基于精细高斯支持向量机的岩相识别模型——以B区块为例

Lithofacies identification model based on fine Gaussian support vector machine--Take block B as an example

B井区长8储层为该地区的主要资源潜力层系,通过其录井资料可以发现该地区主要以泥岩,细砂岩,其主要储层类型为细砂岩,其单砂体的精细刻画对于后期沉积微相特征研究,单砂体特征以及后续的勘探开发工作开展具有一定的重要意义,对于该区块在实际生产过程中具备生产井较多、工作量大等特点,因此高效精准的开展单砂体刻画工作是油田生产试油选层亟待解决的问题。传统的单砂体刻画方法仅需要自然伽马和自然电位两条曲线通过对其曲线形态进行判断,从而对目的层段的单砂体进行识别,其划分效率低,准确性一般,其识别效果从效率到精度上都需要得到进一步提升,因此本文提出一种基于支持向量机模型的岩相识别法,根据前人经验结合研究区实际地质情况选取相适应的特征向量作为其支持向量机模型的输入层对岩相进行预测,分别采用一次支持向量机,二次支持向量机进行建模,并最终以高斯核作为核函数的精细高斯支持向量机对模型进行优化。该方法区别于传统方法对于数据样本数量以及特征维度的需求,分类样本不局限于非线性识别。从而高效地完成单砂体刻画,有效的解决了研究区砂体识别效率低的问题,对于后期开展储层特征研究具有一定的实际意义。

B well area long eight reservoir for the region's main resource potential layer system,through its logging data can be found that the region is mainly mudstone,fine sandstone,its main reservoir type is fine sandstone,the fine engraving of the single sand body for the later study of the sedimentary microphase characteristics,the single sand body characteristics as well as the subsequent exploration and development work is of some significance to the actual production of this block in the production of wells with a large number of characteristics such as large workload,so the efficient and accurate implementation of the single sand body engraving is an urgent problem of oil field production test oil selection,In the actual production process of this block,there are many production wells and a large workload,so the efficient and accurate single sand body engraving is an urgent problem to be solved in the oilfield production test oil layer selection.The traditional method of single sand body engraving only needs two curves of natural gamma and natural potential to identify the single sand body of the target layer by judging its curve morphology,which is low in division efficiency and general in accuracy,and its identification effect needs to be further improved from efficiency to accuracy.Therefore,this paper proposes a lithofacies identification method based on a support vector machine model,which is based on the experience of the previous work combined with the actual geological situation of the study area to select the appropriate feature vector as the input layer of its support vector machine model to predict the lithofacies.Based on the previous experience and the actual geological situation of the study area,we select the appropriate feature vectors as the input layer of the support vector machine model to predict the rock phases,and use primary support vector machine,secondary support vector machine for modelling,and finally use Gaussian kernel as the kernel function of the fine Gaussian support vector machine to optimize the model.The method distinguishes itself from traditional methods in terms of the number of data samples as well as the need for feature dimensionality,and the classification samples are not limited to non-linear identification.Thus,the single sand body engraving is efficiently completed,which effectively solves the problem of low efficiency of sand body identification in the study area,and has certain practical significance for the later development of reservoir characterization research.