This paper presents a novel approach for an energy control of a DC microgrid. It combines decentralized grid management and energy management. For this purpose, the conventional voltage droop curves are extended to a characteristic diagram with electricity costs as a further dimension. The support points of these characteristic diagrams are then optimized with a particle swarm optimizer. The target criterion of this optimization is a monetary cost function, that takes several effects, such as depth of discharge, on the operating costs into account. The optimized characteristic diagrams are designed more robust by a sensitivity analysis. The proposed method has been tested successfully in simulations and experiment and was always more cost-efficient than the initial characteristics diagram.