مرکز منطقه ای اطلاع رساني علوم و فناوري -

چكيده لاتين :

Determination of the effective porosity and water saturation of a hydrocarbon reservoir plays a very important role in the petroleum upstream industry, especially in the economic success of reservoir development and this task is yet encountered with many technical and economical problems. Artificial neural network (ANN) is one of the latest technique s for modeling and simulation that can also be used for prediction of petrophysical parameters. In this study, two separate "error back-propagation" ANNs are used for prediction of effective porosity and water saturation from just wireline logs data. The effective porosity ANN is a three-laye r network using sonic,density, neutron porosity, gamma ray and LLD (resistivity log in uninvaded zone or deep resistivity log) logs as input with 8 neuron s in hidden layer and "Sigmoid" activation function for both hidden layer and output layer. The water saturat ion ANN is also a three-layer network using sonic, density , gamma ray, MSFL (resistivity log in flushed zone), LLS (resistivity log in transition zone or shallow resistivity log) and LLD logs as input with 10 neurons in hidden layer and "Hyperbolic Tangent" activation function for hidden layer and "Sigmoid" activation function for output layer. This study was performed on Sarvak Formation, an important oil reservoir, in the Zagro s Basin, southern Iran and the results show that effective porosity and water saturation can be estimated with high accuracy using ANN.