Prediction of Chiller Thermal Characteristics in Radiant Cooling Systems

This study is intended to develop an artificial-neural-network (ANN)-based model that computes the necessary chiller thermal energy/heat transfer rate in a desiccant and sensible heat recovery wheel integrated vapor absorption-based radiant cooling system (RCS). Solar and natural gas were the main energy sources. The RCS was installed with a 4-floor corporate building of
$4998 \ \mathrm{m}^{2}$
. By applying the time-variant weather data, radiant and dedicated outdoor air system (DOAS) chiller capacities of the RCS may be pre-determined securely in a more convenient manner. During the development, the ANN structure was fed with input neurons containing solar irradiance, outdoor temperature, and moisture contents. These had statistical connection with the output neuron denoting radiant and DOAS chiller heat transfer rates. The entire ANN framework was optimized for the analysis and the parametric predictions. Results were studied using mean-squared error, correlation coefficient (R), and error histograms. For heat transfer rate predictions of radiant and DOAS chillers, R values were determined above 86% and 95%, respectively. Finally, histograms exhibited that for most of the samples, errors of 0.0025 and 0.03366 were involved with parametric predictions of radiant and DOAS chillers, respectively. Therefore, the presented ANN was proven to have the capability to be implemented in sustainable energy-based building designs.