Climate-adaptive control of a hybrid heat pump-desiccant wheel cooling system

Conventional air conditioning systems are energy-intensive and have limited ability to independently control humidity, particularly in hot-arid climates with high summer latent loads. This study proposes a climate-adaptive control framework for a hybrid heat pump-desiccant wheel system serving an office building under extreme arid conditions. The system integrates an energy recovery ventilator, a silica-gel desiccant wheel, and a vapor-compression heat pump. An artificial neural network surrogate model of the desiccant wheel, with R² = 0.998 and RMSE of 0.36 °C, and 0.16 g.kg⁻¹ for outlet temperature and humidity ratio, respectively, is coupled with a physics-based heat pump model within a multi-objective optimization framework. The controller adaptively selects among three operational modes while optimizing process-air temperature entering the desiccant wheel, wheel speed, face velocity, regeneration temperature, and bypass fraction. The framework was evaluated over 24 summer design and representative days, including ≥95th-percentile temperature and humidity conditions. The controller maintained humidity tracking errors within ±0.1 g. kg⁻¹ across all operating hours, with most hours within ±0.05 g. kg⁻¹. Compared with a conventional direct-expansion baseline, the hybrid system achieved energy savings of 45% under low-to-moderate humidity conditions of 12–16 g.kg⁻¹, while savings declined to 0–15% under extreme latent loads above 22 g.kg⁻¹, with an average COP_Sys of 2.07 ± 0.18 and COP_Th of 3.07 ± 0.46. Sensitivity analysis demonstrated that ambient humidity explained 91% of the variance in energy savings, and that adaptive control variables increased COP_Sys performance by ΔR² =0.60. These results indicate that machine-learning-assisted multi-objective control can significantly improve the operational efficiency and humidity control performance of hybrid cooling systems under challenging climatic conditions.