DNN 모델 기반 HVAC 제어 설정값 최적화 알고리즘 개발 및 효과 검증에 대한 연구

Abstract

The purpose of this study is to develop a DNN model-based HVAC control setting value optimization algorithm and verify its effectiveness. A Deep Neural Network (DNN) model was developed to predict operating costs for heat source systems equipped with absorption chiller. Using this, optimal control algorithm for HVAC was developed. In addition the effect was verified by applying it to the actual building located in Suwon. The prediction accuracy of the DNN model developed through actual operation data was 0.769 percent for R2 score and 18.93% for CVRMSE, satisfying the performance criteria suggested by ASHRAE Guideline 14. The algorithm that interlocks with the BEMS(Building Energy Management System) of the target building calculated a optimal control setting value that minimizes operation costs by fetching operation data every 20 minutes. This was transmitted to the BEMS to control the HVAC. As a result of applying the algorithm, it was confirmed that the operating cost was reduced by operating the cooling water supply temperature close to the wet-bulb temperature and reducing steam consumption. It was confirmed that it was reduced by 4.60% when comparing simple costs, and reduced by 31.0% when converted to the same cooling capacity costs.