Design of an Immersion Cooling Data Centers System Based on Deep Hashing Networks

The rapid growth of cloud computing, HPC, and AI has increased data center computing density, making traditional air-cooling inadequate for high-power servers. Immersion cooling, with better thermal conductivity and energy efficiency, is a promising alternative. However, current systems struggle with real-time data management and dynamic heat dissipation control, limiting adaptability to server thermal loads. This study introduces a Dual-Loss Optimized Deep Hashing Network for Immersion Cooling (DHN-IC), which efficiently processes high-dimensional data and optimizes heat dissipation by addressing hash discretization and semantic similarity loss. The DHN module enables real-time coolant flow adjustments, reducing Power Usage Effectiveness (PUE) and enhancing energy efficiency. Experimental results show DHN-IC improves data retrieval accuracy, reduces cooling energy consumption, and stabilizes server temperatures compared to traditional methods. This research provides insights into intelligent liquid cooling optimization, supporting energy-efficient, sustainable data center development.