Modeling and Control of Hybrid Si-Based Micro-Fluid Cooling System for Data Center Application

Liquid cooling system for IT device shows potential of energy efficiency, which is very attractive to data center owner. To exploit the capability of the liquid cooling system, proper controller should be designed. This paper presents the design of model and controller for the liquid cooling system. First, the mathematical models for each individual component are derived from physical laws. Model parameters are determined by the performance curve. Based on the working point selection program and control parameter optimization process, the proportional-integration controller with feedforward path (PI-FF) is designed. Target tracking task and disturbance rejection task are used to evaluate the performance of the PI-FF controller. The results and simulations not only validate the capability of proposed control method, but also in-depth revealed the key physical bottleneck that prevent improving the response of the controller. This result suggests that heat transfer process from junction to the coolant should be as shot as possible, and therefore direct bonding the micro cooler to the chip could be the most efficient way to reduce their interface thickness.