A Permeability Model Considering Stress Concentration At Pore Edge Based On Machine Learning Model

It is of great significance to study the mechanism of permeability stress sensitivity for oil and gas production prediction.From the thin section images of reservoir rock,it can be seen that the rock matrix is not completely separated by pores.According to the theory of elasticity,there is a phenomenon of stress concentration around the pores of rock,but conventional permeability models don’t consider the effect of this phenomenon.Some factors closely related to the degree of stress concentration are ignored.In order to clarify the effect of stress concentration at pore edge on permeability stress sensitivity,This thesis mainly studies the response law of permeability change to stress concentration at the pore edge,and based on this,a permeability model with higher prediction accuracy and better applicability to reservoirs is established.In this thesis,a numerical model of pore deformation is established based on the microstructure of real reservoir rock to verify the existence of stress concentration at pore edge and analyze the effect of this phenomenon on permeability stress sensitivity.The machine learning method is used to establish a permeability model considering stress concentration at pore edge.,and numerical simulation data and experimental data are used to verify its prediction accuracy;the new model is used to analyze the influence of main controlling factors of stress concentration such as pore geometric characteristics and stress direction on permeability change.The results show that stress concentration at pore edge significantly affects permeability stress sensitivity.The degree of stress concentration under elliptical pore is higher,the corresponding permeability change is 4 times that of circular pore.The stress concentration caused by the stress perpendicular to the long axis of pore is higher.,and the corresponding permeability change is 2 times that of the stress parallel to the long axis.The prediction error of the machine learning permeability model is reduced by up to 54% compared with the conventional permeability model,and the new model considers the influence of complex pore geometry,different stresses and stress direction,and is more suitable for actual reservoirs.According to pore shape complexity(C),the permeability stress sensitivity under different pore shapes can be divided into three stages: when C is less than 0.6,the stress sensitivities under different pore shapes are weak and similar,and the permeability changes are similar to that of circular pores;When C is greater than 0.6,the stress sensitivities under different pore shapes are significantly improved and there are significant differences;when C is greater than 0.9,the stress sensitivities under different pore shapes are strong and similar,and the permeability changes are close to that of elliptical pores.In addition,the effect of stress direction is also significant,and its effect increases with the increase of C.When C is greater than 0.9,the permeability change induced by stress direction can reach more than 30%.