Journal of Korean Institute of Architectural Sustainable Environment and Building Systems ISSN:1976-6483(Print) 2586-0666(Online) 한국건축친환경설비학회논문집

Underground utility tunnel is located underground, which makes the ventilation system operation and internal environment monitoring important. By designing a machine learning regression model for predicting the internal environment, the effects of the input variables were verified. Temperature, relative humidity and fan operation hour data was collected hourly by sensors installed near the exhaust fan. The external weather data was obtained from the nearest meteorological station in the same experimental period. The results of machine learning regression analysis model showed accuracy of random forest (R²: 0.891), support vector regression (0.800), k-nearest neighbor (0.774), and multi linear regression (0.744). In order to design the accurate model, the n estimator was set to 157, in which the accuracy of temperature prediction was 0.837 by R², and the accuracy of humidity prediction was 0.950. Using the RF model, the correlating factors correlating were verified and the internal environment of underground utility pipe according to fan operation time was verified.