Development of a probabilistic cooling load prediction-based robust chiller sequencing strategy and its real-world implementation

Abstract

Multiple-chiller systems are widely adopted in large buildings owing to their energy efficiency and flexibility. Robust chiller sequencing is important for the energy-efficient and reliable operation of multiple-chiller systems. Conventional direct and indirect chiller sequencing strategies are not robust enough because they are affected by fluctuations in measurements. To address this challenge, this study proposes a novel chiller sequencing strategy leveraging probabilistic chiller load predictions to make more robust switching decisions. Another challenge for chiller sequencing control in real applications is the variation of chiller maximum cooling capacities and the measurement uncertainties. Therefore, the proposed strategy adapts the chiller sequencing thresholds based on historical data and real-time measurements. To validate the proposed strategy, an in-situ test was conducted in a typical educational building with a multiple-chiller system. The test results show that, compared to the original rule-based strategy, the average daily chiller switching number is reduced by 56.5 %, and the average daily energy savings is approximately 3945.1 kWh while maintaining thermal comfort. The coefficient of performance of the chiller plant is increased by an average of 4.2 %. The in-situ test demonstrates that the proposed strategy has great potential to be widely deployed in multiple-chiller systems for energy-efficient and reliable control.