为何弧闪安全分析对AI数据中心迈向800V直流电源架构至关重要

Why arc flash safety analysis is critical to the industry’s shift to 800 VDC in AI data centers

机架密度的持续攀升对数据中心基础设施带来了更深层次的影响，推动业界再次重新评估既有规范且制定新标准。

The ripple effect from ever expanding rack densities is moving into a new layer of data centers, forcing the industry to once again reevaluate historic norms and create new standards.

AI工作负载增速空前，正推动数据中心基础设施向单机架兆瓦级功耗的方向演进，以支撑加速计算的需求。传统的交流配电系统与低压交流架构已无法满足此类极高密度且高度集中的电力负荷。因此，业界正围绕800V直流电源架构形成共识，将其视为下一代AI数据中心兼具实用与高效的发展路径之一。

The unprecedented pace with which AI workloads are growing is quickly pushing infrastructure toward megawatt-scale racks to support accelerated computing. Traditional AC distribution systems and lower-voltage architectures cannot keep up with these extreme, concentrated power demands. As a result, the industry is actively aligning around 800 VDC architectures as a highly practical and effective path forward for next-generation AI data centers.

随着行业向800V直流架构迈进，安全工程与弧闪风险防控也必须同步提速。更高压系统虽能大幅提升token生成效率，但部署的前提是要对电气安全建立起清晰、全面的认知。

As the industry races toward 800 VDC architectures, safety engineering and arc flash risk mitigation must evolve just as quickly. The transition to higher voltage systems can unlock enormous token production efficiency gains, but deployment will require a clear, comprehensive understanding of electrical safety. 

直流电压等级提升，加之有源电力电子变流器的广泛应用，使得电气故障特性更加复杂。这就要求行业在推进该类系统转型时，必须对故障电流特性、保护配合方案、设备设计以及全体运维人员的安全操作规范，形成更为深入、严谨的专业认知。

With higher DC voltage combined with the use of active power electronic converter, electrical fault behavior inherently becomes far more complex. Transitioning to these systems requires the industry to develop a much deeper, more rigorous understanding of fault current behavior, protection coordination, equipment design, and safe work practices for all operating personnel. 

任何形式的配电系统都可能发生弧闪事件，等离子能量和强压力冲击波除了会造成严重设备损毁，亦可能对现场人员造成致命伤害。若要成功发挥800V直流电源架构的效率优势，先进的安全规程是不可或缺的前提。

Arc flash events exist for all forms of power distribution, and they always carry the risk of potentially exposing personnel to devastating injuries in addition to severe equipment damage due to plasma energy and intense pressure waves. If we are to successfully capture the efficiency benefits of 800 VDC, we must ensure advanced safety protocols.

弧闪分析已成为交流数据中心标准且成熟的方法，但目前尚无行业通用标准或指导规范，用以应对人员在变流器供电的800V直流系统中可能遭遇的电气危害。现有的弧闪分析方法主要针对传统交流电气系统和常规的电池供电直流系统制定。

While arc flash analysis is already a standard, mature practice in traditional AC data centers, no industry-wide standard or guidance currently exists to manage the specific electrical hazards that personnel may encounter in converter-fed 800 VDC systems. Existing arc flash analysis methods were primarily developed for traditional AC electrical systems and conventional, battery-based DC systems. 

这就导致传统的弧闪计算方法无法准确反映AI数据中心中由变流器供电的800V直流电源架构的实际故障特性，用现有标准得出的风险评估结果与实际存在偏差。传统分析方法通常会假设故障电流处于稳态。然而，在800V直流电源系统中，电容放电在故障发生的最初几毫秒内占主导地位，使得故障电流具有高度瞬态性和时变性。由于变流器供电系统与传统电池供电直流系统在故障响应特性上存在根本差异，传统分析方法往往会高估或误判实际的弧闪风险。因此，新架构亟待与其适配的全新安全方法论。

Consequently, these legacy calculation methods do not fully reflect the converter-fed 800 VDC architectures emerging in AI data centers, and current standards have been shown to overestimate actual arc flash risk. Traditional calculations often assume steady-state fault currents. However, in 800 VDC systems, capacitor discharge dominates the earliest milliseconds of an event, making the fault current behavior highly transient and time-dependent. Converter-fed systems behave fundamentally differently than traditional battery-fed DC systems and, as a result, conventional analysis methods may overestimate or inaccurately represent the actual arc flash risk. These new architectures demand entirely new safety methodologies. 

要弥合这一差距，离不开整个技术生态系统内各职能部门以及跨领域的积极合作。

Bridging this gap requires active, cross-functional cooperation across the entire technology ecosystem.

施耐德电气认为：800V直流电压等级对安全提出了新要求，需要行业统一的标准和规范来应对。明确的操作指南——包括维护规范、弧闪风险防控、接地策略和人员培训——将确保800V直流系统基础设施的安全运行。行业工作组和监管机构需要通力合作，更新或完善适用于800V直流供电架构的相关规范与最佳实践。

“Operating at 800 VDC voltages introduces new safety considerations that must be addressed through industry-aligned standards and procedures,” said Bin Lu, Executive Vice President of the Power Products Division at Schneider Electric. “Clear operational guidelines, including maintenance practices, arc flash mitigation, grounding strategies, and training, will ensure safe handling of 800 VDC infrastructure. Industry working groups and regulatory bodies must collaborate to update or expand applicable codes and best practices.”

令人期待的工作已在业界多个重要组织中展开。电气和电子工程师协会（IEEE）正聚焦于直流电力系统的持续研究以及弧闪分析方法的完善，而开放计算社区（OCP）则围绕800V直流配电标准展开积极讨论，助力下一代数据中心架构走向标准化。

Promising work is already underway within key industry organizations. The Institute of Electrical and Electronics Engineers (IEEE) is focused on ongoing work around DC power systems and refining arc flash methodologies, while the Open Compute Project (OCP) is helping to standardize next-generation data center architectures through active discussions around 800 VDC power distribution standards. 

目前，运营人员需依赖多项零散的标准，且这些标准尚未针对新一代架构进行充分优化。例如，NFPA 70E《工作场所电气安全标准》对电气安全作出了系统规定，IEEE 1584标准则为交流系统的弧闪分析提供了方法支撑。此外，还可采用其他分析方法来评估直流弧闪特性，尤其适用于具有时变故障特性的系统，这些方法收录于NFPA 70E的附录D中，包括Stokes & Oppenlander法以及基于Paukert模型的计算方法。但由于NFPA 70E附录D仅作为资料性附录提供参考，其提及的方法尚未纳入正式、统一的标准计算框架，因此在实际应用中仍存在较大的差异。

Currently, operators rely on a patchwork of standards that are not fully optimized for these next-generation designs. For instance, NFPA 70E successfully addresses electrical safety in the workplace, supported by IEEE 1584 which covers arc flash analysis for AC system. Additional analytical approaches are available to evaluate to DC arc flash behavior in Annex D of NFPA 70E, particularly in systems with time-dependent fault characteristics. These include the Stokes and Oppenlander method as well as Paukert-based formulations. Annex D of NFPA 70E is informative, it means that these specific approaches are not yet part of a formalized, standard calculation framework, which unfortunately leads to high variability in how they are applied in practice. 

无论采用上述哪种分析方法，都需对特定系统参数和故障特性有详尽细致的了解，包括独特的电弧特性和复杂的系统响应。由于仅靠简单计算很难精准捕捉这些变量，所以通常需要借助先进的仿真与分析工具，才能精确模拟真实的系统行为。

Applying any of these alternative analytical approaches requires detailed, granular knowledge of specific system parameters and fault behavior, including unique arc characteristics and complex system responses. Because these variables cannot be easily captured through simplified calculations alone, advanced simulation and analytical tools are typically required to accurately model true system behavior. 

尽管行业标准在不断完善，但数据中心运营方当下急需切实可行的方法与实践指导。目前，行业亟待回答的一系列现实问题包括：当下是否可以对采用800V直流电源架构的AI数据中心进行弧闪分析？开展分析需满足哪些条件？现有分析方法是否适用、评估结果是否准确？实际应用中面临哪些挑战？

Industry standards are evolving, but operators need practical guidance now. Some of the questions the industry seeks to answer include: Is it possible to conduct an arc flash analysis today for an 800 VDC AI data center? What are the requirements? Can existing methods be used? How accurate are they? What are the challenges? 

为解答上述问题，施耐德电气正式发布行业首创评估报告《直流弧闪分析：800V直流数据中心的实践研究》：对机柜级和集中式800V直流架构这两款代表性架构进行了弧闪分析。仿真场景基于超大规模数据中心运营商的典型部署设计构建，并以NFPA 70E及与其配套的IEEE指导原则为基础开展评估工作。

To help answer these questions, Schneider Electric conducted a first-of-its-kind analysis evaluating arc flash risk in two emerging 800 VDC architectures: rack-level architecture and centralized architecture. Simulation scenarios were based on typical hyperscaler deployment designs and used the methods defined in NFPA 70E and supporting IEEE guidance as a foundation for our work.

基于此次评估获得的关键发现如下：

Our key findings from this evaluation demonstrate the following:

• 系统架构、电容器布局及故障切除特性是影响弧闪风险评估结果的关键因素。

• Arc flash outcomes depend strongly on architecture, capacitor placement, and fault-clearing behavior. 

• 瞬态行为至关重要：在800V直流系统中，弧闪由随时间变化的故障电流驱动，而在故障发生后的最初几毫秒内，故障电流主要由电容放电所主导。即使在最严苛的电容主导假设条件下，800V直流系统的弧闪风险仍然可控，且在许多工况下与典型交流系统的风险水平相当，甚至更低。

• Transient behavior matters: In 800 VDC systems, arc flash is driven by time-dependent fault currents, with capacitor discharge dominating the first milliseconds of an event.  Even under the most demanding capacitor-dominated assumptions, arc flash risk in 800 VDC systems can be managed, and in many cases, is comparable to or lower than typical AC systems. 

• 借助先进软件和数字孪生进行精细化建模，有助于更准确地评估弧闪风险。瞬态仿真软件和电力系统分析等工具表明，针对直流系统的简化弧闪分析方法，往往会高估电容主导系统中的弧闪风险。运营方可进一步利用先进软件及ETAP等数字孪生工具的分析结果优化保护策略，从而显著降低弧闪风险.

• Better modeling using advanced software and digital twins can help accurately portray arc flash risk. Software (such as transient simulation and power system analysis tools) shows that simplified arc flash analysis methods for DC systems often overestimate arc flash risk in capacitor-dominated systems. Operators can significantly reduce arc flash risk by informing their protection strategy with advanced software and digital twins like ETAP. 

此外，我们的分析进一步证实了弧闪风险水平不只取决于直流配电本身，也取决于具体的系统架构和配置。其中，电容器布局、反向阻断装置以及毫秒级保护响应，是保障800V直流系统安全部署的关键。这一点充分体现在两种架构的具体案例数据中：

Furthermore, our analysis reinforces that arc flash outcomes3 depend on architecture and system configuration, not on DC distribution alone, where capacitor placement, reverse-blocking devices, and millisecond-scale protection are key levers for safe 800 VDC deployments. This becomes evident when looking at the specific case study data for each architecture:

• 机柜级800V直流架构：在采用保守方法和假设条件的侧边电源柜（Sidecar）案例研究中，结果显示，弧闪产生的入射能量远低于1.2cal/cm²的个人防护装备（PPE）参考阈值。

• Rack-level 800 VDC architectures: In a sidecar case study using conservative methods and assumptions, results showed incident energy3 well below the referenced 1.2 cal/cm² PPE threshold3.

• 集中式800V直流架构：在采用保守方法和假设条件下，设施级案例的弧闪入射能量可能略高于机柜级设计。该分析还评估了系统拓扑结构的影响，展示了反向阻断二极管上下游不同故障位置如何影响反向馈电、峰值电流及弧闪风险水平。当通过标准保护装置对故障电流进行限时切除时，弧闪入射能量可降低至符合作业环境的水平，其整体风险表现与常见交流架构基本一致。

• Centralized 800 VDC architectures: The facility-level3 case study shows slightly higher incident energy3 potential than rack-level designs. The analysis also assesses the effect of system topology and demonstrates how fault locations both upstream and downstream of reverse-blocking diodes3 influence back-feed3, peak current3, and arc flash outcomes3. When fault contribution is time-limited with standard protection devices, arc flash incident energy3 is reduced to appropriate levels for the work environments and generally aligned with common AC architectures.

对于AI数据中心而言，向800V直流架构转型意味着巨大的效率提升，且无需以安全为代价。凭借科学的建模、合理的架构设计与快速的保护配合，运营方能够有效管控弧闪风险。施耐德电气的这项研究无疑为行业提供了可量化的安全依据，也为800V直流架构在AI数据中心的规模化部署注入了信心。

The transition to 800 VDC is an impressive leap forward for AI data center efficiency, and the industry doesn’t have to compromise on safety to achieve it. With the right modeling tools, appropriate architectural choices, and rapid protection coordination, operators can confidently mitigate arc flash risks. 

DKC交流群邀请

深知社创立于2017年，是数字能源行业工程技术人员学习、交流、合作、创新的虚拟社区平台。深知社秉承互联网精神，践行工程师文化，提倡全球视野、终身学习、乐于分享的博学深知理念，感兴趣的同学可以扫描下方二维码申请进群交流学习。

公众号声明：

推荐阅读：