The purpose of this paper is to describe a new constrained multivariable predictive control strategy applied in residential HVAC direct load control. In this way, starting from a target load profile prefixed by the utility in order to achieve several objectives-economic, technical-and taking into account customer constraints minimum comfort levels, HVAC limits-this control algorithm provides, in a multiobjective framework, the most suitable HVAC load control strategy to minimize the discrepancies between the controlled load curve and the predefined target load curve. The control system has a closed-loop behavior with a proper and consistent treatment of modeling errors and other disturbances. It is allowed a dynamic modification of the target load curve according to the real-time system behavior. In order to demonstrate its qualities widely, this control system has been applied to modify a real load curve profile under different operating conditions.