| Title |
Assessing the Potential of Virtual Sensors in Predicting Winter Balcony Window Surface Condensation |
| Authors |
김승빈(Kim, Seung-Bin) ; 손유라(Son, Yu-Ra) ; 양정훈(Yang, Jeong-Hoon) |
| DOI |
https://doi.org/10.5659/JAIK.2024.40.3.247 |
| Keywords |
Condensation prediction; Machine learning; Virtual sensor; Classification Algorithms; Data-driven |
| Abstract |
Persistent condensation in residential spaces can lead to structural damage and mold growth, posing health risks to occupants. While existing
studies focus on reducing condensation, there's a gap in research on condensation prediction. This study aims to explore the feasibility of a
virtual sensor for condensation prediction using machine learning and data from prior studies. A high-accuracy virtual sensor model was
developed and verified using condensation measurement data. Data preprocessing and Pearson correlation analysis were conducted, and input
variables were selected through ReliefF evaluation. Indoor and outdoor temperature and humidity were chosen as final input variables. A
prediction model was crafted using classification learning algorithms: Decision Tree(DT), Support Vector Machine (SVM), and Multi-Layer
Perceptron (MLP). Validation of the prediction model was performed using Confusion matrix, Accuracy, and F-1 score. The accuracy of the
virtual sensor model was 97.1% for Decision Tree, 98.5% for SVM, and 98.6% for MLP. The developed model is expected to effectively
prevent condensation in residential spaces susceptible to surface condensation. Future work will focus on integrating virtual sensors into
existing ventilation and air conditioning systems for practical application in residential spaces. |