Energy consumption optimization method for industrial data centers based on multi-layer interactive learning and multi-level safety mechanisms

Optimizing energy consumption in data centers has emerged as a critical industry challenge, driven by the escalating demand for high-performance computing and the urgent need for environmental sustainability. Although various reinforcement learning-based optimization methods have been developed to manage these energy-intensive facilities, their effectiveness in real-world production environments remains uncertain. This uncertainty primarily stems from the significant discrepancy between idealized simulated data and complex, actual operational scenarios. Furthermore, many existing models provide insufficient coverage of industrial-grade safety requirements, failing to adequately address Service Level Agreement (SLA) constraints that are vital for equipment longevity and data integrity.

To address these limitations, this paper proposes a multi-layer interactive learning energy optimization framework built upon real-time operational data and rigorous industrial-grade SLA constraints. This approach aims to achieve the dual objectives of unwavering operational safety and maximized energy efficiency. By leveraging interactive learning, the system can adapt to dynamic environmental shifts that traditional models often overlook.

In comprehensive field tests conducted on an Indirect Evaporative Cooling System (IECS), the proposed method demonstrated exceptional performance. Compared to traditional Proportional-Integral-Derivative (PID) control, the method achieved a reduction in Power Usage Effectiveness (PUE) of 7.556%–8.651% and a significant decrease in the Cooling Load Factor (CLF) of 29.156%–33.867%. Crucially, the system maintained the safety deviation of the supply air temperature within a strict 0.656 threshold, ensuring stable hardware conditions. These experimental results indicate that our approach effectively enhances system energy efficiency while strictly adhering to the rigorous safety and reliability requirements essential for modern data center operations.