---
格式版本: 2
标题: "LLM-Powered Predictive Decision-Making for Sustainable Data Center Operations"
原文链接: "https://arxiv.org/abs/2608.18503"
发布日期: "2026-08-19"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 19 Aug 2026 03:50:32 UTC (6,091 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-20T16:05:13+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-20T15:56:53+08:00"
入库时间: "2026-08-20T08:05:14.043Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=AI&searchtype=all"
匹配关键词:
  - "AI"
  - "GPU"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 49
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文内容为LLM驱动的数据中心调度优化，聚焦能耗与排队时延，未涉及超节点、AI Rack、机柜级硬件架构、供电散热或高速互连等核心主题，与项目关注范围无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T16:11:25+08:00"
AI主题相关性: 4
AI来源权威性: 12
AI新颖性: 15
AI技术细节: 6
AI商业部署信号: 4
AI完整性: 8
AI摘要: "该论文提出一个基于大语言模型的预测性调度系统，通过从源代码预测执行时间和能耗等指标，并实时分配GPU资源，以提升数据中心运营效率并降低环境影响。在与实际数据中心的合作中，该系统实现了32%的能耗降低和30%的排队等待时间减少。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:18:56.869Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.18503"
---

## Computer Science > Machine Learning

## Title:LLM-Powered Predictive Decision-Making for Sustainable Data Center Operations

Authors:[Hanzhao Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+H), [Jingxuan Wu](https://arxiv.org/search/cs?searchtype=author&query=Wu,+J), [Yumeng Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+Y), [Yu Pan](https://arxiv.org/search/cs?searchtype=author&query=Pan,+Y), [Guanting Chen](https://arxiv.org/search/cs?searchtype=author&query=Chen,+G)

[View PDF](https://arxiv.org/pdf/2608.18503) [HTML (experimental)](https://arxiv.org/html/2608.18503v1)

> Abstract:The growing demand for AI-driven workloads, particularly from Large Language Models (LLMs), has raised concerns about the significant energy and resource consumption in data centers. This work introduces a novel LLM-based predictive scheduling system designed to enhance operational efficiency while reducing the environmental impact of data centers. Our system utilizes an LLM to predict key metrics such as execution time and energy consumption from source code, and it has the potential to extend to other sustainability-focused metrics like water usage for cooling and carbon emissions, provided the data center can track such data. The predictive model is followed by a real-time scheduling algorithm that allocates GPU resources, aiming to improve sustainability by optimizing both energy consumption and queuing delays. With fast inference times, the ability to generalize across diverse task types, and minimal data requirements for training, our approach offers a practical solution for data center scheduling. This framework demonstrates strong potential for advancing sustainability objectives in AI-driven infrastructure. Through our collaboration with a data center, we achieved a 32% reduction in energy consumption and a 30% decrease in waiting time.

| Subjects: | Machine Learning (cs.LG) |
| --- | --- |
| Cite as: | [arXiv:2608.18503](https://arxiv.org/abs/2608.18503) \[cs.LG\] |
|  | (or [arXiv:2608.18503v1](https://arxiv.org/abs/2608.18503v1) \[cs.LG\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18503](https://doi.org/10.48550/arXiv.2608.18503) |

## Submission history

From: Jingxuan Wu \[[view email](https://arxiv.org/show-email/58049223/2608.18503)\]  
**\[v1\]** Wed, 19 Aug 2026 03:50:32 UTC (6,091 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18503) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))