The National University of Singapore 𞓜 uses model free reinforcement learning to evaluate the energy efficiency of the energy efficiency data center

【 title 】Energy saving evaluation of an energy efficient data center using a model-free reinforcement learning approach

【 The author team 】Muhammad Haiqal Bin Mahbod, Chin Boon Chng, Poh Seng Lee, Chee Kong Chui

【 Date of publication 】2022.6.21

【 Thesis link 】https://www.sciencedirect.com/sdfe/reader/pii/S0306261922007309/pdf

【 Recommended reasons 】 In order to reduce cooling energy consumption , It is recommended that the data center increase the server inlet air temperature setpoint . However , In tropical climate , Data center operators are still working at low temperatures . This paper proves that , A tropical climate with a floating set point and reduced temperature reduces the overall energy consumption of the data center , Instead of raising the temperature statically . This paper achieves this by applying the deep reinforcement learning algorithm to the hybrid data center model , The model is built by collecting data from an efficient data center . This leads to an optimal control strategy , Minimize energy consumption costs , At the same time, operate under the required set of operation constraints . This paper evaluates the behavior of control strategies , To illustrate the exact source of energy savings . The deep reinforcement learning algorithm learns through continuous interaction with the established data center model , Without prior knowledge of the data center . The algorithm is trained under the full load and partial load configuration of the data center . The test results show that , In a data center that already has cooling efficiency , Supply through targeted cooling , Save energy .