[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84528-en":3,"doc-seo-84528-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84528,962075006959,"Anda","https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297",8,"Research & Report","Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization","Thermal management in AI data centers faces growing stress from bursty workloads and uncertain heat generation, where existing methods either enforce overly conservative fixed bounds that reduce grid responsiveness or depend on forecasts that fail under workload uncertainty and distribution shifts. A Contextual Distributionally Robust Optimization (CDRO) framework dynamically adapts Wasserstein ambiguity sets using real-time AI and grid context, shrinking uncertainty during stable regimes. The resulting control formulation yields tractable ADMM solutions and enforces DR-CVaR thermal safety, achieving near-zero violations in EnergyPlus co-simulations while reducing robustness cost by about 13.7 percentage points versus min-max MPC.","Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization  \nJiachen Shen, Student Member, IEEE, Jian Shi, Senior Member, IEEE, Yijie Yang, Member, IEEE, Chenye Wu, Senior Member, IEEE, Dan Wang, Senior Member, IEEE, Ju Bin Song, Member, IEEE, and Zhu Han, Fellow, IEEE  \narXiv :2607 .00099v1 [ ee ss . SY] 30 Jun 2026  \nAbstract—Thermal management in AI data centers is increasingly challenged by bursty workloads and uncertain heat generation. To prevent thermal violations, existing cooling strategies either enforce conservative, rigid bounds that severely limit grid responsiveness, or rely on forecast-driven controllers that perform poorly under AI workload uncertainty and distribution shifts. To overcome the above challenges, this paper proposes a Contextual Distributionally Robust Optimization (CDRO) framework for grid-interactive cooling control. Unlike standard DRO with fixed ambiguity sets, the proposed approach dynamically adapts the Wasserstein radius using real-time AI and grid context. This safely shrinks uncertainty bounds during stable regimes, unlocking deep demand-side flexibility. Theoretically, we formulate the control as an infinite-dimensional inf–sup problem, derive an exact tractable reformulation for the Wasserstein worstcase expected-cost term, and then derive a tractable conservative deterministic counterpart for the Distributionally Robust Conditional Value at Risk (DR-CVaR) thermal safety constraint. Solved via a scalable nested Alternating Direction Method of Multipliers (ADMM) algorithm, the CDRO controller achieves near-zero thermal violations under extreme workload spikes in highfidelity EnergyPlus co-simulations. Simultaneously, it reduces the operational cost premium of robustness by approximately 13.7 percentage points relative to standard Min–Max Model Predictive Control (MPC).  \nIndex Terms—AI Data Centers, Contextual Distributionally Robust Optimization, AI Workload Uncertainty, Thermal Management.  \nI. INTRODUCTION  \nA. Motivation and Challenges  \nThe rapid growth of Generative Artificial Intelligence (GenAI) has significantly changed the operational landscape of data centers [1] . This elevates thermal management toa critical infrastructure bottleneck alongside computational capacity. Modern AI accelerators, like NVIDIA’s H100 GPUs, exhibit a Thermal Design Power (TDP) exceeding 700W per chip. This pushes rack power densities beyond 50kW, which is ten times that of traditional CPU-based servers [2] . This rapid rise in heat flux requires powerful cooling capacities. As a result, the cooling system becomes the largest auxiliary energy consumer (accounting for 30%–40% of total facility energy) and the most critical subsystem for hardware reliability. Also, the central operational challenge has shifted from general energy scheduling to the precise control of cooling loads: maintaining thermal safety under extreme heat densities while attempting to adjust these massive cooling loads to respond to volatile power grid signals (e.g., locational marginal prices (LMP) or carbon intensity) .  \nMoreover, the inflexible operational strategies of legacy cooling systems are increasingly in conflict with the decarbonization goals of the energy sector. As the grid integrates higher shares of variable renewable energy, electricity prices  \nand carbon intensity become highly volatile [3], [4] . Static cooling strategies (e.g., constant setpoint or ProportionalIntegral-Derivative (PID) control), which consume power regardless of these external signals, not only incur excessive operational costs but also worsen grid congestion during peak hours [5] . There is an urgent need to utilize the unused flexibility within the cooling infrastructure. This allows it to act as a responsive demand-side resource that can absorb renewable generation when it is abundant and shed load when the grid is stressed.  \nTo address this, the concept of the grid-interactive datacenter has ","cbCaiepewrJ8Rwmx","https://ap.wps.com/l/cbCaiepewrJ8Rwmx","pdf",14446993,1,10,"English","en",105,"# Introduction\n## Motivation and Challenges\n## Grid-Interactive Datacenter Concept\n## Control Challenges: Timescale Mismatch and Over-Cooling Trap","[{\"question\":\"Why do thermal violations become more likely in AI data centers?\",\"answer\":\"Bursty AI workloads can cause rapid heat generation and localized thermal shocks, while heat rejection dynamics are much slower due to thermo-fluid components. This mismatch makes it difficult for traditional controllers to ramp cooling in time, leading to thermal safety risks.\"},{\"question\":\"What limitation do existing cooling strategies have with respect to grid interaction?\",\"answer\":\"Many approaches rely on conservative over-cooling or forecast-driven controllers. Over-cooling keeps cooling power unnecessarily high and restricts participation in demand response, while forecast-based methods degrade under uncertainty and distribution shifts.\"},{\"question\":\"How does the CDRO framework improve robustness and flexibility for cooling control?\",\"answer\":\"CDRO dynamically adjusts the Wasserstein radius based on real-time AI and grid context, shrinking ambiguity when conditions are stable. This enables safe thermal constraints under uncertainty while improving grid responsiveness and lowering the operational cost premium.\"}]",1784196423,25,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"grid-interactive-thermal-management-of-ai-data-centers-via-contextual-distributionally-robust-optimization","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/grid-interactive-thermal-management-of-ai-data-centers-via-contextual-distributionally-robust-optimization/84528/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do thermal violations become more likely in AI data centers?","Question",{"text":75,"@type":76},"Bursty AI workloads can cause rapid heat generation and localized thermal shocks, while heat rejection dynamics are much slower due to thermo-fluid components. This mismatch makes it difficult for traditional controllers to ramp cooling in time, leading to thermal safety risks.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What limitation do existing cooling strategies have with respect to grid interaction?",{"text":80,"@type":76},"Many approaches rely on conservative over-cooling or forecast-driven controllers. Over-cooling keeps cooling power unnecessarily high and restricts participation in demand response, while forecast-based methods degrade under uncertainty and distribution shifts.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the CDRO framework improve robustness and flexibility for cooling control?",{"text":84,"@type":76},"CDRO dynamically adjusts the Wasserstein radius based on real-time AI and grid context, shrinking ambiguity when conditions are stable. This enables safe thermal constraints under uncertainty while improving grid responsiveness and lowering the operational cost premium.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]