Abstract. With the arrival of electric vehicles, battery storage and smart appliances, households now have the opportunity to actively participate in balancing supply and demand in electricity networks. We propose to coordinate this multi-agent system using distributed optimisation, in order to scale to large systems whilst preserving agent privacy. However, the practical applicability of distributed optimisation remains an open question in this context, as AC power flows are inherently nonconvex and households often make discrete decisions about how to schedule their loads. In this paper we show that one such method, the alternating direction method of multipliers (ADMM), can be adapted to remain practical in this challenging microgrid setting. We formulate and solve a multi-period optimal power flow (OPF) problem featuring household agents with shiftable loads, and study the results obtained with a range of power flow models and approaches to managing discrete decisions. Our experiments on a suburb-sized microgrid show that the AC power flows and a simple two-stage approach to handling discrete decisions do not appear to cause convergence issues, and provide near optimal results in a time that is practical for receding horizon control. This brings distributed control of microgrids several steps closer to reality.