| 논문명 |
터보 냉동기의 기계학습 모델과 하이브리드 모델 비교 / Machine Learning model vs. Hybrid model of a turbo chiller / 춘계-05. 건축환경및설비 |
| 저자명 |
박성호(Park, SungHo) ; 안기언(Ahn, Ki Uhn) ; 황승호(Hwang, Aaron) ; 최선규(Choi, Sunkyu) ; 박철수(Park, Cheol Soo) |
| 수록사항 |
대한건축학회 학술발표대회 논문집, Vol.38 No.1 (2018-04) |
| 주제어 |
인공신경망 ; 하이브리드 모델 ; 기계학습 ; 그레이 박스 모델 ; 제 1법칙 기반 모델 ; 데이터 기반 모델 ; Artificial Neural Network ; Hybrid model ; Machine learning ; Grey-box model ; Physics-based model ; Data-driven model |
| 요약2 |
In the previous research, the authors identified a valid range of inputs using Gaussian Mixture Model (GMM) and developed an Artificial Neural Network (ANN) model for optimizing the operation of a turbo chiller in an offifce building. In this study, the authors developed a new hybrid model which is different from the gray-box model and compared it with the ANN model developed in the previous study. The hybrid approach is a kind of a data-driven model but it also uses minimal physical knowledge (physics-based regression equations from the EnergyPlus). It is found that the hybrid model performs closely to the ANN model (MBE=2.05%, CVRMSE=7.98%). It is believed that the hybrid ANN model can be applied for the wider range of inputs since it is data-driven and based on physical knowledge, which will be validated in future study. |
