Abstract: In the digital age, data centers have emerged as the cornerstone of modern information systems. However, the distribution of data centers in China is highly uneven, with a significant concentration on economically prosperous eastern regions such as the Beijing-Tianjin-Hebei area, the Yangtze River Delta, and the Guangdong-Hong Kong-Macao greater bay area. Despite their economic advantages, these regions face substantial challenges in hosting data centers. High land prices, power shortages, and unsuitable climates that necessitate energy-intensive cooling systems all contribute to the rise of operational costs. Whereas, western China offers substantial potential for data center development, characterized by low land costs, abundant power resources, and a more favorable climate. To address this imbalance and achieve coordinated regional development, China has introduced the “Eastern Data and Western Computing” strategy.
The strategic location and resource allocation of data centers are complex decision-making processes that involve multiple factors, including geographical location, climate, energy supply, and land cost. Existing research in this area is limited. Many studies focus only on partial factors and lack precision in optimizing data center locations. Additionally, there is a significant void in research that comprehensively considers carbon emissions and overall resource utilization. This study aims to address the multi-objective location and resource allocation problem of naturally cooled data centers under the “Eastern Data and Western Computing” strategy.
A multi-objective optimization model centered on natural cooling technology is proposed, which integrates economic, environmental, and social benefits. The model defines parameters for user demand nodes and resource endowment nodes, such as the number of existing cabinets, power generation structure, and energy utilization efficiency, and determines decision variables related to data center construction and node connections. By combining integer and linear programming techniques, the model solves complex problems in data center location and scale configuration and employs the Multi-Objective Genetic Algorithm (MOGA) for optimization. MOGA, based on the principle of natural selection, can simultaneously optimize multiple objectives, such as minimizing costs and maximizing carbon emission reduction.
Data for this study come from a wide range of reliable sources, including statistical yearbooks, national economic development bulletins, operator data, meteorological datasets, and information from local development and reform commissions. The improved K-means clustering method is used as an analytical technique. It standardizes data to eliminate dimensional differences, determines the optimal number of clusters through the elbow method and silhouette coefficient, and classifies data center construction areas according to user demands and resource endowments.
Theoretical analysis validates the effectiveness of the proposed model, which optimizes the layout of data centers to achieve the lowest cost and minimal environmental impact, thereby effectively reducing carbon emissions. The empirical results show that different natural cooling technologies significantly impact data center location, cost, and environmental performance. For example, liquid-cooling technology, despite its high initial investment, offers high energy efficiency, leading to lower long-term operating costs and better environmental performance. In practical applications, the integration of cluster analysis and multi-objective optimization generates a scientific layout plan. In the case studies of different cooling scenarios, the model provides guidance for rational resource allocation. This study offers practical guidance for data center operators to optimize layout and resource allocation, for policymakers to formulate relevant policies, and for equipment suppliers to develop suitable products. It serves as a crucial reference for data center planning and management under the “Eastern Data and Western Computing” strategy, facilitating the green transformation and energy-saving efforts of the data center industry.

Key words: Eastern Data and Western Computing, data center, cluster analysis, natural cooling

摘要： 在数字化时代,数据中心战略选址与资源配置对平衡经济发展和环境保护意义重大。本研究聚焦中国“东数西算”战略下的数据中心布局问题,提出以自然冷却技术为核心的多目标优化模型。该模型旨在实现数据中心在最低成本和最小环境影响下的最优选址与资源配置。研究首先综合分析地价、能源供应、气候条件等关键选址因素,并运用K-means聚类方法科学划分全国用户需求和资源禀赋。接着构建含投资成本、运营成本和碳减排量的多目标函数,采用多目标遗传算法求解,以获成本效益最大化和环境影响最小化的布局方案。实证分析表明,该模型能有效指导不同自然冷却技术条件下的数据中心选址决策,兼顾经济效益和环境可持续性。研究结果为“东数西算”战略下的数据中心布局提供理论和实践指导,助力行业绿色转型与节能减排,为相关方数据中心规划管理提供重要参考。

关键词: 东数西算, 数据中心, 聚类分析, 自然冷却