页岩润湿性的神经网络预测模型

Prediction model of shale wettability based on neural network

页岩润湿性对油水微观分布以及井壁失稳等工程响应特征影响显著。文中实验表征了润湿性对页岩岩石组构及油基钻井液关键参数的响应特征,并利用广义回归神经网络(GRNN)方法开展了润湿性的定量表征研究。研究发现,页岩润湿性对岩石组构与钻井液性能呈现复杂的非线性响应特征。基于GRNN方法,建立的包含岩石组构与钻井液性能特征的润湿性定量表征模型具有较高的精度和泛化能力。其中,有机碳质量分数(TOC)对润湿性影响最大,其次是黏土矿物、石英质量分数和油基钻井液的油水比,岩石孔容特征影响最为微弱。基于GRNN方法的页岩润湿性定量表征模型,可实现页岩润湿性预测,从而为钻井液性能优化和井壁稳定控制提供指导。

Shale wettability has significant influence on the microscopic distribution of reservoir fluid and wellbore instability-. The response characteristics of wettability to shale tabric and oil-based drilling fluid parameters were analyzed by experiment. Then generalized regression neural network （GRNN） method was used to carry out the research on quantitative characterization model of shale wettability. The research indicates that shale wettability presents complex nonlinear response characteristics to rock tabric and drilling fluid performance; the quantitative characterization model of wettabiliU based on GRNN method has high accuracy and generalization ability-, which considers 12 influencing factors of rock tabric and drilling fluid performance. The sensitivity- analysis shows that the content of TOC has the greatest influence on wettability, tollowed by clay- minerals, quartz content and oil-water ratio of oil-based drilling fluid, and the pore volume characteristics of rock has the least influence on wettability; based on the GRNN method, the study on wettability quantitative model can be used to predict the wettability of rock and fluid system, so as to provide guidance tor drilling fluid performance optimization and wellbore stability control in shale drilling.