데이터센터 에너지 예측 모델을 활용한 외기습구온도별 최적 유체 온도 도출에 대한 연구

Abstract

As societal demands for carbon neutrality continue to rise, reducing energy consumption in the building sector has become essential. Among various building types, data centers consume a significant amount of energy, prompting numerous studies to explore the application of Water-side Economizer (WSE) systems that enable Free Cooling operation for enhanced energy efficiency. This study aims to optimize the performance of WSE systems by proposing a control strategy that minimizes cooling energy consumption in response to varying outdoor conditions. As a preliminary step, a cooling energy prediction model for a data center was developed to derive optimal fluid temperature setpoints that minimize energy usage according to outdoor wet-bulb temperature. The predictive model was constructed using Long Short-Term Memory (LSTM) and Light Gradient Boosting Machine (LightGBM), and demonstrated high accuracy and reliability with R-squared value of 0.995, CV (RMSE) of 4.31%, and an MAE of approximately 1.13 kW. Based on this model, optimal fluid temperature combinations were determined for each outdoor wet-bulb condition throughout the year. These results are expected to serve as a foundational guideline for developing future control algorithms and establishing energy-efficient operation strategies.