Improved understanding of reaction kinetic identification problems using different nonlinear optimization algorithms

The correct answer regarding which nonlinear optimization algorithm should we use for a given problem is that “it depends.” In this paper, we would like to add that “it depends, but use multiple programs whenever possible.” Here we consider 23 algorithms, implemented in MATLAB, evaluating their performance both on a lumped kinetic model for vacuum gas oil hydrocracking and a few-step kinetic model for ethane pyrolysis; the former particularly raised our interest as the kinetic parameters have no reference values in such models. We can use the results of such a study to estimate model variance; moreover, the statistical analysis of the identified minimum values can also quantify the parameter uncertainty. We can also identify key operating conditions where the applied kinetic model shows the highest sensitivity to the identified parameters, opening up the possibility to further reduce the uncertainty by targeting additional experimental work or by refining the identification problem.