AI数据中心面临电网新规
TechRepublic 报道,华盛顿特区附近的一次短暂电压扰动导致 3.8 GW 的数据中心切换至备用电源凸显了大型 AI 负载对电网的影响。德州参议院第 6 号法案要求数据中心承担新增电网基础设施费用,NERC 正制定类似的穿越故障规则并预计于 2026 年底前获批,PJM 也提出了相关提案。Nvidia 在 DSX 架构中推出机架级功率平滑功能,ON.energy 则发布了采用双向电力电子和电池的 AI UPS。
电网兼容性正成为大型 AI 数据中心必须满足的合规与基础设施要求,其重要性已与算力、散热和供电并列。
Data centers used to need protection from grid disturbances. They still do. However, the scale of the latest AI data centers has reversed the equation. Now the grid needs to be safeguarded from problems caused by AI data centers.
A downed power line that generated a brief voltage disturbance near Washington, D.C. caused 3.8 GW of data centers to revert to on‑site backup power within seconds. This represented about 3% of demand and caused lights to flicker as far away as Chicago.
The system operator (PJM Interconnect) scrambled and restored stability within about ten minutes. Similar events occurred in 2024 and 2005 in Virginia after scores of data centers tripped off the grid and reverted to backup generation.
That’s just one way data centers can suddenly cause grid destabilization. Large AI factories suffer from massive swings in power use. They can surge up and down – sometimes from 1 GW to almost nothing and back again several times a second in extreme cases.
If not properly managed, rapid changes in demand can contribute to voltage or frequency disturbances, potentially affecting grid reliability and sensitive equipment.
#Nvidia#ON.energy#PJM Interconnect#North American Electric Reliability Corporation#Oracle Cloud#Nvidia DSX architecture#AI UPS#Texas Senate Bill 6