苏里格地区三水模型与数学方法结合的致密砂岩储层评价  

Tight sandstone reservoir evaluation by the combination of three-water model and mathematical method in Sulige Area

作　　者：郭宇航[1] 潘保芝[1] 蒋必辞[1] 刘思慧[1] 房春慧[1] 李丁[1] 

机构地区：[1]吉林大学地球探测科学与技术学院,吉林长春130026

出　　处：《石油物探》2015年第5期621-626,共6页Geophysical Prospecting For Petroleum

基　　金：国家自然科学基金项目(41174096);国家科技重大专项项目(2011ZX05009;2011ZX05044)联合资助

摘　　要：应用三水模型评价苏里格地区致密砂岩储层时,在孔隙度较高的层段,孔隙度和含水饱和度预测结果与岩心数据符合度很好;在孔隙度较低的层段,孔隙度预测结果符合度较好,但含水饱和度预测结果存在很大偏差。造成这一现象的原因是孔隙度较低的层段岩性更加致密,孔隙结构更加复杂,三水模型中的参数难以赋值。为此,提出三水模型与数学方法结合的致密砂岩储层评价方法,通过全区密闭取心资料分析确定三水模型处理下限,在下限之下的层段结合广义回归神经网络(GRNN)和粒子群-支持向量机(PSO-SVM)算法得到处理结果。三水模型结合数学方法在苏里格地区综合处理的结果与该区岩心数据符合度较好,说明方法是可行的。When three-water model was applied to evaluate tight sandstone reservoir of Sulige Area,the prediction results of the layers with high-porosity fit well with the core data; however, in the layers with low-porosity, the porosity prediction re- sult is good while there is a big deviation in the water saturation prediction result. The phenomenon is caused by the more tight pores and more complex porous structures in the layers with lower porosity, and the parameters of three-water model are difficult for evaluation. Through sealed core data analysis, the lower limits of the parameters of the three-model is identified. Combined with the Generalized Regression Neural Network （GRNN） and Partial Swarm Optimization-Support Vector Machine （PSO-SVM） algorithm, the layers below the lower limits is predicted by mathematical method. The prediction re- sults obtained by the combination of three-water model and mathematical method is coinciding well with the core data, which provides good criteria for the logging evaluation of the tight sandstone reservoirs in Sulige Area.

关 键 词：致密砂岩 含水饱和度 三水模型 粒子群-支持向量机算法(PSO-SVM) 广义回归神经网络(GRNN) 

分 类 号：P631[天文地球—地质矿产勘探]