Data-Driven Prediction of Compressor Efficiency for High-Temperature Heat Pumps

Compressors are decisive components in high-temperature heat pumps (HTHPs), as they dominate both system efficiency and cost. Yet in many design studies, compressor behavior is reduced to a single assumed isentropic efficiency (), regardless of refrigerant or operating conditions. This simplification introduces substantial errors in predicted performance and often leads to suboptimal system configurations. In reality, varies strongly with pressure ratio, suction state, and refrigerant properties, making accurate prediction essential for both thermodynamic and techno-economic evaluations. This contribution presents a data-driven framework for predicting compressor efficiency using only a small set of thermodynamic inputs typically available in early-stage design: inlet gas density, pressure ratio, evaporating temperature, condensing temperature, and temperature difference. An artificial neural network (ANN) was trained on a large dataset of commercial compressors spanning multiple refrigerants and technologies. The model learns to approximate across diverse operating conditions without requiring compressor-specific parameters such as displacement or rotational speed. Validation against withheld test data shows that the ANN consistently outperforms fixed-efficiency assumptions, reducing predictive errors while preserving generality across fluids. To bridge prediction with practical component choice, the ANN can be coupled with a k-nearest neighbors (kNN) search that retrieves actual compressor entries from the dataset matching the predicted performance. This hybrid approach provides not only an efficiency estimate but also a shortlist of feasible compressor options under the given operating conditions. The model is designed to be embedded in ASTRA (AI-based Superstructure for Thermal system Automated design), where reliable compressor performance models are essential for evaluating feasible configurations and guiding cost-effective designs. By replacing fixed efficiency assumptions with condition-dependent predictions, the framework provides a more realistic basis for automated system design and optimization. Such integration is expected to significantly influence overall results in terms of both thermodynamic performance and economic assessment, underlining the importance of accurate compressor modeling in advancing industrial heat pump deployment.