Optimal Design of an Integrated Radiant Syngas Cooler and Steam Methane Reformer using NLP and Meta-heuristic Algorithms

In this study, optimal designs for a novel integrated radiant syngas cooler and steam methane reformer are explored. Previously, feasible base-case designs were established but were sub-optimal given the complexity of the model that included more than 200,000 equations. The two optimization approaches used in this study include: (i) conventional non-linear programming solvers (within gPROMS) and (ii) meta-heuristic techniques (Differential Evolution and Particle Swarm Optimization). The model for the integrated device was implemented in gPROMS, and hence the built-in NLP solver was used. However, the NLP solver can only guarantee local optimality, and finding good initial guesses manually can be prohibitively time consuming. Therefore, the results were compared with meta-heuristic methods that are easy to implement, are parallelizable and can cover a wide search space. The results using both methods showed significant improvement in capital cost, as much as 40% and with improved methane conversion. The advantage of parallel computing when using meta-heuristic techniques for optimizing multi-scale models was clear: for example, the CPU time reduced by up to 50% when twice the number of cores was utilised to simulate the multi-scale model in gPROMS.