Objective In low-temperature environment, the two-stage compression heat pump system exhibits higher heating efficiency compared to the single-stage heat pump system. It is more energy-efficient and environmentally friendly to apply the former system at urban rail transit stations in northern regions. However, the refrigerant leakage in heat pump systems leads to significant energy waste. When the refrigerant leakage exceeds 25% of the rated charge, the system’s energy efficiency ratio decreases by 15%, and its cooling/heating capacity drops by 20%. Therefore, real-time monitoring of refrigerant charge in heat pump systems and research on refrigerant fault diagnosis are crucial.

Method A quasi-two-stage heat pump system is studied, and a hierarchical feature screening method is proposed to identify fault characteristic parameters for refrigerant charge in the heat pump system. The Pearson correlation coefficient is used to analyze the state parameters of the system. The key characteristic parameters of refrigerant charge faults are finally determined by eliminating redundancy and other methods. Based on the characteristic parameters, a segmented empirical formula VRC (virtual sensor model) is established to predict the refrigerant charge.

Result & Conclusion Under different refrigerant charges, the fault characteristic parameters obtained through the hierarchical feature screening method are used to refine the VRC model, effectively improving the accuracy of the model. Test results show that an overall accuracy of 90.2% is achieved in diagnosing the system charge using the segmented VRC model, demonstrating its excellent diagnostic performance.