Share

An empirical data-driven model for energy consumption prediction of a heat-pump-driven liquid-desiccant air-conditioning system.

Number: 2286

Author(s) : LEE J. H., KIM M., CHO H-Y., JEONG J. W.

Summary

In response to the growing significance of indoor humidity control in energy-efficient buildings, liquid-desiccant systems, capable of independently controlling air temperature and humidity, have emerged as alternatives to traditional vapor compression cooling systems. Subsequently, the integration of a heat pump into the liquid-desiccant system, facilitating control over both desiccant solution and air temperatures, gives rise to a heat-pump-driven liquid-desiccant (HPLD) air-conditioning system. However, there is a lack of previous studies on model derivation to predict its operating energy consumption in response to real-time variations in outdoor conditions and building thermal loads. Notably, the compressor energy consumption, a primary energy consumer in the heat pump, exhibits a pronounced sensitivity to variations in environmental conditions—encompassing outdoor and indoor conditions, building thermal loads, and desiccant-solution conditions. Relying solely on a physics-based model or performance specifications of the compressor under fixed rating conditions poses limitations in accurately predicting the operating energy consumption of the HPLD air-conditioning system in practical building applications.
This study aims to develop an empirical data-driven model to predict the compressor energy consumption of the HPLD air-conditioning system required to handle real-time variations in building thermal loads during summer. The data is acquired through on-site measurements from an actual building application, and the distribution of training to testing data sets follows an 8:2 ratios. The empirical data-driven model is derived using the Polynomial Regression (PR) method to provide an equation-based model. This study integrates domain knowledge from building engineering and mechanical systems into organizing the data-driven model structure. The model incorporates outdoor temperature and humidity, as well as sensible and latent heat capacity processed by the system, as input factors, with compressor input power designated as the output factor. The PR model, derived through a forward stepwise regression method, achieves R-square, root mean squared error, and mean absolute percentage error values of 0.976, 0.068, and 1.84%, respectively. Upon implementing the PR model, the predicted percentage error for compressor energy consumption during summer is determined to be 0.357%. Furthermore, the domain knowledge-encoded data-driven approach revealed reliable prediction accuracy, regardless of the quantity and quality of input data, in contrast to a pure data-driven method. These results highlight the exceptional accuracy and practical utility of our model in real-world scenarios, contributing valuable insights to the dynamic prediction of energy consumption in the HPLD air-conditioning system and promoting its implementation in building practices.

Available documents

Format PDF

Pages: 8 p.

Available

Free

Download

Details

  • Original title: An empirical data-driven model for energy consumption prediction of a heat-pump-driven liquid-desiccant air-conditioning system.
  • Record ID : 30033189
  • Languages: English
  • Subject: Technology
  • Source: 2024 Purdue Conferences. 20th International Refrigeration and Air-Conditioning Conference at Purdue.
  • Publication date: 2024/07/17

Links


See other articles from the proceedings (104)
See the conference proceedings

Indexing