多元地震属性分析提高薄储层预测精度

Multivariate seismic attribute analysis improve prediction accuracy of thin reservoirs

针对叠后地震属性预测薄储层信息时存在属性单一、多解性及定性表征等问题,文中提出了基于主成分分析-支持向量机算法(PCA-SVR)对多种地震属性进行定量表征预测薄储层的方法。以松辽盆地某井区葡I4层为例,利用主成分分析法对9种地震属性进行降维分析,优选累计贡献率大于90%的3种主成分,利用支持向量回归机算法,建立了薄储层预测模型。对井区内葡I4层砂岩厚度进行预测,预测结果与实际钻遇砂岩厚度吻合程度较好。研究结果表明,预测结果与实际钻遇砂岩厚度的相对误差,与目的层段内砂岩发育层数相关,随着发育砂岩的层数增多,平均相对误差逐渐增大。

Aiming at single attribute,multi-solution and qualitative representation of post-stack seismic attributes when extracting seismic attributes in thin reservoirs,a new method was proposed to quantitatively analyze and predict the thin reservoir thinkness based on the principal component analysis-support vector machine algorithm(PCA-SVR).Taking the PI4 Formation of a well block in Songliao Basin as an example,the principal component analysis method was used to analyze the nine seismic attributes,and the first three principal components with cumulative contribution rate greater than 90%were selected,and then the support vector regression algorithm was used to establish the thin reservoir prediction model.The sandstone thickness of PI4 Formation was predicted and the prediction result was in good agreement with the actual drilling thickness.The results show that the relative error between the prediction result and the actual reservoir thickness was related to the number of sandstone development layers in the objective interval.With the increase of the number of sandstone development layers,the average relative error increased gradually.