Generative artificial intelligence-based energy process design: the case study of post-combustion carbon capture

Post-combustion carbon capture is a critical technology for mitigating CO2 emissions, but its widespread adoption is hindered by high energy consumption. This study addresses this problem by introducing a novel generative artificial intelligence-based design paradigm to rapidly identify optimal, energy-efficient process configurations. The primary scientific novelty lies in integrating a flexible Artificial Neural Network surrogate model directly within a gradient-based optimization framework. This method overcomes the speed and constraint-handling limitations of traditional derivative-free optimization approaches and offers greater adaptability than previous gradient-based methods that rely on less versatile surrogate models. Applied to an amine-based carbon capture process, the framework successfully minimized the specific reboiler duty by optimizing design and operational variables. The results provide quantitative insights into key engineering trade-offs. For example, doubling the absorber packing height from 12 to 24 m consistently reduces the specific reboiler duty by 0.6–0.7 MJ/kg CO2, whereas increasing capture efficiency from 85 % to 95 % raises it by 0.5–0.7 MJ/kg CO2. Furthermore, optimizing the lean-rich heat exchanger can yield additional energy savings of up to 0.3 MJ/kg CO2. The study concludes that this generative design approach effectively balances capital and operational costs which leads to achieving optimal specific reboiler duty values between 3.5 and 4.3 MJ/kg CO2 across various flue gas CO2 compositions. This fast and robust methodology provides a replicable blueprint for optimizing not only carbon capture systems but also other energy-intensive industrial processes.