Predicting the energy consumption of a VRF heat pump using manufacturer performance data and limited experimentation for dynamic data collection

초록

The variable refrigerant flow (VRF) air conditioner is widely used because it can control indoor air conditioning units individually, allowing for efficient energy use. However, accurately modeling a VRF system is challenging because of its complex operating mechanism. Despite this challenge, manufacturers often only provide basic system information that adheres to regulatory standards, and they do not typically disclose detailed product specifications. In this study, for describing actual VRF system operation and estimating system performance reasonably, calibrated catalog-based performance estimation model is proposed using a small amount of experimental data. The proposed model uses a machine learning method to predict the power input of a VRF via the XGBoost algorithm. The results show that the prediction performance of the proposed model has an R2 higher than 0.9 and root mean squared error (RMSE) less than 0.2, whereas the typical catalog-based model has an R2 of 0.07 and RMSE of 0.54. The proposed model using only 20 % or more experimental data collected during the 12 h period outperforms the existing catalog-based model significantly.

키워드