Physics-inspired reduced order and refrigerant flexible model for the characterization of vapor injected positive displacement compressors

This study introduces a physics-inspired model for vapor injected, positive displacement, compressors, inspired from the principles of polytropic compression processes. The proposed model is pressure-ratio-based model utilizing suction, injection, and discharge pressures to predict key output parameters, including evaporator mass flow rate, injection mass flow rate, compressor power consumption, and discharge temperature. To validate the predictive accuracy of the model, an extensive collection of datasets was utilized, comprising four, in-house, combinations of rotary and scroll compressors operating with refrigerants R410A and R454B, as well as three additional datasets sourced from the literature totaling 221 data points. The proposed model was then compared with the best performing models in the literature for performance comparison. Models were trained and tested under interpolation, extrapolation and training with full datasets, to explore the robustness of the model. The proposed model demonstrated exceptional performance, with deviations in evaporator mass flow rate and compressor power consistently within 2% Mean Absolute Percentage Error (MAPE) under training and testing with full dataset and interpolation scenarios. The injection mass flow rate is predicted with an MAPE of less than 5% for 6 out of 7 datasets. Finally, the model predicts discharge temperature with deviations below 2 K Mean Absolute Error (MAE) across all cases evaluated. Additionally, the model includes a feature which allows accurate predictions on a particular refrigerant even if the model is trained on a different refrigerant. To evaluate this, a dataset comprised of two compressors, one rotary and one scroll compressor, each tested with refrigerants R410A and R454B were collected. The proposed model demonstrates 5% prediction accuracy when evaluated for sensitivity to different refrigerant, showing excellent performance extrapolation compared to the model in literature.