Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adaptive Federated Learning to Optimize Integrated Flows in Cyber-Physical Data Centers

Junhong Liu, Lanxin Du, Yujia Li, Rong-Peng Liu, Yunfeng Li, Fei Teng, Francis Yunhe Hou

TL;DR

The work tackles energy management for hyperscale data centers by jointly optimizing electricity, heat, and data flows under privacy constraints. It introduces an adaptive federated learning-to-optimization framework with a privacy-preserving learning stage and a verifiable data-sharing scheme using double aggregation to ensure data integrity. The approach yields convergence-guaranteed training, near-optimal distributed optimization, and significant computational efficiency compared with centralized MILP. The results demonstrate robust performance under heterogeneity and privacy requirements, enabling scalable, privacy-preserving operation of large-scale data-center networks.

Abstract

Data centers play an increasingly critical role in societal digitalization, yet their rapidly growing energy demand poses significant challenges for sustainable operation. To enhance the energy efficiency of geographically distributed data centers, this paper formulates a multi-period optimization model that captures the interdependence of electricity, heat, and data flows. The optimization of such integrated multi-domain flows inherently involves mixed-integer formulations and the access to proprietary or sensitive datasets, which correspondingly exacerbate computational complexity and raise data-privacy concerns. To address these challenges, an adaptive federated learning-to-optimization approach is proposed, accounting for the heterogeneity of datasets across distributed data centers. To safeguard privacy, cryptography techniques are leveraged in both the learning and optimization processes. A model acceptance criterion with convergence guarantee is developed to improve learning performance and filter out potentially contaminated data, while a verifiable double aggregation mechanism is further proposed to simultaneously ensure privacy and integrity of shared data during optimization. Theoretical analysis and numerical simulations demonstrate that the proposed approach preserves the privacy and integrity of shared data, achieves near-optimal performance, and exhibits high computational efficiency, making it suitable for large-scale data center optimization under privacy constraints.

Adaptive Federated Learning to Optimize Integrated Flows in Cyber-Physical Data Centers

TL;DR

The work tackles energy management for hyperscale data centers by jointly optimizing electricity, heat, and data flows under privacy constraints. It introduces an adaptive federated learning-to-optimization framework with a privacy-preserving learning stage and a verifiable data-sharing scheme using double aggregation to ensure data integrity. The approach yields convergence-guaranteed training, near-optimal distributed optimization, and significant computational efficiency compared with centralized MILP. The results demonstrate robust performance under heterogeneity and privacy requirements, enabling scalable, privacy-preserving operation of large-scale data-center networks.

Abstract

Data centers play an increasingly critical role in societal digitalization, yet their rapidly growing energy demand poses significant challenges for sustainable operation. To enhance the energy efficiency of geographically distributed data centers, this paper formulates a multi-period optimization model that captures the interdependence of electricity, heat, and data flows. The optimization of such integrated multi-domain flows inherently involves mixed-integer formulations and the access to proprietary or sensitive datasets, which correspondingly exacerbate computational complexity and raise data-privacy concerns. To address these challenges, an adaptive federated learning-to-optimization approach is proposed, accounting for the heterogeneity of datasets across distributed data centers. To safeguard privacy, cryptography techniques are leveraged in both the learning and optimization processes. A model acceptance criterion with convergence guarantee is developed to improve learning performance and filter out potentially contaminated data, while a verifiable double aggregation mechanism is further proposed to simultaneously ensure privacy and integrity of shared data during optimization. Theoretical analysis and numerical simulations demonstrate that the proposed approach preserves the privacy and integrity of shared data, achieves near-optimal performance, and exhibits high computational efficiency, making it suitable for large-scale data center optimization under privacy constraints.

Paper Structure

This paper contains 15 sections, 2 theorems, 34 equations, 8 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Formulations of the Energy Management for Hyper-scale Data Centers
  3. Centralized Optimization Formulation
  4. Personalized Federated Learning-to-Optimization Approach
  5. Adaptive Federated Learning Approach
  6. Federated Learning-to-Optimization Approach
  7. Verifiable Private Data-Sharing Scheme
  8. Privacy and Security Analysis
  9. Numerical Simulations
  10. Simulation Setup
  11. Accuracy and Optimality
  12. Computational Efficiency
  13. Conclusions
  14. Preliminaries of the Adaptive Federated Learning
  15. Convergence for the Adaptive Federated Learning

Key Result

Lemma 1

For $L$-smooth $f$ and $d_r=\bar{\theta}_{r+1}-\bar{\theta}_r$, it can be obtained:

Figures (8)

  • Figure 1: Scheme of the data center energy management.
  • Figure 2: Personalized adaptive federated learning (offline phase).
  • Figure 3: Personalized adaptive federated learning (online phase).
  • Figure 4: Verifiable data sharing via double aggregation mechanism.
  • Figure 5: Training results for the adaptive federated learning.
  • ...and 3 more figures

Theorems & Definitions (2)

  • Lemma 1: Smoothness descent boyd2004convex
  • Lemma 2: Exponential Moving Average energy recursion yu2019linear