Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HyperLoad: A Cross-Modality Enhanced Large Language Model-Based Framework for Green Data Center Cooling Load Prediction

Haoyu Jiang, Boan Qu, Junjie Zhu, Fanjie Zeng, Xiaojie Lin, Wei Zhong

TL;DR

HyperLoad tackles cooling-load forecasting for green data centers under data scarcity and distribution shifts by fusing textual priors with time-series data through cross-modality alignment. The two-phase framework combines KARI for cross-modal representation alignment with ADPT and EGIA for domain-aware prefix tuning and cross-variable interaction modeling, powered by a pre-trained LLaMA-7B backbone. The approach achieves state-of-the-art performance in both data-sufficient and data-scarce settings on the DCData dataset, demonstrating strong generalization and rapid adaptation. The work provides a practical, scalable solution for sustainable data-center management and releases the DCData benchmark to spur further research.

Abstract

The rapid growth of artificial intelligence is exponentially escalating computational demand, inflating data center energy use and carbon emissions, and spurring rapid deployment of green data centers to relieve resource and environmental stress. Achieving sub-minute orchestration of renewables, storage, and loads, while minimizing PUE and lifecycle carbon intensity, hinges on accurate load forecasting. However, existing methods struggle to address small-sample scenarios caused by cold start, load distortion, multi-source data fragmentation, and distribution shifts in green data centers. We introduce HyperLoad, a cross-modality framework that exploits pre-trained large language models (LLMs) to overcome data scarcity. In the Cross-Modality Knowledge Alignment phase, textual priors and time-series data are mapped to a common latent space, maximizing the utility of prior knowledge. In the Multi-Scale Feature Modeling phase, domain-aligned priors are injected through adaptive prefix-tuning, enabling rapid scenario adaptation, while an Enhanced Global Interaction Attention mechanism captures cross-device temporal dependencies. The public DCData dataset is released for benchmarking. Under both data sufficient and data scarce settings, HyperLoad consistently surpasses state-of-the-art (SOTA) baselines, demonstrating its practicality for sustainable green data center management.

HyperLoad: A Cross-Modality Enhanced Large Language Model-Based Framework for Green Data Center Cooling Load Prediction

TL;DR

HyperLoad tackles cooling-load forecasting for green data centers under data scarcity and distribution shifts by fusing textual priors with time-series data through cross-modality alignment. The two-phase framework combines KARI for cross-modal representation alignment with ADPT and EGIA for domain-aware prefix tuning and cross-variable interaction modeling, powered by a pre-trained LLaMA-7B backbone. The approach achieves state-of-the-art performance in both data-sufficient and data-scarce settings on the DCData dataset, demonstrating strong generalization and rapid adaptation. The work provides a practical, scalable solution for sustainable data-center management and releases the DCData benchmark to spur further research.

Abstract

The rapid growth of artificial intelligence is exponentially escalating computational demand, inflating data center energy use and carbon emissions, and spurring rapid deployment of green data centers to relieve resource and environmental stress. Achieving sub-minute orchestration of renewables, storage, and loads, while minimizing PUE and lifecycle carbon intensity, hinges on accurate load forecasting. However, existing methods struggle to address small-sample scenarios caused by cold start, load distortion, multi-source data fragmentation, and distribution shifts in green data centers. We introduce HyperLoad, a cross-modality framework that exploits pre-trained large language models (LLMs) to overcome data scarcity. In the Cross-Modality Knowledge Alignment phase, textual priors and time-series data are mapped to a common latent space, maximizing the utility of prior knowledge. In the Multi-Scale Feature Modeling phase, domain-aligned priors are injected through adaptive prefix-tuning, enabling rapid scenario adaptation, while an Enhanced Global Interaction Attention mechanism captures cross-device temporal dependencies. The public DCData dataset is released for benchmarking. Under both data sufficient and data scarce settings, HyperLoad consistently surpasses state-of-the-art (SOTA) baselines, demonstrating its practicality for sustainable green data center management.
Paper Structure (22 sections, 12 equations, 5 figures, 3 tables)

This paper contains 22 sections, 12 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Definition
  4. Cross-Modality Input Construction
  5. Construction of Time-Series Modality Inputs.
  6. Construction of Text Modality Inputs.
  7. Cross-Modality Knowledge Alignment Phase
  8. Time-series modality feature encoding process.
  9. Text modality feature encoding process.
  10. KARI Strategy.
  11. Multi-Scale Feature Modeling Phase
  12. ADPT Strategy.
  13. EGIA Mechanism.
  14. Pre-trained LLM-based Backbone.
  15. Experiments
  16. ...and 7 more sections

Figures (5)

  • Figure 1: Schematic diagram of HyperLoad performing green data center cooling load prediction tasks.
  • Figure 2: Limitations of time-series forecasting algorithms.
  • Figure 3: (a) Data collection and training process. (b) Framework of HyperLoad. (c) Schematic illustration of the KARI loss. ( represents text modality, $\bigcirc$ represents time-series modality.) (d) Schematic illustration of the EGIA mechanism.
  • Figure 4: Prediction results for different horizons.
  • Figure 5: Model performance in the data scarce setting.