滩坝砂储集体测井产能等级划分与地震属性横向预测

滩坝砂储层因其储层物性差、单层厚度薄等特点而导致产能预测较为困难,提出了一种新的基于数据和模型双驱动的井震结合产能等级划分方法。通过机器学习中的降维算法结合聚类算法,并利用聚类算法中的肘部法确定最佳聚类数,进行储层测井产能等级自动划分。首次在平面径向流公式的基础上建立了测井产能等级指示模型,为地震属性横向预测提供井点刻度。以胜利油田W工区为例,利用最小二乘法推导出了测井产能等级指数公式,计算了10口井目的层单井产能等级指数。对提取的多种地震属性在井点处的属性值与测井产能等级分类结果进行Pearson相关性分析,优选了3种与产能等级指数显著相关且两两之间相互独立的地震属性。利用支持向量回归算法,建立了地震属性融合的储层产能等级指数平面图,经3口验证井检验,产能等级指数平面图与实际产能吻合程度较好。研究结果表明,基于测井产能等级划分的地震属性横向预测方法可以有效预测靶区储层产能等级。

 Predicting the productivity of beach-bar sandstone reservoirs is difficult owing to their poor physical properties and the low thickness of the individual layers.In this study,a new productivity grading method based on well logging and seismic data was proposed.Machine-learning dimensionality reduction and clustering algorithms were used to automatically evaluate the logging productivity grading.The optimal clustering number was determined using the elbow rule.For the first time,by using the plane radial flow formula,a model for the logging productivity grading index was established.This model can be used to provide a good one-point calibration for the lateral prediction of seismic attributes.Taking the W well area of the Shengli Oilfield as an example,a practical formula for the logging productivity grading index was derived using the least squares method.The single-well target layer logging productivity grading indices of 10 wells were calculated.Pearsons correlation analysis between the various attribute values extracted at the well point and the logging productivity classification results was conducted.Finally,three seismic attributes,which are significantly related to the productivity grading index and are independent of each other,were selected.Using the support vector regression algorithm,a plane graph of the reservoir productivity grading index combined with seismic attributes was generated.The plane graph of the reservoir productivity grading index was in good agreement with the actual production capacity according to data from three verification wells.The results of this study showed that the combined lateral prediction of seismic attributes based on the logging productivity grading can be effective in predicting the reservoir productivity grade in the target area

国家自然科学基金项目(41874138)和国家重大科技专项(2016ZX05006002-004)共同资助。