Optimization of infrared emissivity design for radiative cooling windows using artificial neural networks: Considering the diversity of climate and building features

Proper design of emissivity is key to promoting passive cooling for radiative cooling (RC) windows. This work developed an artificial neural network (ANN) to guide the design of RC window emissivity. The ANN model demonstrates good predictive performance with a coefficient of determination greater than 0.8 in all verification cases. Long-term predictions were made for RC effects of adjustment of window emissivity under eight tropical climate cities and different building characteristics. Considering the building characteristics, vertical windows are more suitable for selective high emissivity (SE) designs in most cases; while skylights are more recommended for broadband high emissivity (BE) designs when the window-to-wall ratio is large. There exists a critical window-to-wall ratio that makes RC windows more suitable for SE designs into more suitable for BE designs; this value generally fluctuates around 75 %, depending on specific orientations and climates. Considering the climate characteristics, BE designs should be prioritized when RC windows are in dry climates with low ambient temperatures and strong solar radiation; whereas SE designs should be prioritized when RC windows are in humid climates with high ambient temperatures but relatively weak solar radiation. This work provides more convenient and targeted guidance for the design of RC window emissivity.