An Optimization Algorithm to Design Compact Plate Heat Exchangers for Waste Heat Recovery Applications in High Power Datacenter Racks

摘要

Heat dissipations of servers in datacenter racks are following an ever-increasing trend, breaking the economical heat removal limits of traditional air-based cooling technologies. Currently, an average of 40-45% of the total datacenter energy consumption is needed to cool servers, presenting significant challenges to maintain energy efficiencies and also noise levels within US OSHA standards. The present paper focuses on the determination of the optimal design of a compact plate heat exchanger (PHE), acting as an overhead refrigerant-to-water condenser of a macro-scale thermosyphon, which dissipates the total heat from a datacenter rack into a cooling loop for waste heat recovery applications (e.g. district heating network). PHEs are already the preferred solution for many industrial and domestic applications (especially small to medium size refrigeration and heat pump systems), since they provide higher heat transfer performance, higher flexibility toward the targeted application and lower pressure drops compared to the conventional tube-in-tube and shell-and-tube heat exchangers. Furthermore, due to the numerous variables involved in the design of PHEs, such as plate number, plate footprint size, geometry of the corrugation pattern (i.e. chevron angle, pressing depth, etc.), an optimization analysis and corresponding simulation tool is auspicious to finding the optimal design of these units to accommodate the targeted heat rates of datacenter racks. Hence, this study proposes a novel optimization process which incorporates a local simulator (an improved version compared to the one presented at ITHERM 2018)for accurately rating and designing PHEs over a wide range of operating conditions, plate geometries and working fluids. The improved simulator uses a local one-dimensional effectiveness-NTU approach, with a local implementation of mass, momentum and energy equations, coupled with newly upgraded methods for condensation heat transfer coefficients and frictional pressure drops, as well as newly added prediction methods to handle the inlet and outlet port pressure drops on the overall thermal-hydraulic performance of PHEs. The local simulator is the central function upon which the optimization analysis is performed by employing a genetic algorithm. The latter has been proven to work exceptionally well when dealing with many, highly non-linear functions, such as heat transfer coefficients and pressure drops in two-phase flow. The primary variables which determine the design of the PHE to be optimized are subject to several constraints. The objective functions chosen for this optimization analysis are the total heat rate and pressure drop on the condensing side of the macroscale rack thermosyphon. While the former is to be maximized, the latter is to be minimized in order to achieve a highperformance index (defined as the ratio of the heat rate over the pressure drop).