기계학습 모델별 공조시스템 고장현상 검출 및 진단성능 평가
신재윤, 채영태 · 대한설비공학회 동계학술발표회 논문집 · 2018
HVAC system is a important part of the building and consumes over 50% of building energy. The fault in HVAC system causes energy waste, so the real-time FDD(Fault Detection and Diagnosis) can improve building energy performance. In this study, the FDD performance has been evaluated using ASHRAE-1312 dataset based on machine learning. The three machine learning models which neural network, decision tree and ensemble are selected for evaluation with 11,520 training dataset and 720 new dataset. As a result, the classification performance of training data shows ensemble(92.5%), neural network(82.9%) and decision tree(72.7%) and the FDD performance of new data shows ensemble(99.1%), neural network(85.6%) and decision tree(76.1%). FDD based on machine learning results effective performance so that the FDD module have possibility to enhance building energy performance.
Cited by 0 works.