Review of clustering algorithms for microgrid formation

Transitioning from a traditional distribution network grid or diesel only systems to microgrids, offers end-users economic benefits and higher power quality at a reduced environmental cost. Particularly, an upcoming research area, multi-microgrids, aims to provide a more reliable network capable of self-healing. The aim of this paper is to assess well-known clustering algorithms for cost effective microgrid formation and develop a planning framework for uncoupled multi-microgrid networks. In each microgrid, a minimum spanning tree represented the network, resulting in a linear relationship between the microgrid cost and the transmission/power demand. In addition, a diversity factor was introduced to showcase the ability of larger microgrids to more reliably meet peak power demands. Simulation results from three real life datasets suggested that hierarchical clustering algorithms were more suited for microgrid planning due to their adaptability to any datasets, complete solution space search guaranteeing global optimum networks and relative computational efficiency.