基于图卷积神经网络的数据中心冷却模型预测控制优化方法研究

扫码查看

原文链接

万方数据
维普

中文摘要：在大型数据中心末端冷却系统中，由于系统复杂、真实系统运行历史数据有限，导致节能优化困难的问题。本文提出了一种基于图卷积神经网络(Graph Convolutional Neural Networks，GCNNs)的数据中心冷却模型预测控制(Model Predictive Control，MPC)优化方法。通过对末端冷却系统中房间级组件(如冷却单元、传感器)之间的空间关系及相互作用进行抽象，建立与之对应的图卷积神经网络模型和末端控制算法。结果表明:该方法在大型数据中心的场景下，与基于比例-积分-微分(Proportional-Integral-Derivative，PID)的传统控制方法相比，末端冷却系统的控制节能效率提升了5％～15％。

外文标题：Research on Predictive Control Optimization Method for Data Center Cooling Model Based on Graph Convolutional Neural Network

外文摘要：In terminal cooling systems of large data centers,due to the complexity of the system and limited historical operational data of actual system operation,energy-saving optimization is difficult. A data center cooling model predictive control (MPC) optimization method based on graph convolutional neural networks (GCNNs) is proposed in this paper. By abstracting the spatial relationships and interactions between room-level components (such as cooling units and sensors) in the terminal cooling system,a corresponding GCNN model and terminal control algorithms are established. The results show that this method improves the energy-saving efficiency of terminal cooling system control by 5％-15％ compared to traditional control methods based on proportional-integral-derivative (PID).

外文关键词：

Data centersCooling systemEnergy conservationModel predictive controlGraph convolutional neural network

作者：

李慧

展开 >

作者单位：

万国数据服务有限公司,上海 201114

关键词：

数据中心冷却系统节能模型预测控制图卷积神经网络

出版年：

2024

DOI：