SUMMARY

Chillers with centrifugal compressors provide chilled fluid for air conditioning and process cooling in commercial buildings, schools, hospitals, data centers, and other applications. Water-cooled chillers are developed with the objective of lowering energy consumption and minimizing carbon footprint; however, unit-to-unit variations, widely varying operating conditions across installations, generic control schemes, and inaccurate sensors prevent the realization of these operational goals. An improved understanding of the underlying chiller phenomena can enable performance improvements while avoiding expensive system enhancements. This study addresses the performance degradation of water-cooled centrifugal chillers from design values by establishing the needed diagnostic tools using characteristics and data from commercially available systems to guide maintenance routines, fault detection, and control applications.

A review of centrifugal chiller modeling literature is conducted, and the methodology and deficiencies of available models are discussed. A steady-state model that relies only on the minimal measurements (i.e., coupling fluid measurements) typically available from field installation is developed to predict the performance of a commercial 1758 kW centrifugal chiller within 3% of experimental values. An additional steady model is also developed to assess system and centrifugal compressor performance prediction accuracy using performance mapping methods and common commercial sensor data (i.e., refrigerant measurements and compressor speed/power). Chiller transients between AHRI operating conditions are dynamically modeled using moving-boundary and state-space techniques for the key heat exchangers. These models are validated using field data. Insights from these performance tools will guide future machine-learning-based models that facilitate real-time chiller performance prediction, fault detection, and control for commercial chillers.