Chia-Chin Chen
National Center for High-Performance Computing
E-mail:chester@narlabs.org.tw
Yo-Ching Lin
National Center for High-Performance Computing
E-mail:1203043@narlabs.org.tw
Wei-Cheng Huang
National Center for High-Performance Computing
E-mail:whuang@narlabs.org.tw
Yu-Bin Fang
National Center for High-Performance Computing
E-mail:ybfang@narlabs.org.tw
Abstract
The International Energy Agency (IEA) points out that the power consumption of global data centers continues to increase significantly and is expected to exceed 1,000 terawatt hours (TWh) in 2026. The cost of air conditioning accounts for about 40% of the total electricity cost. Therefore, the air conditioning energy saving of the data center computer room is very important. This research aims to build a simulation training platform that combined AI models, which were trained by long-term computer room cooling and air conditioning system data, with the “Artificial Intelligence Internet of Things Cloud Service(AI2CS)” or the hybrid from AI2CS and similar techniques. Personnel input simulated operation data through the “Operation and Display Interface” of the simulation training platform. Then, these simulated data was sent to the inference engine for prediction. The prediction data and the input of simulated operation data were stored in the data hub and displayed in the “Operation and Display Interface” ”, allowing personnel to review the results of operations. In this way, the seemingly disordered data of air conditioning system operation and monitoring was transformed into orderly operations and output through the AI model. Thereby, personnel were able to repeatedly simulate the various operation of air conditioning system in the computer room to understand the power consumption of the chiller unit. The operation of the system and the experience of energy-saving were learned and inherited smoothly.
Keywords :AI; Computer room energy saving; Digital learning