Applying Artificial Neural Networks to Permeability Determination

The main aim of this project was to establish permeability model using artificial neural networks (ANN). This methodology enables us to create an effective permeability model in uncored wells. First we train the network using representative data in wells with permeability core samples and then we apply trained network to uncored wells. Few approaches and tests were performed in the project including tests of cycle number, amount of hidden layers and excluding some input data. Data used in this project come from the Rotliegend formation from the Polish Lowland (P1 well, Cz2 well, Z1 well). All calculations were done using K.mod in Techlog® (Schlumberger).