데이터센터 최적 열환경 제공을 위한 적응형 인공신경망 냉매 유량 예측 모델 개발

Development of Adaptive ANN Mass Flow Rate Prediction Model for Providing Optimal Thermal Environment in Data Center
  • 최영재; 
  • 박보랑; 
  • 최은지; 
  • 문진우

초록

Purpose: The purpose of this paper is to develop an ANN model for predicting the refrigerant flow rate of the cooling system in order to provide optimal thermal environment in the data center as well as an adaptive control algorithm which allows the predictive model to adapt to the diverse data center environments. Method: Two data center models were consisted by using ANSYS Fluent CFD tool. Model A was used to obtain data for ANN model training and the adaptive control algorithm was tested on the Model B. After developing the ANN model, the optimization process was conducted and the optimized model was employed in the adaptive control algorithm. Result: A model consisted with 4 hidden layers and 11 hidden neurons presented the highest accuracy of CVRMSE 0.47%. In addition, the adaptation test of the control algorithm was conducted by changing the air flow rate and the setpoint temperature of supply air. When the setpoint temperature was set to 20℃ and 25℃, the supply air temperature reached properly to the designated setpoint temperature in 3 cycles and 20 cycles, respectively. Each case presented maximum error of the temperature as much as 19.6% and 18.12%. After entering steady state, the ANN model adapted well on the setpoint temperature. In conclusion, the ANN model and the adaptive algorithm demonstrated a probability to be applied to the data center.

키워드

데이터센터; 인공신경망; 냉매 유량 제어; Data Center; Artificial Neural Network; Refrigerant Flow Control
제목
데이터센터 최적 열환경 제공을 위한 적응형 인공신경망 냉매 유량 예측 모델 개발
제목 (타언어)
Development of Adaptive ANN Mass Flow Rate Prediction Model for Providing Optimal Thermal Environment in Data Center
저자
최영재; 박보랑; 최은지; 문진우
DOI
10.12813/kieae.2019.19.6.073
발행일
2019-12
저널명
KIEAE Journal
권
19
호
6
페이지
73 ~ 79