This study extends the existing secondary droop control models for frequency synchronization and power-sharing problem to enhance the efficiency of converter-interfaced generators integration into community microgrids. We address this problem by developing models of hierarchical community EMS, which provide synchronization conditions by frequency for an autonomous microgrids community by means of graph theory and Kuramoto-based consensus algorithms for multi-agent reinforcement systems (MARLs). Moreover, we consider the optimal consensus of MARLs in the problem of design controllers for each agent in such a way that all agents will be consensual with the leader agent. The proposed control approach validates via a simulation example of a community microgrid based on the modified CIGRE distribution network.