컨테인먼트형 데이터센터 최적 제어 알고리즘을 위한 열환경 예측모델 개발

Abstract

Purpose: This study aimed at developing a temperature prediction model for a containment data center. The predictive model must be guaranteed with stability and accuracy in order to be used for real-time control. Therefore, statistical evaluation was conducted to verify the prediction performance of the proposed model. Method: The predictive models were developed using four representative machine learning algorithms. A thermodynamic based containment data center and cooling system were modeled by MATLAB & Simulink software. The initial and optimized models were evaluated by R2 and Cv(RMSE), and the model with the highest performance was applied to the simulation. Result: In the initial models, RF and ANN presented highest accuracy on R2 (0.89) and Cv(RMSE) (17.85%), respectively. After the optimization, ANN presented the best prediction performance on both R2 (0.99) and Cv(RMSE) (0.94%). The result supports the accuracy and stability of the ANN model to be used for real-time control, and based on which the optimal control algorithm will be developed on further study.