Predicting the energetic performance of an automobile heat pump utilising a fixed capacity compressor and R1234yf using ANN modelling

Abstract

This study used experimental data to illustrate the accuracy of artificial neural network modelling for vehicle heat pump systems. The system had a four-way valve, thermostatic expansion valves, and a fixed-capacity compressor. The system used R1234yf refrigerant instead of R134a in automotive air conditioning systems. The system was evaluated using varying compressor speeds, indoor unit intake air flow rates, interior and outdoor unit inlet air flow temperatures, and relative humidity. The experimental system was tested 72 times using different control and data-collecting technologies to determine steady-state performance and how artificial intelligence may enhance it. The projected performance parameter of the automotive heat pump system employing R1234yf refrigerant was assessed using an artificial neural network model. Six scenarios were examined: compressor discharge temperature, indoor unit output airflow temperature, refrigerant mass flow rate, compressor power, heating capacity, and performance coefficient. Data was divided into training (269 patterns, 68.27 %) and testing sets (125 patterns, 31.73 %) to ensure accurate model development and performance assessment across different experimental configurations. This approach guarantees robust data handling and reliable artificial neural network predictions. The training and testing of the artificial neural network model of the automobile heat pump system with R1234yf was evaluated. In the best case, training R² was 0.99817, MSE 0.0012, and MEP 0.005. High prediction accuracy and robust linear associations were observed with R² = 0.99969, MSE = 0.0008, and MEP = 0.003. Future vehicle heat pump research using alternative refrigerants will benefit from this study's shortened experimental techniques and system performance estimates.