Thermal analysis and performance optimization of supercritical carbon dioxide Brayton cycle based on ship waste heat

• SCO 2 power cycle system using CO 2 -propanebinary mixture. • Established a Convolutional Neural Network model for predicting system performance. • Using response surface methodology for multi-objective optimization to determine the optimal operating conditions. The global shipping industry is increasingly focused on energy conservation and emission reduction, driving the development of green technologies. Utilizing supercritical carbon dioxide (SCO 2 ) power generation to recover waste heat from marine engines has proven to be an effective approach to improve energy efficiency and reduce carbon emissions. In this study, waste heat from the main engine of a 9000 TEU container ship serves as the heat source for an SCO 2 power cycle incorporating a CO 2 -propane (C 3 H 8 ) binary mixture. A convolutional neural network (CNN) model was developed to predict system performance, using split ratio, mixing ratio, main compressor (MC) inlet temperature, and turbine inlet temperature as inputs. The results show that the determination coefficients R 2 of the model were 0.941 and 0.931 for the training set, while they were 0.922 and 0.912 for the test set. Multi-objective optimization based on response surface methodology (RSM) identified the optimal operating conditions as a split ratio of 0.39, mixing ratio of 14.2 %, MC inlet temperature of 34.17 °C, and turbine inlet temperature of 472.85 °C. Under these conditions, the SCO 2 system achieves a thermal efficiency of 25.81 % and an exergy efficiency of 41.55 %. These results demonstrate the significant potential of CO 2 -C 3 H 8 mixtures to enhance waste heat recovery in marine applications, contributing to cleaner and more efficient shipping energy systems.