Model-Free Control of Data Center Compressor-Based Cooling Systems

Data centers are energy intensive facilities and nearly half of the total energy consumption is used to cool the IT equipment. In particular, data center cooling systems have traditionally relied on compressor-based cooling. Even though, more recently a number of alternatives for reliable cooling have been introduced (e.g. free-cooling technologies), a minimal amount of vapour compression cooling is used in order to ensure backup cooling and to reduce humidity or to peak temperature inside the computer rooms. The efficient operation of this kind of energy-intensive cooling systems may decrease their energy consumption and diminish the environmental impact. In this paper, a model-free control approach is presented for the efficient management of this kind of cooling system. Indeed, data center cooling needs can be met with various combinations of modes of operation that can be set by the control system to maximize the compressor-based cooling system overall operating efficiency. Specifically, we have developed a Matlab-based Computer Room Air Conditioning (CRAC) simulation environment in order to design an Extremum Seeking Control scheme. The effectiveness and performance of the proposed model-free control approach have been evaluated by means of simulations in CRAC different operating conditions.