| Title |
Predicting Supply Air Temperature in Air Handling Unit Using Machine Learning-Based Automation Algorithm |
| Authors |
Yusun Ahn ; Goopyo Hong ; Byungseon Sean Kim |
| DOI |
https://doi.org/10.6110/KJACR.2020.32.1.037 |
| Keywords |
공조기; 예측; 급기온도; AutoML(Automated Machine Learning); Auto-sklearn(Auto-sklearn); ANN(Artificial Neural Networks) Air handling unit; Prediction; Supply air temperature; AutoML(Automated Machine Learning); Auto-sklearn(Auto-sklearn); ANN(Artificial Neural Networks) |
| Abstract |
More than 50% of the energy used in commercial buildings is consumed by building cooling, heating, and ventilation energy. To conserve building energy, some of the new buildings have begun to operate through the BEMS (Building Energy Management System). But, existing buildings are operated based on the empirical judgment of the building manager. The purpose of this study was to predict air supply temperature of air-handling units by applying an automatically implementation machine learning model and hyper-parameters suitable for non-professional users. The building used for the experiment was a large hospital building, using four AHU (Air Handling Unit) measured data from same floor on the different zones. The automated-data algorithm for analysis uapplies Auto-sklearn, which is AutoML (Automated Machine Learning) as finding an optimal model. As a result, various models (Gradient Boosting, Support Vector Regressor, Adaboosting, Random Forest) were recommended. For comparison, the ANN (Artificial Neural Network) model, used most frequently in the field of building energy, was compared with the hyper-parameters derived through iterative tuning in previous studies. Comparing ANNs with the recommended models and parameters, the accuracy of similar or recommended models was 0.53% higher on average with CvRMSE and 0.54% with MBE. It is reasonable to apply models and parameters through an automated data analysis algorithm. |