基于支持向量回归机的地层孔隙压力预测方法

Pore pressure evaluation method based on support vector machines for regression

（1.中国石油化工股份有限公司胜利油田分公司钻井工艺研究院信息中心，山东东营257017； 2.中国石油大学信息与控制工程学院，山东东营257061）

The Information Center of Drilling Technology Research Institute, Shengli Oil Field Company Ltd., Dongying 257017, China

测井资料是确定地层孔隙压力的基础性资料，为此，利用测井资料来研究准噶尔盆地某区块的地层孔隙压力，开展地层孔隙压力区域研究，以充分认识异常地层孔隙压力分布规律。在分析地层孔隙压力预测传统方法局限性的基础上，提出了一种基于有效应力定理和声波速度模型的地层孔隙压力预测方法。由相关测井资料计算泥质含量、孔隙度和声波速度，利用声波速度模型计算垂直有效应力，利用密度测井资料计算上覆岩层压力，最后根据有效应力定理计算地层孔隙压力。声波速度模型由支持向量回归机（SVR）通过对相关测井、测压资料的非线性回归得到。实际应用表明，该方法能够以较高精度预测到异常地层孔隙压力，为钻井工程设计提供依据，提高钻井工艺水平，对钻井过程中防止工程事故发生，减少地层污染，节省钻井成本有着重要的应用价值。

Well logging data is the basic data for pore pressure evaluation. The well logging data from an area in Junggar basin was used to study the pore pressure. Through regional study of the pore pressure, abnormal high pressure distribution was recognized. After analyzing the limitation of the traditional pore pressure prediction methods, a new method based on the effective pressure theorem and the acoustic velocity model was proposed. The clay content, permeability and acoustic velocity were calculated from related well logging data and the vertical effective stress was computed from sonic velocity model. In addition, the overburden pressure was also calculated from density logging data. Finally, the pore pressure was calculated according to the effective pressure theorem. The acoustic velocity model was built by regression support vector machine through nonlinear regression of the related logging data and the pressure data. The actual application indicates that the method can predict abnormal pore pressure accurately, which provides foundations for drilling engineering design and improves the level of drilling technology. What is more, it has a significant application value in preventing engineering accidents, decreasing stratum pollution, and saving drilling cost.