데이터센터 냉방 에너지 절약을 위한 최적 냉수 유량 예측·제어 알고리즘 개발

Abstract

Purpose: This study aimed at developing an adaptive control algorithm using ANN based on chilled water mass flow control to provide the optimal indoor thermal environment of the data center and save cooling energy. Method: The predictive model inherent in the control algorithm uses the model developed in preliminary research. The control algorithm including the predictive model was developed using three techniques with relearning function. To verify the adaptability of the finally selected ANN prediction-based adaptive control algorithm, it is compared and evaluated through simulations with the existing widely applied ON-OFF and PID controllers. Result: Among the three relearning techniques, the RMSE of the control algorithm to which the sliding window was applied is 0.04 (℃), which has the highest prediction accuracy. Thus, was selected as the final control algorithm model. To verify the adaptability and scalability of the selected ANN control algorithm, the containment size and the set-temperature scenario were changed into two and compared and analyzed with ON-OFF, PID controller using simulation. As a result, the ANN controller showed the highest accuracy with RMSE 0.23℃ and 0.24℃, respectively, in both control scenarios. Also, It showed the best performance in terms of maximum temperature difference and energy consumption. That is, the adaptability and scalability of the ANN-based control algorithm to the new environment have been reasonably verified.