论文部分内容阅读
数理统计方法和地质理论结合已逐渐成为研究复杂储层的方向。由于致密砂岩储层的复杂性,其孔隙度、渗透率与测井资料不只是简单的线性关系,常用的线性回归方法难以满足致密砂岩储层物性研究的精度要求。选取致密砂岩储层研究区内3口取心井511个岩心分析样品,首先运用贝叶斯判别法将储层砂岩分为三类:石英砂岩、岩屑砂岩和岩屑石英砂岩,然后采用多元逐步回归、主成分分析和支持向量机方法对不同类型砂岩分别进行建模,对比各方法得出的复相关系数,发现支持向量机回归得到的复相关系数明显高于其他两种方法 ,且支持向量机回归方法得到的预测值与原始值的平均绝对误差也是最小的。支持向量机模型效果检验结果表明,孔隙度绝对误差小于1.5%,渗透率绝对误差小于0.25×10~(-3)μm~2,说明该模型预测效果较好,适用于研究区致密砂岩储层物性参数建模。
The combination of mathematical statistics and geological theory has gradually become the direction of the study of complex reservoirs. Because of the complexity of tight sandstone reservoirs, its porosity and permeability are not only simple linear correlations with logging data. The commonly used linear regression methods are difficult to meet the precision requirements of the physical properties of tight sandstone reservoirs. In this paper, 511 core samples of 3 coring wells in the tight sandstone reservoir are selected. First, Bayesian discriminant method is used to classify the reservoir sandstone into three types: quartz sandstone, lithic sandstone and lithic quartzite sandstone, Stepwise regression, principal component analysis and support vector machine method were used to model different types of sandstone. Comparing the complex correlation coefficients obtained from each method, it was found that the complex correlation coefficient obtained by support vector regression was significantly higher than the other two methods and supported The mean absolute error between the predicted value and the original value obtained by the vector machine regression method is also the smallest. The result of support vector machine model test shows that the absolute error of porosity is less than 1.5%, and the absolute error of permeability is less than 0.25 × 10 ~ (-3) μm ~ 2, indicating that this model has good prediction effect and is suitable for the study of tight sandstone reservoirs Physical parameter modeling.