Optimal soft early-stop of multi-chiller plants for effective utilization of passive cold storage in buildings

Abstract In large-scale buildings, conventional load-based sequencing strategies often control the chiller plants to fully meet cooling demand until the end of office hours to maintain the indoor comfort, leading to significant waste of cooling energy. Stopping cooling supply earlier could significantly reduce the cooling energy use but may fail to maintain necessary thermal comfort. To address this critical issue, this study proposes an innovative and optimal “soft early-stop” strategy that optimizes the trade-off between the early reduction of cooling supply and thermal comfort. To achieve this objective precisely and reliably, the key technical challenge is to predict the cooling energy demand during the office ending hours. This is solved by introducing the novel concept of “minimum cooling energy demand,” which bridges the gap between usable cold energy stored in the building thermal mass and cooling demand during office ending hours. Usable cold storage is predicted via a resistance-capacitance (RC) model, while cooling demand is forecasted using a TPE-LightGBM model, enabling dynamic optimization of early-stop times and reduced chiller capacity. Simulation test results show that, compared to conventional load-based control, the proposed strategy achieves energy savings of 11.3% and 23.3% for indoor air temperature rises of 1 K and 2 K, respectively, by proactively reducing chiller cooling capacity 30–180 minutes before the end of office hours. This strategy provides robust performance with prediction errors below 10%, offering potential for reducing cooling energy consumption without the need for infrastructure upgrades.