Studying the Influence of Parallelization on the Performance of Evolutionary Algorithms When Solving an Optimal Control Problem of Hydrogenation of Hydrocarbons

This paper studies the gains and losses of parallelization of evolutionary algorithms. It is known that the computing system’s parallel architecture should be taken into account to successfully parallelize an evolutionary optimization algorithm. However, it can lead to a modification of the sequential version of the algorithm, which results in poorer yet faster performance. We study an optimal control problem for a complex chemical reaction, namely, the hydrogenation of polycyclic aromatic hydrocarbons in the presence of catalysts. This optimal control problem was transformed into a global nonlinear optimization problem, which was solved by the Mind Evolutionary Computation (MEC) algorithm. We propose a few parallel modifications of the MEC algorithm which utilize different task parallelization techniques and algorithm parallelization approaches. Those modifications were used to solve the optimal control problem of hydrogenation of hydrocarbons in the presence of the nickel-kieselguhr catalyst. For the efficiency analysis, several metrics were used to evaluate a solution’s speed and quality. The results of all numerical experiments are presented in the paper with the obtained optimal control for the chemical reaction under investigation.