Model predictive control of a building energy system including thermal energy storage

Abstract

The present paper aims to verify the effectiveness of a model predictive control (MPC) strategy for an office building subject to an occupancy disturbance and time-varying electricity pricing by comparison with the conventional rule-based control (RBC) strategy. The energy system of the building includes an air-cooled chiller, stratified thermal energy storage (TES) system, two fan coil units, three heat exchangers, and five pumps. The chiller and TES operation were optimally determined by manipulating the mass flow rate of five pumps to minimize the operation cost. In order to construct reliable but computationally light prediction models, an artificial neural network (ANN) was utilized and the epsilon constrained differential evolution with a random jumping (eDE-RJ) algorithm was employed for solving an optimization problem. The simulation was performed for four days in the cooling season with a discrete prediction time horizon of 24 h and control time horizon at 1-h intervals. As a result, the MPC could save approximately 8.3% in terms of the total operation cost in comparison to the RBC, which prioritizes the TES operation to manage the thermal load.