基于无监督学习的多参数储层评价:以蒲包山地区下三叠统飞仙关组礁滩储层为例

Multiparameter reservoir evaluation method based on unsupervised learning:A case study of the reef beach reservoir of the Lower Triassic Feixianguan Formation in the Pubaoshan area

四川盆地东北部蒲包山地区下三叠统飞仙关组礁滩储层的形成与发育是地质历史时期综合作用的结果,因此储层评价时仅使用单一因素难免会产生偏差。采用k-means聚类分析和主成分分析相结合的方法对研究区储层进行了分类与评价。研究结果表明:在已知蒲包山地区下三叠统飞仙关组礁滩储层的白云岩厚度、平均孔隙度以及储层有效厚度3种不同影响因素平面图的前提下,对不同平面进行等大网格化,提取不同影响因素的储层特征数据,结合使用肘部法和轮廓法对储层特征数据进行分析并将储层划分成4种发育类型,然后应用k-means聚类分析方法对已知的数据点进行类别属性分配;使用主成分分析对不同储层特征数据进行降维处理形成一个新的综合参数,参数贡献率达0.882,按k-means的分类结果,计算4种储层不同类型主成分分析综合参数的均值,分别为0.404,0.640,0.716,作为储层评价分区的分界点。最终使用此种量化的方法将研究区不同特征平面图合理融合在一起,形成储层综合评价图。研究结果可有效地对研究区储层进行分类评价及有利勘探区预测。

[Objective]The formation and development of the reef-shoal reservoirs in the Lower Triassic Feixianguan Formation in the Pobaoshan area are the result of the comprehensive action of the geological historical period.Therefore,only using a single factor in reservoir evaluation will inevitably lead to deviations.[Methods]The k-means cluster analysis method and principal component analysis method were used to classify and evaluate the reservoir in the study area.[Results]The results show that:On the premise that three different influencing factors of dolomite thickness,average porosity and effective reservoir thickness of the Lower Triassic Feixianguan Formation reef-shoal reservoir in the Pubaoshan area are known,gridding different planes to extract reservoir characteristic data of different influencing factors.The combined elbow method and contour method are used to analyze reservoir characteristic data and divide the reservoir into 4 development types.Then k-means cluster analysis method is applied to assign class attributes to the known data points.Using principal component analysis to reduce the dimensionality of different reservoir characteristic data to form a new comprehensive parameter.The parameter contribution rate can reach 0.882.According to the classification results of k-means,the mean values of the comprehensive parameters of different types of principal component analysis of the four reservoirs were calculated,which were 0.404,0.640 and 0.716,respectively,as the demarcation point of the reservoir evaluation zone.Finally,this quantitative method is used to reasonably integrate the different characteristic plans of the study area to form a comprehensive evaluation map of the reservoir.[Conclusion]The research results can effectively classify and evaluate the reservoir in the study area and predict favorable exploration areas.