A pore network model reconstruction method via genetic algorithm

Abstract

A pore network model could represent the porous medium space to predict rocks' various flow properties. This paper proposes a new method via genetic algorithm (GA) to reconstruct the pore network model with irregular distribution and topologically equivalent based on the real porous medium. Firstly, the properties of pore and throat are based on the Weilbull distribution function. A random pore network model is built by some assumptions and Weilbell parameters. In order to obtain the reliable parameters of pore and throat, the Avizo Fire analysis software is used to analysis the sample's micro-computed tomography (micro-CT) images. The theory of pore network simulation also is introduced simply in this section. Second, two objective functions for GA are given. We build up an objective function to reduce the discrepancy of slices of numerical model with the real micro-CT images, which is controlled by the core porosity. The second objective function makes model much more agreed with the experiment value. Thirdly, according to the characteristics of control parameters, we choose three steps alternating optimization to construct the model via GA. It demonstrates that this method has a quick convergence speed than before experiments. We adopt two kinds of complex geological porous medium to test the validity of our proposed method. One is a rock with high porosity (Sample One). The other is the dense sandstone (Sample Two). After the optimization, two appropriate pore network models are performed excellent of the prediction displacement in primary oil flooding of the target porous medium.