Ultra-efficient Power Delivery Architecture and Topologies for IT Systems (UPDATE-IT)

数据中心是当今信息时代的支柱，预计到2020年将消耗约730亿千瓦时，仅在美国就相当于约73亿美元的电力成本。由于客户对在线数据和云计算的需求飙升，这是美国乃至全球电网负荷增长最快的一次。这种快速增长的消费极大地促进了全球碳排放。因此，以更高的效率更新数据中心的供电是重要的社会需求。该项目的目标是通过使用紧凑、可扩展、超高效、低热量、低成本和可靠的新的配电和转换架构来解决这一功耗瓶颈。如果成功，建议的系统的广泛应用将为仅在美国的数据中心每年节省超过81亿千瓦时或约8.1亿美元的电力消耗。这些节约反过来将减少美国道路上130万辆乘用车的有害温室气体排放。通过在通信系统、汽车、高性能便携式设备等其他信息技术应用中采用拟议的系统及其子系统，可以进一步扩大影响。这项研究将由研究生和本科生进行，他们将为未来的专业和教育发展机会配备集成电力电子和能源效率方面的知识和技能。这项研究还将涉及教育努力，包括提供官方课程和面向K-12学生的外联活动。为了实现研究目标，该团队将为未来的绿色数据中心和其他信息技术系统探索一种截然不同的配电和管理架构。采用能够有效支持大转换比的新的转换器拓扑，该架构将从电网到核心电压的功率转换级数从四级或更多减少到仅两级。第一阶段，将电网交流电压转换为直流母线电压，将服务器主板紧密地堆叠在服务器机架中，使每个服务器主板只能处理输入电网交流电压的一小部分，从而实现超高的功率转换效率、低成本和高可靠性。第一个转换阶段还利用了一种经过实验验证的新型智能电缆方法，该方法可以显著降低板载热量，从而大幅降低热管理成本。对于将直流母线转换为核心电压的第二阶段，研究团队将探索一种新的直接转换混合型DC-DC转换器拓扑系列，以在极大的转换比下实现更高的效率和功率密度。该研究计划包括设计、制造和验证多种新的混合转换器拓扑，集成不同类型的AC-DC和DC-DC混合转换器，智能电缆解决方案的设计和实施，系统故障保护电路的实施和测试，针对不同规格和应用的系统扩展，以及针对现有解决方案的基准测试。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Data centers, the backbone of today's information era, are expected to consume ~73 billion kWh in 2020, which amounts to ~$7.3 billion in electricity cost in the U.S. alone. This is one of the fastest growing loads on the electric grid in the U.S. as well as worldwide, because of customers' soaring demands for online data and cloud computing. This rapid-increasing consumption significantly contributes to global carbon emissions. Therefore, updating the power delivery for data centers with improved efficiency is an important societal need. The goal of this project is to address this power consumption bottleneck by using a new power distribution and conversion architecture that is compact, scalable, ultra-efficient, low-heat, low-cost, and reliable. If successful, a wide application of the proposed system could result in more than 8.1 billion kWh or ~$810 million annual savings in electricity consumption for data centers in the U.S. alone. These savings would, in turn, reduce harmful greenhouse gas emission of ~1.3 million passenger cars from U.S. roads. The impact can be further amplified by adoptions of the proposed system and its sub-systems in other information technology applications, such as communication systems, automotive, high-performance portable devices, etc. The research will be conducted by graduate students and undergraduate students who will be equipped with the knowledge and skills in integrated power electronics and energy efficiency for future opportunities in both professional and educational development. The research will also involve educational efforts including official curriculum offerings and outreach activities to K-12 students.To achieve the research goal, the team will explore a radically different power distribution and management architecture for future green data centers and other information technology systems. Employing new converter topologies that can efficiently support large conversion ratios, the architecture reduces the number of power conversion stages from the grid to core voltages from four or more to only two. The first stage, converting grid AC voltage to a DC bus voltage, electrically stack server boards in close proximity in a server rack to allow each of them to handle only a fraction of the input grid AC voltage, leading to ultra-high power conversion efficiency, low cost and high reliability. This first conversion stage also exploits a novel, experimentally validated smart-cable method that can significantly reduce on-board heat, leading to substantial thermal management cost reductions. For the second stage converting the DC bus to core voltages, the research team will explore a new direct-conversion hybrid DC-DC converter topology family to enable superior efficiency and power density at extremely large conversion ratios. The research plan includes design, fabrication, and verification of multiple new hybrid converter topologies, integrating different types of AC-DC and DC-DC hybrid converters, smart-cable solution design and implementation, system failure protection circuit implementation and test, system scaling for different specifications and applications, and benchmarking against existing solutions.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.