Estimating Asphaltene precipitation in the presence of co2 injection in oil reservoirs

Abstract

In this research, use of multi layer perceptron (MLP) and radial basis function (RBF) structures of artificial neural network (ANN) for prediction of asphaltene precipitation were described and the models were contrasted with the modified Hirschberg et al., model. The essential data were gathered and after pre-treating was employed for training of ANN models. The performance of the best obtained model was checked by its generalization ability in predicting 30% of the unseen data. Excellent prediction with Mean Squared Error (MSE) of 0.0018 and Average Absolute Deviation (AAD %) of 1.4108 was observed. However the accuracies of RBF and MLP models may be evaluated relatively similar, it was obtained that the constructed MLP according to Levenberg-Marquardt (LM) optimization exhibited a high performance than RBF structure, and the modified Hirschberg to predict asphaltene precipitation.