基于AI Agent的数据中心冷水机组健康监测方法

A Health Monitoring Method for Data Center Chiller Units Based on AI Agent

摘要

针对数据中心冷水机组健康监测中存在的诊断模型固化、跨场景迁移困难以及算法模型与运维认知割裂等问题，提出一种基于智能体 (AI Agent) 的健康监测方法。该方法借鉴人体健康诊断中初步筛查与综合会诊的协同逻辑，构建融合单变量阈值监测与多变量核主元分析 (kernel principal component analysis, KPCA) 的诊断框架，并基于 DeepSeek-V3 大语言模型实现监测策略自动规划、算法工具调用和可执行代码生成。基于 ASHRAE RP-1043 公开数据集的实验结果表明，该智能体能够准确复现冷却水不足、冷冻水不足、制冷剂泄漏等 7 类典型故障的诊断逻辑，并通过标准化函数调用机制提升诊断模型在不同机组间的迁移效率与部署灵活性。研究结果进一步表明，在单变量阈值监测与多变量 KPCA 工具协同作用下，该方法可有效降低多工况和不同机组健康监测的误报率；同时，在保证诊断准确性的前提下，智能体生成的监测代码较人工编写代码行数减少约 8.6%，为数据中心关键基础设施故障预测与健康管理提供了一种可扩展、易部署的自动化解决方案。

Abstract

To address challenges in health monitoring of data center chiller units, including inflexible diagnostic models, poor cross-scenario transferability, and the disconnect between algorithmic models and operation-and-maintenance knowledge, this paper proposes an AI Agent-based health monitoring method. Inspired by the coordinated strategy of preliminary screening and comprehensive diagnosis in clinical practice, the proposed method establishes a diagnostic framework integrating univariate threshold monitoring with multivariate kernel principal component analysis (KPCA). Based on the DeepSeek-V3 large language model, the AI Agent can automatically plan monitoring strategies, invoke algorithmic tools, and generate executable code. Experiments on the public ASHRAE RP-1043 dataset show that the proposed AI Agent can accurately reproduce the diagnostic logic for seven typical faults, including insufficient condenser water flow, insufficient chilled water flow, and refrigerant leakage. Through a standardized function-calling mechanism, the method improves the transferability and deployment flexibility of diagnostic models across different chiller units. The results further demonstrate that the integration of univariate threshold monitoring and multivariate KPCA can effectively reduce false alarms under varying operating conditions and across different units. Meanwhile, while maintaining diagnostic accuracy, the monitoring code generated by the AI Agent reduces the number of lines of code by approximately 8.6% compared with manual coding. This study provides a scalable and easy-to-deploy automated solution for fault prediction and health management of critical data center infrastructure.

关键词

工业大数据 / 大语言模型 / AI Agent / 设备健康监测 / 冷水机组 / 智能决策

Key words

industrial big data / large language models (LLMs) / AI agent / equipment health monitoring / chiller units / intelligent decision-making

参考文献

[1] 王智秀, 李建龙. 基于 AI 应用的建筑行业算力可达性提升研究[J]. 建筑经济, 2026, 47(1): 13-18.

[2] Wang Zhixiu, Li Jianlong. Research on computing power accessibility promotion of construction industry based on AI application[J]. Construction Economy, 2026, 47(1): 13-18.

[3] Lee Y T, Wen C Y, Shih Y C, et al. Numerical and experimental investigations on thermal management for data center with cold aisle containment configuration[J]. Applied Energy, 2022, 307: 118213.

[4] Pan X G, Liang K T, Jiao Y Y, et al. Interpretable chiller fault diagnosis based on physics-guided neural networks[J]. Journal of Building Engineering, 2024, 86: 108991.

[5] Bezyan Y, Nasiri F, Nik-Bakht M. Feature selection and fault detection under dynamic conditions of chiller systems[J]. Electronics, 2026, 15(1): 208.

[6] Deng Q, Liu Y F, Chen Z W, et al. A novel physical information-guided predictive maintenance method for chillers[J]. Applied Energy, 2025, 402: 126856.

[7] Ssembatya M, Claridge D E. Quantitative fault detection and diagnosis methods for vapour compression chillers: Exploring the potential for field-implementation[J]. Renewable and Sustainable Energy Reviews, 2024, 197: 114418.

[8] Wang J L, Xu C Q, Zhang J, et al. Big data analytics for intelligent manufacturing systems: a review[J]. Journal of Manufacturing Systems, 2022, 62: 738-752.

[9] Li G N, Chen L, Liu J Y, et al. Comparative study on deep transfer learning strategies for cross-system and cross-operation-condition building energy systems fault diagnosis[J]. Energy, 2023, 263: 125943.

[10] Han S Y, Shao H D, Huo Z Q, et al. End-to-end chiller fault diagnosis using fused attention mechanism and dynamic cross-entropy under imbalanced datasets[J]. Building and Environment, 2022, 212: 108821.

[11] 成卉, 刘勤明, 叶春明, 等 . 考虑外部冲击的风电机组机会维护策略研究[J]. 工业工程, 2023, 26(3): 159-167.

[12] Cheng Hui, Liu Qinming, Ye Chunming, et al. An opportunistic maintenance strategy for wind turbines under external environment shocks[J]. Industrial Engineering Journal, 2023, 26(3): 159-167.

[13] 高学金, 王璇, 姜渴鑫, 等 . 基于多尺度域生成网络的冷水机组故障诊断[J]. 中国机械工程, 2026, 37(3): 688-696.

[14] Gao Xuejin, Wang Xuan, Jiang Kexin, et al. Fault diagnosis of chillers based on multi-scale domain generative networks[J]. China Mechanical Engineering, 2026, 37(3): 688-696.

[15] 袁锐炜, 陈兆祥, 卫宇杰, 等 . 基于改进 Stacking 与 D-S 证据理论的两阶段多源信息融合轴承故障诊断[J]. 工业工程, 2025, 28(4): 56-67.

[16] Yuan Ruiwei, Chen Zhaoxiang, Wei Yujie, et al. Bearing fault diagnosis with two-stage multi-source information fusion based on improved stacking algorithm and D-S evidence theory[J]. Industrial Engineering Journal, 2025, 28(4): 56-67.

[17] Fan C, Lei Y T, Sun Y J, et al. Novel transformer-based self-supervised learning methods for improved HVAC fault diagnosis performance with limited labeled data[J]. Energy, 2023, 278: 127972.

[18] Yang W X, Zhang H Y, Lim J B, et al. A new chiller fault diagnosis method under the imbalanced data environment via combining an improved generative adversarial network with an enhanced deep extreme learning machine[J]. Engineering Applications of Artificial Intelligence, 2024, 137: 109218.

[19] Leng J W, Wang J J, Zhou L H, et al. A review of multi-modal deep learning towards agentic smart manufacturing[J]. Advanced Engineering Informatics, 2026, 70: 104117.

[20] Ren Y W, Liu Y Y, Ji T, et al. AI Agents and Agentic AI-navigating a plethora of concepts for future manufacturing[J]. Journal of Manufacturing Systems, 2025, 83: 126-133.

[21] Huang J, Teng Y, Liu Q H, et al. Leveraging large language models for efficient scheduling in Human-Robot collaborative flexible manufacturing systems[J]. npj Advanced Manufacturing, 2025, 2: 47.

[22] Wang L, Ma C, Feng X Y, et al. A survey on large language model based autonomous agents[J]. Frontiers of Computer Science, 2024, 18(6): 186345.

[23] Cao Y, Li S, Liu Y, et al. A comprehensive survey of AI-generated content (AIGC): a history of generative AI from GAN to ChatGPT[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(7): 4432-4456.

[24] 姚锡凡, 胡坤, 李佳蔚. 面向工业 5.0 的多智能体云边端协同智慧制造体系架构与作业调度[J/OL]. 机械工程学报, 2026. https://link.cnki.net/urlid/11.2187.TH.20251223.1147.044.

[25] Yao Xifan, Hu Kun, Li Jiawei. Wisdom manufacturing architecture and scheduling for Industry 5.0 based on cloud-edge-end collaboration of AI agents[J/OL]. Journal of Mechanical Engineering, 2026. https://link.cnki.net/urlid/11.2187.TH.20251223.1147.044.

[26] 张晴雪, 张欣亮, 武凌杰. 多模态 AI 智能体在工业缺陷检测中的应用[J]. 机器人技术与应用, 2025(1): 37-40.

[27] 戴希文, 王军武. 公路工程造价数字化管理成熟度评价研究[J]. 建筑经济, 2026, 47(5): 39-47.

[28] Dai Xiwen, Wangjunwu. Research on evaluation of digital management maturity of project cost for highway engineering[J]. Construction Economy, 2026, 47(5): 39-47.

[29] Comstock M C. Development of analysis tools for the evaluation of fault detection and diagnostics in chillers[D]. West Lafayette: Purdue University, 1999.

基金资助

国家自然科学基金项目(52305545)