Edge Computing-Enabled Energy Efficiency Prediction in Supercomputing Centers

Strategic Insights -

Research at Xi'an University of Posts and Telecommunications has proposed an innovative architecture for predicting energy consumption in immersion cooling systems for supercomputing centers. The edge computing-enabled energy efficiency prediction system combines time-series generative adversarial networks (TimeGAN) for data augmentation with neural basis expansion analysis for time series (N-BEATS) for precise predictions. Experimental results demonstrate the effectiveness of the proposed architecture, outperforming traditional models in key metrics such as root mean square error (RMSE), mean squared error (MSE), and R-squared (R2).

Key Takeaways:

  • The proposed architecture combines TimeGAN and N-BEATS to address the challenges in predicting energy consumption in immersion cooling systems for supercomputing centers.
  • TimeGAN enhances the training dataset by generating high-quality synthetic time-series data, mitigating issues of data sparsity and imbalance.
  • N-BEATS ensures precise predictions by capturing both global trends and local variations in energy usage.
  • The experimental results demonstrate superior performance of the proposed architecture compared to traditional models, with RMSE reduced by more than 8%, MSE decreased by over 18%, and R2 reaching 97.31%.
  • The research highlights the potential of integrating generative and predictive models to optimize energy efficiency in liquid cooling systems.
  • The proposed architecture is designed to provide a robust solution for accurate energy consumption prediction in supercomputing centers.
  • The authors, Yichun Cao and Shuaiyin Ma, propose that the edge computing-enabled energy efficiency prediction system can offer valuable insights for sustainable supercomputing operations.

Statistics:

  • RMSE reduced by more than 8% using the proposed architecture.
  • MSE decreased by over 18% using the proposed architecture.
  • R2 reached 97.31% using the proposed architecture.
  • The proposed architecture outperformed baseline models such as long short-term memory (LSTM) and gated recurrent unit (GRU) neural networks.
  • The research proposes a novel approach to address the challenges in predicting energy consumption in immersion cooling systems for supercomputing centers.

Sources:

  • Cao, Y., & Ma, S. (2025). Edge computing-enabled energy efficiency prediction of immersion cooling system for supercomputing centers. Cleaner Engineering and Technology, 2025, 27(), 101014. doi: 10.1016/j.clet.2025.101014
  • NewsRx (2025, July 11). Researchers at Xi'an University of Posts and Telecommunications Report Research in Cleaner Engineering and Technology (Edge computing-enabled energy efficiency prediction of immersion cooling system for supercomputing centers). Ecology, Environment & Conservation, p 1963.