This paper presents a framework for the optimal operation of data centers, leveraging their heating, ventilation, and air conditioning unit, delay-tolerant information technology workload and battery storage system for participating in demand response programs. In this context, an model predictive control based control framework has been developed that guarantees the reliable operation of the data centers core activities. We derive a modeling approach to represent the dynamics of the data centers subsystems and validate it for a data center test bed via practical experiments. Hereby, the thermal subsystem leads to deviations of less than 0.60 K in the modeled outlet temperature. The validated model is used for incremental prototyping of the proposed control via simulations under uncertainties. The results demonstrate a mean absolute error of the relative deviations between the data center consumption and the target load profile of 2.71% for an incentive-based scenario and a cost reduction of 3.86% for a price-based scenario.

中文翻译：

---

参与需求响应计划的数据中心控制策略


本文提出了一个数据中心优化运行的框架，利用其供暖、通风和空调装置、延迟容忍信息技术工作负载和电池存储系统来参与需求响应计划。在此背景下，开发了基于模型预测控制的控制框架，保证数据中心核心活动的可靠运行。我们推导出一种建模方法来表示数据中心子系统的动态，并通过实际实验在数据中心测试台上对其进行验证。因此，热子系统导致建模出口温度的偏差小于 0.60 K。经验证的模型用于通过不确定性下的模拟来对所提出的控制进行增量原型设计。结果表明，在基于激励的场景中，数据中心消耗与目标负载曲线之间的相对偏差的平均绝对误差为 2.71%，而在基于价格的场景中，成本降低了 3.86%。