Support Vector Machine based modeling of an industrial natural gas sweetening plant

Abstract

In this study Support Vector Machine (SVM) is employed to develop a model to estimate process output variables of an industrial natural gas sweetening plant. The developed model is evaluated by process operating data of Hashemi Nejad natural gas refinery in Khorasan/Iran. A set of 13 input/output plant data each consisting of 145 data has been used to train, optimize, and test the model. Model development that consists of training, optimization and test was performed using randomly selected 80%, 10%, and 10% of available data respectively. Model estimations are compared with those obtained from an ANN based model developed using the same dataset as used for training and test of SVM based model. Test results from the SVM based model showed to be in better agreement with operating plant data compared to artificial neural networks based model. The minimum calculated squared correlation coefficient for estimated process variables is 0.99. Based on the results of this case study SVM proved that it can be a reliable accurate estimation method.