Improved model and optimization for the energy performance of chiller system with diverse component staging

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)


Accurate data-driven models are required to optimize the energy performance of chiller systems with individual configurations and control strategies. This study develops stepwise regression models at successive time steps to estimate accurately the coefficient of performance (COP) of a system with diverse component staging. Comprehensive trend logging data were collected at 15-min intervals year-round for a system with five chillers of two capacities. Data at two successive time points were compiled into 19 input variables relating to the components’ operating statuses, load sharing variation, temperature-related and flow-related variables and weather conditions. Stepwise regression with the Akaike information criterion was performed to select the predictor variables and time points giving the highest achievable coefficient of determination R2. The developed system COP models with different numbers of operating chillers and multiple time points has an increased R2 of 0.75–0.94 compared with 0.37–0.89 for a conventional model with the chiller part load ratio (PLR) only. The model coefficients help prioritize variables to be optimized. The L-BFGS-B optimization algorithm with operating constraints was applied to predict the maximum system COP for the system operating for the cooling load profile of an educational building. A reduction of 9.14–16.67% was estimated in the annual electricity consumption when compared with the existing operation. The optimal control involves implementing pair-up operation of chillers, pumps and cooling towers, and resetting the temperature of water leaving the condensers at specific ranges based on the number of operating chillers.

Original languageEnglish
Article number119376
Publication statusPublished - 15 Feb 2021


  • Coefficient of performance
  • L-BFGS-B optimization
  • Part load ratio
  • Stepwise regression
  • Water-cooled chiller


Dive into the research topics of 'Improved model and optimization for the energy performance of chiller system with diverse component staging'. Together they form a unique fingerprint.

Cite this