基于机器学习的岩芯渗透率及裂缝开度预测

Prediction Method of Core Permeability and Fracture Aperture Based on Machine Learning

应力敏感是致密砂岩气藏损害的主要原因之一,预测应力敏感损害下岩芯渗透率和裂缝开度的变化规律一直是致密砂岩储层保护领域的重点。以塔里木盆地克拉苏构造带岩样为研究对象,基于应力敏感实验及调研数据,采用机器学习多元线性回归算法,耦合了围压渗透率关系模型和K−p函数参数预测模型,建立了岩芯渗透率预测模型和裂缝开度预测模型,并通过决定系数、均方根误差和相对误差检验模型精度。结果表明,围压渗透率关系模型在裂缝性岩芯和非裂缝性岩芯中预测结果决定系数平均值均大于0.960;K−p函数参数预测模型在裂缝性岩芯中的均方根误差高于非裂缝性岩芯,但裂缝性岩芯的相对误差要低于非裂缝性岩芯,综合来看岩芯渗透率预测模型更适用于裂缝性岩芯;裂缝开度预测模型与实测值决定系数0.978,预测精度较高。建立的渗透率预测模型和裂缝开度预测模型可为致密砂岩储层的开采与保护提供指导。

Stress sensitivity is one of the main reasons for the damage of tight sandstone gas reservoirs.The prediction of the change law of core permeability and fracture aperture under stress sensitivity damage is always the key point in the field of tight sandstone reservoir protection.Based on the stress-sensitive experiment and survey data,the core permeability prediction model and fracture opening prediction model were established by using machine learning multiple linear regression algorithm coupled with the confining pressure permeability relationship model and the K−p function parameter prediction model.The accuracy of the model was tested by correlation coefficient,root mean square error and relative error.The results show that the average correlation coefficient of the prediction results of the confining pressure permeability model in fractured and non-fractured cores is greater than 0.96.The prediction results of K−p function parameter prediction model show that the root mean square error in fractured cores is higher than that in non-fractured cores,but the relative error of fractured cores is lower than that of non-fractured cores.It shows that the permeability prediction model is more suitable for fractured cores.The coefficient of determination between the fracture opening prediction model and the measured value is 0.978,indicating a high prediction accuracy.The permeability prediction model and fracture aperture prediction model can provide guidance for the exploitation and protection of tight sandstone reservoir.