Advancing multi-generation energy system optimization through artificial intelligence-driven configuration optimization: a novel framework for simultaneous component selection and parameter adjustment

In most renewable energy studies, optimization has been limited to adjusting the operating parameters (e.g. mass flow rate, pinch point temperature) of predetermined components, a strategy known as parametric optimization. Although useful, this practice narrows the design domain by excluding alternative technologies that might achieve superior outcomes. To address this gap, the present research proposes a formal two-stage configurational optimization framework for comparing and ranking fundamentally different system architectures. Unlike integrated co-design, which is suited for interchangeable components within a single model, our hierarchical approach is designed for high-level architectural selection where systems have unique, incompatible models. The methodology is applied to a multi-generation plant in Bandar Abbas, Iran, where two types of hydrogen electrolysis namely, proton exchange membrane (PEM) and alkaline electrolysis and two types of solar collectors namely, parabolic trough (PTSC) and parabolic dish (PDSC) are compared. The configurational optimization encompasses four possible configurations. The findings indicate that configurational optimization identifies the PTSC–alkaline setup as the optimal configuration, producing 18,390.576 kg/year of liquid hydrogen with an exergy efficiency of 19.10% and a cost rate of 304.97 $/Day. The remaining configurations, ranked by the configurational optimization in order of preference, are: PDSC–alkaline, PDSC–PEM, and PTSC–PEM.