Volume 67

Frosting Image Recognition detection for R290 Secondary Loop Air Conditioning Heat Pump Systems in Electric Vehicles Panlin Gu, Kang Li, Jingxuan Xu, Zilong Wang, Jinjun Yan, Liaokuo Song, Qize He, Chao Li, Soheil Mohtaram

Abstract

R290 secondary-loop air conditioning heat pump
systems are a promising alternative for electric vehicles
due to their environmental friendliness and high thermal
efficiency. At ambient temperatures around 0°C, heat
exchangers are prone to frosting, which reduces heating
performance; rapid frost detection can significantly
enhance system efficiency. In this study, a deep learning
based image recognition method was developed for frost
detection in R290 heat pump systems. Experiments were
conducted to observe the frosting process and obtain
heat transfer characteristics. The results indicate that
frosting can be divided into three stages: a sharp decline
in the initial stage, a stable development stage, and a
rapid decrease during the frosting stage. A modified
deep learning approach was used to detect these stages,
achieving over 95% accuracy across different
temperatures after 150 training rounds. These results
demonstrate that the method is reliable and effective
under varying conditions.

Keywords R290, air conditioning heat pump system, defrosting, image recognition