Model Predictive Control, supervisory control, optimal control, simulation
Model based predictive control (MPC) in building HVAC systems incorporate predictions of weather and occupancy to determine the optimal operating setpoints. However, application of MPC strategies to large buildings might not be real time feasible due to the large number of degrees of freedom in the underlying optimization problem. Decomposing the problem into several smaller sub-problems to be solved in parallel is one way to circumvent the high computational requirements. Such an approach, termed Distributed MPC, requires certain approximations about the underlying sub-problems to converge to a consistent solution thus leading to a trade off between computational load and optimality. In this paper, we present a simulation based evaluation for a Distributed MPC formulation for a case study based on a medium sized commercial building. Results indicate that distributed MPC can offer near optimal control at a fraction of the computational time that centralized optimization based MPC requires while maintaining occupant comfort. Comparison with a few other viable control algorithms will be performed and merits and drawbacks of each approach pointed out.