A NEURAL-NETWORK PREDICTION MODEL OF FLUID DISPLACEMENTS IN POROUS-MEDIA

This paper presents the development and design of an artificial neural network that is able to predict the breakthrough oil recovery of immi scible displacement of oil by water in a two-dimensional vertical cros s section. The data used in training the neural network was obtained f rom the results of fine-mesh numerical simulations. Several network ar chitectures were investigated and trained using the back propagation w ith momentum algorithm. The neural network that gave the best predicti ve performance was a two-hidden layer network with 8 neurons in the fi rst hidden layer and 8 neurons in the second hidden layer. This networ k also performed well against a cross validation test. The reservoir s imulation data used so far in the training process was for a homogeneo us reservoir, the more general case is still under investigation.