인공지능을 활용한 Internet Data Centre의 제어조건별 에너지사용량 예측 모델 개발

Abstract

Currently, most of the fluid-related elements such as chilled water, condensing water, and air volume of the internet data centre HVAC system are operated at fixed values annually or seasonally, which causes a decrease in energy efficiency. Therefore, it is necessary to establish a control method that minimizes energy usage by controlling the chilled water and condensing water temperature according to the outdoor air temperature. A model to predict energy usage was developed for efficient control of the entire system. Through PCC (Pearson Correlation Coefficient) analysis, elements affecting energy usage were selected as control variables, and energy usage at each temperature was derived by varying the chilled and condensing water temperature. Afterwards, a prediction model was developed using the results, and it was confirmed that the reliability was secured as CV (RMSE) was 2.19%, NMBE was 0.024%, MAPE was 0.72%, and R2 was 0.885.