Multi-objective parameter optimization of a methanol-mediated carbon-closed power system assisted by a neural network prediction model

A novel methanol-mediated carbon-closed power system integrating green methanol synthesis, carbon capture and utilization, and high-efficiency power generation is proposed to address the challenges of renewable energy intermittency and carbon emission reduction. The system couples alkaline electrolysis, methanol synthesis, methanol reforming, gas turbine combined cycle, and carbon capture into a highly integrated energy conversion framework. To efficiently optimize this complex system under strict operating constraints, a novel staged optimization framework combining a neural-network surrogate model with the NSGA-II algorithm is developed. High-efficiency energy utilization is achieved by employing extraction steam for reforming and carbon capture, facilitating a carbon-closed cycle by recycling captured CO 2 for methanol synthesis. During the methanol synthesis system optimization phase, the NSGA-II algorithm identified optimal parameters that reduced the unit exergy cost and unit exergy carbon intensity of methanol by 3.5% and 2.01%, respectively, compared to the base case. For the methanol utilization system, a neural network surrogate model was introduced to achieve effective prediction of turbine inlet temperature and flue gas temperature constraints. This approach, validated against simulated values, eliminated over 90% of invalid operating cases and significantly accelerated the search process. Optimization results reveal that the unit exergy cost and unit exergy carbon intensity of power decreased by 1.4% and 6.2%, respectively. Ultimately, the system efficiency increased to 32.9%, reaching peak operational performance within the defined flue gas temperature constraints. The results demonstrate the feasibility of integrating neural-network-assisted optimization with carbon-closed energy systems and provide a scalable framework for future low-carbon energy system design.