Sign in Subscribe
Data centers

Advanced Cooling System Fault Diagnosis Method for Data Centers

Strategic Insights

— 1 min read

Researchers from North China Electric Power University have proposed a new method for early adaptive multivariate state estimation fault diagnosis in multi-mode cooling systems used in data centers. The method combines deep agent reinforcement learning with traditional state estimation techniques to improve fault diagnosis accuracy and reduce energy consumption. According to the study, the proposed method can advance the fault warning time by more than 9 hours and improve the average fault diagnosis rate by up to 10.38%. The research has been funded by the National Natural Science Foundation of China and has been peer-reviewed for publication in the journal Energy.

Key Takeaways:

  • The proposed method uses deep agent reinforcement learning to improve fault diagnosis performance in multi-mode cooling systems.
  • The method can advance the fault warning time by more than 9 hours and improve the average fault diagnosis rate by up to 10.38%.
  • The study found that the proposed method can reduce the fault alarm rate by 8.05% compared to traditional methods.
  • The research was funded by the National Natural Science Foundation of China and the Fundamental Research Funds for the Central Universities.
  • The proposed method was evaluated using three state-of-the-art comparison methods and five typical cooling system faults.
  • The study concluded that the proposed method can lead to significant energy savings and improved fault diagnosis accuracy in data centers.

Statistics:

  • 39.6% energy consumption reduction achieved by the proposed method.
  • 9 hours advance in fault warning time by the proposed method.
  • 10.38% improvement in average fault diagnosis rate by the proposed method.
  • 8.05% reduction in fault alarm rate by the proposed method.
  • 3 state-of-the-art comparison methods used to evaluate the proposed method.
  • 5 typical cooling system faults simulated to evaluate the proposed method.

Sources:

  • A Deep Agent Reinforcement Learning-based Early Adaptive Multivariate State Estimation Fault Diagnosis Method for Multi-mode Cooling System In Data Center. Energy, 2025;334. (Elsevier - www.elsevier.com; Energy - www.journals.elsevier.com/energy/)
  • NewsRx. New Data from North China Electric Power University Illuminate Findings in Data Centers (A Deep Agent Reinforcement Learning-based Early Adaptive Multivariate State Estimation Fault Diagnosis Method for Multi-mode Cooling System In Data Center). Information Technology Newsweekly. October 21, 2025; p 441.

Read more