Determination of Reservoir Model from Well Test Data, Using an Artificial Neural Network

Nowadays, neural networks have a wide range of usage in different fields of engineering. In the present work, this method is used to determine a reservoir model. Model identification, followed by parameter estimation, is a kind of visual process. Pressure derivative curves showing more features are usually used to determine the reservoir model based on the shape of the curve and no calculation is included. So, it is difficult to convert this kind of visual process to an applicable algorithm for computers. In fact, the model identification is a pattern recognition which is best done by an Artificial Neural Network (ANN). If neural networks were learned successfully, they would be able to categorize different shapes into different groups, due to their visual characterization. So, their use in such a job would seem to be useful. In this work, it is shown how to train, examine and use neural networks to determine a reservoir model. The input of an ANN is fifty points of the normalized pressure derivative type curve. Each ANN is trained, based on a specific model, and the output of the ANN is the probability of occurrence of a fed curve to the related model