A visual defrosting control method for air source heat pump system based on machine vision

Frosting on the coil of the air source heat pump (ASHP) significantly affects its performance. An accurate defrosting control method is essential for the ASHP unit. Here, a novel visual defrosting control method based on machine vision is proposed. The frosting degree of the unit is accurately obtained by frosting image analysis with digital image processing to achieve on-demand defrosting. Fractal dimension, frosting degree coefficient, and single-circuit frosting evenness value are introduced to evaluate the frosting degree of coil images. The image acquisition point is determined with confusion matrix. Based on this, the logical framework of the proposed method is constructed, in which the defrosting starting and termination conditions are set as the moment when the frosting degree coefficient is equal to 0.9 and 0, respectively. A multi-variable model of fractal dimension is fitted, which improves the generalization of the proposed method. Finally, comparative experiments are conducted between the proposed method and the traditional temperature-time (T-T) method in typical cases. The results indicate that the COP and total heating capacity increase by 11.8%, 10.6% and, 7.4%, 6.6%, respectively, and the defrosting frequency is reduced by 75%, which demonstrates dominant advantages of the proposed method in accuracy, energy-saving, and indoor thermal comfort management.