| 引用本文: | 俞露杰,赵博文,朱介北,李炳森,周欢,贾宏杰.面向需求响应的数据中心热力学动态系统建模与关键参数辨识方法[J].电力自动化设备,2025,45(4):186-193 |
| YU Lujie,ZHAO Bowen,ZHU Jiebei,LI Bingsen,ZHOU Huan,JIA Hongjie.Modeling and key parameter identification method of data center thermal dynamic system towards demand response[J].Electric Power Automation Equipment,2025,45(4):186-193 |
|
| 摘要: |
| 数据中心热力学动态模型能够为负载管理和制冷系统优化运行提供依据。建立了引入自适应噪声的数据中心热力学动态(ADTD)模型，在传统热力学模型基础上引入过程噪声和测量噪声以补偿数学模型和物理系统之间的偏差；建立自适应噪声更新算法以及数据中心热状态最优估计算法；辨识模型热容、热阻等关键参数。搭建了私有云数据中心实验平台，基于实测运行数据，建立ADTD模型并验证其模拟数据中心室内温度变化的性能。结果表明，与传统建模和参数辨识方法相比，所建立的模型可获得更为精确的参数辨识结果，在数据中心不同运行场景下可准确模拟室内温度变化，并具有良好的适用性。 |
| 关键词: 数据中心 热力学模型 自适应噪声 参数辨识 状态估计 需求响应 |
| DOI：10.16081/j.epae.202412021 |
| 分类号:TM716 |
| 基金项目:国家重点研发计划资助项目(2018YFA0702200) |
|
| Modeling and key parameter identification method of data center thermal dynamic system towards demand response |
|
YU Lujie1, ZHAO Bowen2, ZHU Jiebei1, LI Bingsen1,3, ZHOU Huan1, JIA Hongjie1
|
|
1.School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China;2.Operation and Maintenance Center of State Grid Chengdu Power Supply Company, Chengdu 610041, China;3.State Grid Information & Telecommunication Group Co.,Ltd.,Beijing 102211, China
|
| Abstract: |
| The thermal dynamic model of data center can provide a basis for load management and cooling system optimal operation. An adaptive noise-incorporated data center thermal dynamic(ADTD) model is proposed. On the basis of conventional thermal dynamic model, process noise and measurement noise are introduced to compensate the deviation between mathematical model and physical system. An adaptive noise update algorithm and an optimal thermal state estimation algorithm for the data center are developed. Key parameters such as heat capacity and heat resistance are identified. A private cloud data center experimental platform is built. Based on the measured operation data, the ADTD model is established and its performance of simulating indoor temperature variations in the data center is verified. The results indicate that compared with conventional modeling and parameter identification methods, the established model can obtain more accurate parameter identification results, and can accurately simulate indoor temperature variations in different operating scenarios of data center, and has good applicability. |
| Key words: data center thermal dynamic model adaptive noise parameter identification state estimation demand response |
