Model-based predictive control of an ice storage device in a building cooling system

•Simplified building linear thermal model obtained from building simulation model.•Linear model used for the determination of required cooling power.•Use of cooling power exchange as the basis for problem formulation.•Focus on the optimal distribution of chiller/TES to supply cooling requirements.•Significant cost reductions (5–30%) with respect to benchmark control strategies. This paper describes an approach to the formulation of a model-based predictive control (MPC) algorithm for the cooling plant of a building under a time-dependent electricity price profile. The mechanical system includes a three-stage chiller and an ice bank used for thermal energy storage (TES). Cooling can be provided to the indoor space either by directly using the chiller or by discharging the ice bank when electricity prices are high. The chiller is also used to charge the ice bank at night. By applying system identification techniques, a simplified linear thermal model for the building was derived from a detailed building simulation previously developed in EnergyPlus. The use of a simplified linear model – along with weather and internal gains forecasts – allows to readily calculate the required cooling power for a given temperature setpoint trajectory. By making use of simple parametric models for the chiller and the ice bank, an optimization algorithm is applied to decide on the optimal combination of chiller and ice bank cooling power contributions at discrete hourly intervals over the prediction horizon. The length of the prediction horizon alternates between 24 and 30h in order to coincide with the beginning or end of charge/discharge periods. The formulation of the optimization problem is considerably facilitated by using cooling power as the main working variable and then writing the equations accordingly. The proposed MPC strategy is compared with two rule-based control strategies: a modified storage-priority algorithm (similar to the one currently used in the case study building) and a chiller-priority algorithm. With the considered pricing structure and mechanical system, the MPC algorithm results in typical savings of about 5–20% with respect to the modified storage-priority strategy and about 20–30% with respect to the chiller-priority strategy.