN)算法对输入参数设置敏感,以及传统的边缘计算框架计算成本高、计算时间过长等问题,创新性地提出了一种单遍权重K-means(SPWK)聚类算法。构建了电力通信网络故障及入侵监测模型,并将深度强化学习技术与边缘计算相结合,以降低计算成本和计算时长。仿真试验结果表明:SPWK聚类算法的迭代次数更少,平均执行时间以及总聚类时间分别低于其他算法67.5%、37.5%,加速比高出76.4%以上,聚类效率更高;边缘计算优化方法的服务器占用时间以及计算等待时间分别低于其他算法70.4%以上和79.2%以上,性能更优;电力通信监测模型对异常数据的平均识别准确率高出其他算法23.86%以上,入侵检测率高出其他算法4.8%以上,误报率降低65.4%以上,具备优异的检测性能。综上所述,所提故障及入侵监测模型以及边缘计算优化方法的性能均优于其他流行方法,适合在电力通信监测研究中推广使用。

A single pass weighted K-means(SPWK) clustering algorithm is innovatively proposed to solve the problems of the traditional density-based spatial chustering of applications with noise(DBSCAN) algorithm which is sensitive to the input parameter settings, and the traditional edge computing framework with high computational cost and long computation time. A fault and intrusion monitoring model for power communication networks is constructed, and deep reinforcement learning techniques are combined with edge computing to reduce the computational cost and computational time. The simulation test results show that: the SPWK clustering algorithm has fewer iterations, the average execution time and the total clustering time are lower than other algorithms by 67.5% and 37.5%, respectively, and the acceleration ratio is higher by more than 76.4%, so the clustering efficiency is higher;the server occupation time and the computation waiting time of the edge computing optimization method are lower than other algorithms by more than 70.4% and 79.2%, respectively, so the performance is better;the average recognition accuracy of the power communication monitoring model for abnormal data is more than 23.86% higher than that of other algorithms, the intrusion detection rate is more than 4.8% higher than that of other algorithms, and the false alarm rate is more than 65.4% lower, with excellent detection performance. In summary, the performance of the proposed fault and intrusion monitoring model and the edge computing optimization method are better than other popular methods, and are suitable for promotion and use in power communication monitoring research.