A distributed predictive control approach to building temperature regulation

We study the problem of temperature regulation in a network of building thermal zones. The control objective is to keep zone temperatures within a comfort range while consuming the least energy by using predictive knowledge of weather and occupancy. First, we present a simplified two-mass nonlinear system for modeling thermal zone dynamics. Model identification and validation based on historical measured data are presented. Second, a distributed model-based predictive control (DMPC) is designed for optimal heating and cooling. The DMPC is implemented by using sequential quadratic program and dual decomposition. Simulation results show good performance and computational tractability of the resulting scheme.