Comparison of Modeling and Optimization of Methanol to Propylene (MTP) Over High Silica H-ZSM-5 NANO Catalyst Using Black-Box Modeling (ANN) and Meta-heuristic Optimizers (GA-PSO)

چکیده مقاله

The methanol to propylene (MTP) process is one of the most important methods of propylene production. The operating conditions of the process are of great importance as far as the product yield is concerned and can be detected by the means of modeling and optimization. The modeling done in this paper is based on black-box approach and in order to optimize the operation conditions the meta-heuristic algorithms have been developed. Two types of artificial neural network (ANN) known as multilayer perceptron (MLP) and radial basis function (RBF) were used in this black-box modeling and the results of optimization were compared. The artificial neural network includes two inputs, which are the reaction temperature (between 623.15 and −823.15 K) and the space–time of the feed-in reactor (between 0 and −0.25 h) along with one output that is the weight fraction of propylene. The correlation coefficient of multilayer perceptron (MLP) and radial basis function (RBF) of the modeling were obtained to be 0.99909 and 0.99703, respectively. To achieve proper optimization, genetic algorithm (GA) and particle swarm optimization (PSO) were applied. The genetic algorithm and the particle swarm optimization optimized the modeling in eight and three forms, respectively, and both of them validated the obtained results.