基于改进支持向量机的致密砂岩储层参数预测研究

Research on tight sandstone reservoir parameter prediction based on improved support vector machine

致密砂岩储层的评价技术既是油气勘探开发的重点,也是难点。目前对致密砂岩储层的储层参数的预测与评价,依然采用传统的储层参数预测方法,结合测井曲线进行建模,用以对渗透率、孔隙度等参数进行拟合,主要运用的方法有经验公式、回归分析等,其中大部分方法都是基于线性的,无法反映致密储层特有的沉积和成岩作用所导致的非均质性强的特点,无法揭示致密储层中测井曲线与储层参数之间的复杂非线性关系。针对此问题,提出在传统储层参数预测模型的基础上,对测井曲线与储层参数的非线性关系进行分析,挖掘更多现有测井信息,进行支持向量机储层参数预测模型的建构,并采用粒子群算法、头脑风暴算法、布谷鸟算法等三种支持向量机的改进优化算法对模型参数进行测试,筛选出最优的储层参数预测模型。将该模型应用于研究区储层参数预测评价中,有效提高了预测评价精度,为致密储层精细预测评价和非常规油气田的高效开发提供了有力的技术保障。

The evaluation technology of tight sandstone reservoir is not only the focus but also the difficulty of oil and gas exploration and development.At present,the traditional methods are still adopted in the prediction and evaluation of reservoir parameters of tight sandstone reservoir.In these methods,the modeling is carried out in combination with the well logging curves,so as to fit parameters such as permeability and porosity.The main methods used are empirical formulas and regression analysis.Most of these methods are based on linearity,which fails to reflect the strong heterogeneity caused by the unique sedimentation and diagenesis of tight reservoirs and fails to reveal the complex nonlinear relationship between well logging curves and reservoir parameters in tight reservoirs.In view of the above,on the basis of the traditional reservoir parameter prediction model,the nonlinear relationship between well logging curves and reservoir parameters is analyzed and the existing well logging information is more fully explored to construct a reservoir parameter prediction model based on support vector machine(SVM).The model parameters are tested with three improved optimization algorithms of SVM,including particle swarm optimization(PSO),brainstorming algorithm and cuckoo search(CS)algorithm,so as to select the optimal reservoir parameter prediction model.The model improves the accuracy of prediction and evaluation effectively when it is applied to the prediction and evaluation of the parameters of the reservoir in the study area.Therefore,the proposed model can provide strong technical support for fine prediction and evaluation of tight reservoirs and efficient development of unconventional oil and gas fields.