Mohammad Sadegh momeni - Department of Petroleum engineering, Omidiyeh branch, Islamic Azad University Omidiyeh, Iran Author’s address:
nasser teymourei khanesary - Corresponding Author’s Gas engineering, Petroleum University of Technology-Ahwaz
alireza khoshroo - University of Yasouj, Yasouj, Iran

Miscible gas injection is one of the most effective methods in enhancing oil recovery.The most important Parameter in the design of such processes is minimum miscibility pressure (MMP).MMP is the lowest pressure that can be injected through multi contact with reservoir fluid to be miscible. MMP can be obtained from Rising bubble apparatus, s lim tube and mixing cell in laboratory that these methods are costly and time consuming. A multilayer perceptron neural networks with 6 neuron of input layer, 12 neuron of hidden layer and a neuron of output layer was trained for estimating the MMP. The coefficent of determination (R2) between experimental MMP and predicted MMP using neural networksfor the training data, testing data and all data were 0.988, 0.954 and 0.970 respectively. The results states neural networks has higher accuracy that Alston method in predicting MMP.

MINImum miscibility pressure MMP,artificial neural network,miscible gas injection