Model development and application of a comprehensive performance evaluation framework for immersion cooling data centers

Existing studies on immersion cooling data centers have advanced heat transfer analysis, system modeling, and operational optimization, but multi-cabinet thermal heterogeneity, coordinated system operation, and thermal safety assessment remain insufficiently integrated. This study developed a coupled lumped-parameter model for a single-phase immersion cooling system comprising six cabinets, a coolant distribution unit (CDU), and a cooling tower. Across different loads and allocations, cabinet and CDU model predictions agreed with measurements within approximately ±10% and 5%, respectively; measurement uncertainty and local sensitivity were also assessed. The validated model was applied to coordinated operational optimization, cabinet-level thermal safety control under load-imbalanced conditions, and annual thermal safety assessment. The minimum cooling-system power consumption was 2.134 kW at a CDU coolant outlet temperature of 29 °C, with coolant and cooling-water flow rates of 14.2 and 4.8 m 3 /h, respectively. Using 2025 hourly meteorological data for Shanghai, the annual unmet hours (AUH) totaled 255 h (2.91%), and a minimum safe coolant flow rate of 6.2 m 3 /h per cabinet was identified. The framework provides a common system-level basis for differentiated cabinet loads and cabinet-specific thermal states, coordinated regulation of the coolant and cooling-water loops, cabinet-level thermal safety control, and annual thermal safety assessment, thereby supporting energy-efficient operation and thermal risk management in multi-cabinet immersion cooling systems.