Novel Functional Community Detection in Networked Smart Grid Systems-Based Improved Louvain Algorithm

Abstract

Despite using Community Detection Algorithms (CDA) in various network partitioning real-world applications, these algorithms tend to fail to partition complex weighted functional networks such as power grids. When contingency events occur, appropriate power grid partitioning enables an effective distributed control strategy to detect system faults and cyberattacks and accelerate electrical assets recovery through fast remedial control. This paper addresses the CDA by considering the functional structure of the power grid. When comparing the topological and functional structures of the power grid network, Electrical Coupling Strength (ECS) reflects both direct and indirect power flows, unlike a conventional adjacency matrix, which only considers an unweighted direct connection between buses. The proposed Improved Louvain Algorithm (ILA) employs the ECS patterns to represent the network connection. Here, the quality function is modified as electrical modularity Qe to assess the goodness of the produced clusters. Large Qe indicates that the power transmission within the partition is dense compared to power transmission with other partitions. The algorithm has been tested on IEEE 39-bus and 118-bus systems. Using different case scenarios, the simulation results demonstrate the high effectiveness of the proposed ILA to the traditional Louvain algorithm in partitioning the complex networked electrical grid into adequate subsystems.