在可持续AI数据中心的最佳可再生能源战略中利用深度强化学习
Tianqi Xiao1, Fengqi You1,2,3,4
1College of Engineering, Cornell University, Ithaca, New York 14853, United States.
Environmental science & technology
|December 17, 2025
概括
深度强化学习 (DRL) 优化了人工智能数据中心的效率,减少了能源和水的使用. 这种智能控制策略与可再生能源相结合,为人工智能基础设施提供了具有成本效益的脱碳化.
科学领域:
- 人工智能的人工智能
- 可持续的计算 可持续的计算
- 数据中心运营数据中心运营
背景情况:
- 人工智能计算的增长推动了数据中心的大量能源消耗,引发了可持续性问题.
- 现有的数据中心控制方法缺乏效率和具有成本效益的可再生能源集成.
- 减少人工智能基础设施的碳排放对于实现全球净零目标至关重要.
研究的目的:
- 开发和评估一个新的框架,将深度强化学习 (DRL) 与AI数据中心的成本效益优化相结合.
- 提高能源效率,减少用水量,并使经济可行的可再生能源整合成为可能.
- 通过智能控制和可再生能源战略,评估成本和排放减少的潜力.
主要方法:
- 开发了一个框架,将DRL控制与成本效益优化的合.
- 利用了七个现实世界的人工智能工作负载和开源网格/可再生成本数据.
- 评估了全球十个地点的能源,水和碳性能.
- 与ASHRAE标准一致的基线控制器进行基准测试.
- 评估需求响应和电池存储集成.
主要成果:
- DRL实现了近乎最佳的自由冷却,节省了超过6%的能源和8%的水.
- 在特定条件下,更高的服务器利用率可以将辅助冷却降低高达60%.
- 综合战略将50%的减排总成本降低了9-28%.
- 现场可再生能源的减排成本在各个地点从每107美元到500美元不等.
结论:
- 智能DRL控制与可再生能源战略相结合,为人工智能基础设施提供可扩展和具有成本效益的脱碳.
- 开发的框架与人工智能数据中心的全球效率和净零目标保持一致.
- 这种方法证明了一条可行的途径,可以减少人工智能计算对环境的影响.
相关概念视频
01:16Non-equilibrium in the Cell
5.3K
An important concept in studying metabolism and energy is that of chemical equilibrium. Most chemical reactions are reversible. They can proceed in both directions, releasing energy into their environment in one direction, and absorbing it from the environment in the other direction. The same is true for the chemical reactions involved in cell metabolism, such as the breaking down and building up of proteins into and from individual amino acids, respectively. Reactants within a closed system...
5.3K
00:50Unrenewable Cells
2.8K
In humans, the photoreceptor cells of the eye and sensory hair cells of the ear lack stem cells. These cells are thus unrenewable and cannot be replaced when they are damaged or destroyed.
Photoreceptors
The retina is composed of several layers and contains specialized cells called photoreceptors. The photoreceptors (rods and cones) change their membrane potential when stimulated by light energy. There are two types of photoreceptors—rods and cones—which differ in the shape of...
Photoreceptors
The retina is composed of several layers and contains specialized cells called photoreceptors. The photoreceptors (rods and cones) change their membrane potential when stimulated by light energy. There are two types of photoreceptors—rods and cones—which differ in the shape of...
2.8K
01:23Reinforcement
777
Positive and negative reinforcement are key concepts in operant conditioning, a learning process where the consequences of a behavior affect the likelihood of that behavior being repeated.
Positive reinforcement occurs when a behavior is followed by the presentation of a rewarding stimulus, increasing the frequency of that behavior. For example:
Positive reinforcement occurs when a behavior is followed by the presentation of a rewarding stimulus, increasing the frequency of that behavior. For example:
777
01:24Reinforcement Schedules
427
Positive reinforcement is a powerful method for teaching new behaviors to both animals and humans. B.F. Skinner demonstrated this with his experiments using rats in a Skinner box. When a rat pressed a lever, it received a food pellet. This immediate reward encouraged the rat to repeat the behavior. This method, where a reward follows every instance of the behavior, is known as continuous reinforcement. It is highly effective for establishing new behaviors quickly.
Once a behavior is learned,...
Once a behavior is learned,...
427