Data-driven model for predicting power consumption of heat-pump-driven liquid-desiccant systems in building applications

Abstract

With the growing emphasis on indoor humidity control in energy-efficient buildings, heat-pump-driven liquid-desiccant (HPLD) systems have emerged for their ability to independently control air temperature and humidity. Previous studies have estimated their power consumption using theoretical models, which are often limited by structural complexity and challenges in physical interpretation. Additionally, theoretical models yield prediction inaccuracies when applied to buildings because they lack sensitivity to dynamic environmental variations typically observed in real-building conditions. This study develops a simplified data-driven model using real-building measurements to predict power consumption, capturing partial-load compressor performance under variable outdoor conditions and indoor thermal loads during the summer season. A polynomial regression method is used to develop the model in a simplified equation-based form. The developed model achieves Rsquared, root mean squared error, and mean absolute percentage error (MAPE) values of 0.9583, 0.0668, and 8.37 %, respectively, in predicting the partial-load compressor power. Moreover, the model predicts the compressor energy consumption during summer operations with a percentage error of 0.36 %. Its adaptability is further validated against previous studies on HPLD systems with diverse features and specifications, within an acceptable error bound of +/- 20 % and a MAPE of 11.1 %. These results highlight the exceptional prediction accuracy and practical utility of the model developed in this study, supporting its adoption in various building application scenarios and replacement of theoretical models.