This work studies how the communication network between proactive consumers affects the power utilization and fairness in a simplified direct-current micro-grid model, composed by three coupled layers: physical (an electric circuit that represents a micro-grid), communication (a peer-to-peer network within the micro-grid) and regulatory (individual decision strategies). Our results show that, for optimal power utilization and fairness, a global knowledge about the system is needed, demonstrating the importance of a micro-grid aggregator to inform about the power consumption for different time periods.Index Terms-Communication networks, Agents-based systems, Network topology, Smart grids, Decision making. Our model is described as follows. The physical layer (a DC micro-grid) is a circuit composed by a power source and resistors in parallel. Individual agents (the proactive consumers, the "prosumers") can add, remove or keep the resistors they have. Agents' decisions aim at maximizing their own delivered power, which is a non-linear function dependent on the others' behavior, and they are based on (i) their internal state, (ii) their global state perception, (iii) the information received from their neighbors in the communication network, and (iv) a randomized selfishness related to their willingness of answering to a demand-side control request. Different peerto-peer communication network topologies and randomized communication errors in the deployment of micro-grids are important aspects to be included as in [17], [18]. But differently from other papers, we analyze here their systemic effects; for detailed results of communication network implementation in power grids using, for instance, spatial and/or temporal spectrum sharing, refer to [19]- [23].Looking at the proposed model, by individually modifying the peer-to-peer communication network topology and the demand-side management policies keeping fixed the other