Comprehensive review and future prospects of multi-level fan control strategies in data centers for joint optimization of thermal management systems

With the rapid advancement of information and communication technology, particularly in artificial intelligence and cloud computing, the deployment of data centers has surged, making energy consumption in these facilities a critical issue. This review aims to address the joint control strategy for thermal management in data centers, focusing on control strategies for server fans. We examine various control methods, from model-free to model-based approaches, including physical, data-driven, and reinforcement learning models. Our analysis extends to multi-level fan control strategies, emphasizing the tight coupling between server operation and facility-level cooling systems. We also discuss control strategies encompassing data center IT equipment and external cooling sources, such as chillers and cooling towers. Our findings highlight the significant potential of multi-level fan control strategies for optimizing energy management. The review advocates for developing an integrated control ecosystem across all levels of data centers, optimizing energy management from chips and servers to air conditioning and cooling infrastructure. This integrated approach addresses data centers' critical energy consumption issue, promoting sustainable development. The novelty of this research lies in its holistic perspective on thermal management and the introduction of joint control strategies that coordinate cooling systems with IT operations. Our insights provide valuable guidance for future implementations, aiming to improve thermal management and computational optimization, thereby enhancing energy efficiency and overall data center performance. This work addresses the pressing need for more efficient energy management in rapidly growing data centers, which is crucial for sustainable development.