Optimal real-time dispatching of chillers and thermal storage tank in a university campus central plant

Recent studies have indicated a great potential for applying predictive methodologies to the operation of Heating, Ventilation and Air Conditioning (HVAC) systems. Particularly for Centralized Chiller Plants with Thermal Energy Storage (TES) used for District Cooling, there is a substantial opportunity for cost savings when responding to variations in electricity price and ambient temperature. The present work addresses the problem of closed-loop scheduling of a large-scale chiller plant with TES tank under a Day-Ahead (DA) electricity price program. The main contributions include: (i) formulating the problem for a real large-scale complex system; (ii) comparing different dispatching policies with varying degrees of optimality, constraint satisfaction and operational complexity; and (iii) designing an optimization-based scheduling tool and describing its implementation and results. The proposed Mixed-Integer Linear Programming (MILP) formulation extends existing models by explicitly accounting for the effect of chilled water return temperature on the energy balances, considering real plant features, and performing pre-calculation of bounds on equipment and storage capacity. The study is performed with real process data from the central chiller plant located on the campus of the University of California, Davis. The comparison between different dispatching policies highlights the benefits of adopting an optimization-based operational strategy, indicating a cost reduction of up to 32% compared to other suboptimal policies. The study also demonstrates how additional constraints that reduce operational complexity affect the closed-loop performance of the optimization-based method, serving as a guideline for designing the optimization tool.