Analysis of Clustering Algorithms Applied to Distribution Network Reconfiguration

Objective: The study aims to speed up the reconfiguration process of electrical distribution networks. Using graph theory, it explores methods for the network's topological decomposition and searches for lower-cost paths to optimize the distribution network configuration. Related Studies: The research is based on graph theory, which allows modeling the connections of an electrical network as an interconnected graph. Important references include clustering techniques and algorithms for finding minimum paths in graphs, applying methods such as Malgrange, K-means, DBSCAN, SOM, and classical graph search algorithms like Dijkstra and Floyd-Warshall. Method: The methodology includes applying different algorithms for topological reduction and path search, analyzing their performances. The IEEE 37-bus system was used as a test model. The studied algorithms included Malgrange, K-means, Ward (hierarchical), DBSCAN, SOM, and minimum path algorithms. Results and Discussion: The results indicated that the Malgrange method brought a significant reduction in the search space, with a decrease of approximately 83.5% in possible switch combinations. However, in larger and more complex networks, increasing the number of clusters could contribute to even greater efficiency. The research concluded that minimum-path search algorithms produced similar results. Research Implications: The research offers a methodology that can be applied in power distribution networks to reduce computational time and improve the efficiency of reconfiguration processes. Practical implications include direct use by power distribution companies looking to optimize their networks. Theoretically, it reinforces the use of graph theory in network optimization problems. Originality/Value: The originality lies in combining graph techniques for decomposition and minimum-path search in electrical networks, with significant optimization results. The research contributes insights for future implementations of reconfiguration in complex networks, highlighting the value of graph analysis as a robust approach to electrical distribution problems.