The dispatch problem has received considerable attention in the context of large as well as micro grids, especially in the presence of renewable resources and storage. The preferred approach of using model predictive control (MPC) for dispatch is computationally complicated by the presence of complex intertemporal constraints. This paper uses recent advancements in MPC and approximate dynamic programming to improve the computational efficiency of the approach. Specifically, the paper focuses on the design of terminal cost function on the MPC problem. The relative value function of a related problem with average cost criterion is proposed as the terminal cost and demonstrated to save significant computational efforts.