This paper outlines the first steps to enable a set of large industrial freezer warehouse units to act as flexible loads in a smart grid setting. We describe the modeling and identification of individual freezer units, focusing on the governing thermal dynamics and corresponding electric-energy consumption. By relying on the model identified, we show how optimal supervisory control inputs can be computed with the goal of tracking a predefined power consumption schedule. Our algorithm is applied to a real freezer warehouse, and its performance with respect to power reference tracking is evaluated and discussed. We show that the freezer system can follow a predefined power schedule despite the currently limited availability of actuators. © 2013 IEEE.