Investigating Performance and Reliability of Stochastic Methods for Optimization of Low Temperature Gas Separation Processes

The synthesis and optimization of low-temperature gas separation processes is complex owing to the large number of design options. In this paper, we have verified the performance of two popular stochastic methods, Genetic Algorithm and Simulated Annealing, in optimization of sub-ambient systems. GA has been applied to many problem areas, but evidently not previously to low-temperature problems. Despite the feasibility of GA optimizer in low-temperature processes, which has recently been addressed, here we have studied the quality of GA answers in this subject. We designed a very large experimental data set to systematically explore a range of parameter settings in genetic algorithm and simulated annealing and afterwards we have investigated the potential of achieving global optima. In other words, the optimization task is not only to converge in a feasible region, but also to give the best qualitative solution. Having identified the GA and SA parameters and also optimized the solutions in three different case studies, we observed that: (1) SA is more robust and reliable than GA when applying to low temperature gas separation processes. (2) By adjusting the key parameters in SA method, the optimization process will avoid pre-mature convergence and will be able to give the best near-global results. 1Introduction