Automatic Reservoir Model Identification using Artificial Neural Networks in Pressure Transient Analysis

Abstract

Abstract Pressure transient well test analysis is commonly used to help characterize oil and gas reservoirs. In a typical well test, interpretation of rate/pressure data usually yield information about permeability, skin factor, boundary conditions, and character of the reservoir. Some well tests, however, suffer from ambiguity and non-unique interpretation. The objective of this study is to apply the Artificial Neural Network (ANN) technology to identify the reservoir model. A multilayer neural network, with back propagation optimization algorithm, is used to identify the reservoir model. The required training and test datasets were generated by using the analytical solutions of commonly used reservoir models. Nine ANN networks were constructed with each one capable of differentiating among six boundary models. Most commonly found reservoir models of different inner, outer boundary and reservoir medium are included (e.g. vertical, fractured and horizontal wells; homogenous, dual porosity and radial composite reservoirs; and infinite, one sealing fault, two sealing faults, rectangle and circle boundaries). Each of the ANN of the nine networks has been constructed by one input layer, two hidden layers; and one output layer with six nodes characterizing the different reservoir boundary models. Different network structures and training intensity were tested during this work to arrive at optimum network design. The performance of the proposed ANN was tested against simulated noisy and smooth datasets. The results indicate that the proposed multilayer neural network can recognize the reservoir models with acceptable accuracy even with complex models. The comprehensive testing of different ANN designs showed that success rate increases significantly by distributing the commonly used reservoir models into nine networks. This ANN design can still yield good results even with some noise in the pressure data. After testing the ANN, they were then used in the interpretation of several field cases (including complex tests) and results are presented in the paper.