Prediction of Desulfurization from Diesel of Isfahan Refinery by Membrane Method Using Intelligent Approach

In this study, modeling and investigation of desulfurization of diesel Isfahan Oil Refinery Company were performed using membrane process. The total sulfur of diesel as feed to membrane module is 6380ppm. In the research, four kinds of membrane including Polydimethylsiloxane, blend of Polydimethylsiloxane with Polyethylene glycol, blend of Polydimethylsiloxane with Polyethersulfone and blend of Polydimethylsiloxane with Polyacrylonitrile are used. The process variables in this research are pump pressure of membrane module (5-9 bar), crosslinking agent concentration (1.5 and 3 wt. %), crosslinking temperature (65-85°ċ), and crosslinking time (0.5-2.5 h). In the modeling procedure, Artificial Neural Network (ANN) and genetic programming (GP) were employed. The Levenberg-Marquardt training algorithm was used to train the ANN. ANN architecture with 6 neurons was determined as optimal architecture. ANN and GP are beneficial tools for predicting the performance of RO with high accuracy (R2=0.93 and 0.89 respectively).