This paper explores the supervisory control of a microgrid with a flexible cooling system in order to meet load shaping constraints in an economical manner. This load shaping explicitly limits the uncertainty and variability imposed on utility companies by distributed generation. A hierarchical approach is formulated for the integrated operation of the microgrid's controllable loads and dispatchable generation/storage units. Stochastic optimization is used at the hourly time scale to coordinate external energy exchange and determine target profiles for the storage level and building temperature. Deterministic optimization is used at the minute time scale to reject disturbances, determine setpoint trajectories for each unit, and ensure the system is on track to meet long-term goals. The proposed approach is shown to effectively coordinate the decision making at these two different time scales to result in satisfactory closed-loop performance. Moreover, a case study demonstrates that load shaping can be achieved at the microgrid scale at only a small opportunity cost. (C) 2017 Elsevier Ltd. All rights reserved.