智能电网中基于极大团的社团结构挖掘算法

Maximal clique based community structure mining algorithm for smart grid

针对智能电网落后的分区方式无法适应日益复杂的运行状态和无法及时排查故障的问题,提出了一种基于极大团的智能电网社团结构挖掘算法(MCBCA)。首先,搜索网络中的低阶极大团,通过合并矩阵将其合并得到网络中极大团;然后,定义了极大团相似度,确定了合并极大团与生成候选子图的标准,进行初步社团挖掘;最后,对网络中的孤立节点进行隶属度划分,形成最终的社团结构。实验结果表明,在空手道俱乐部网络、美国足球网络、美国国家西部网络及我国省级电力通信骨干网络数据集中,所提算法与KL算法相比,在准确率、模块度及网络抗毁性方面平均提高了50.1%、36.8%和36.2%;与标签传播算法(LPA)相比,在准确率、模块度及网络抗毁性方面平均提高了31.2%,17.7%和3.25%;与改进的GN算法相比,准确率和模块度方面平均提高了3.6%和2.1%。可见基于极大团的智能电网社团挖掘算法所挖掘的网络社团结构更为合理,具备更高的安全性,有利于及时排查故障.

For the problem that the backward zoning method of the smart grid cannot adapt to the increasingly complex operating conditions and cannot check the faults in time,a Maximal Clique Based Community detection Algorithm(MCBCA)for smart grid was proposed. Firstly,triangle maximal cliques were searched in the network,and then were merged through the merging matrix to obtain the maximal cliques in the network. Secondly,the similarity of the maximal cliques was defined,and the criteria for merging maximal cliques and generating candidate subgraphs were determined,so that community mining was preliminarily conducted. Finally,the isolated nodes in the network were divided according to membership degree to form the final community structure. The experimental results show that,on the datasets of Karate Club Network,American Football Network,American National West Network and Chinese Provincial Power Communication Backbone Network,the proposed algorithm averagely has the precison 50. 1%,the modularity 36. 8% and the network survivability when attacked 36. 2% higher than KL(Kernighan-Lin)algorithm;the precison 31. 2%,the modularity 17. 7%and the network survivability when attacked 3. 25% higher than Label Propagation Algorithm(LPA);the precison 3. 6% and the modularity 2. 1% higher than improved GN(Girvan-Newman)algorithm. It can be seen that the network community structure mined by the smart grid community mining algorithm based on maximal cliques is more reasonable,and the higher security is conducive to timely troubleshooting.