粒子群优化支持向量机算法在页岩储层总有机碳含量预测中的应用——以渝西地区Z井区为例

页岩总有机碳(TOC)含量是反映页岩生烃潜力及页岩气富集程度的关键参数之一。四川盆地渝西地区Z井区的页岩气勘探,存在钻探程度低、地质资料少和岩心样品实验分析数据不全等不利因素,难以利用测井资料得到精确的总有机碳含量曲线。而通过地震正演、地震多属性反演和叠前反演等方法预测的总有机碳含量又存在精度较低的问题。为了实现总有机碳含量精细预测,提出了一种基于粒子群(PSO)优化支持向量机(SVM)算法的页岩气总有机碳含量计算方法。首先,根据总有机碳含量与测井资料的交会关系,确定自然伽马、密度和纵横波速度比等与计算总有机碳含量相关的敏感测井参数,利用支持向量机和粒子群优化算法的方法原理,建立适用于研究区与计算总有机碳含量相应的粒子群优化支持向量机算法;其次,采用该算法计算总有机碳含量测井曲线并与岩心总有机碳含量数据进行对比以修正算法的预测精度;最后,在常规地震反演数据体基础上,利用粒子群优化支持向量机算法计算出储层总有机碳含量数据体,进而开展页岩总有机碳含量有利勘探区的预测与页岩气储层的评价。研究结果表明,粒子群优化支持向量机算法预测总有机碳含量曲线与岩心实测总有机碳含量较为吻合,误差较小,同时,通过地震资料预测总有机碳含量的结果与测井解释的总有机碳含量结果对应较好。表明在非均质性较强的页岩气储层中,利用粒子群优化支持向量机算法进行总有机碳含量预测,可以有效提高页岩气储层总有机碳含量的预测精度,对四川盆地渝西地区页岩气勘探开发具有一定的指导意义。

The total organic carbon (TOC) content is a key parameter for shale gas reservoirs as it can inform on their hydrocarbon generation potential and degree of shale gas enrichment.Shale gas exploration in Well Z in the Yuxi area of the Sichuan Basin is challenging owing to limited drilling and,consequently,incomplete geological and core sample data.Consequently,obtaining accurate TOC content curves from logging data is difficult.Moreover,the accuracy of TOC content predictions from seismic forward modeling,seismic multi-attribute inversion,and pre-stack inversion is relatively low.To overcome the aforementioned issues,a method of TOC content calculation based on the particle swarm optimization (PSO) and support vector machine (SVM) algorithms has been proposed.As a first step,core logging parameters relevant to TOC determination were evaluated,namely:natural gamma,density,and P-wave-to-S-wave velocity ratio.Then,a PSO-SVM algorithm was established,by which TOC logging curves were estimated.These curves were compared with those evaluated from core logging data to verify the algorithm’s prediction accuracy.Finally,using conventional seismic inversion data as an input,the PSO-SVM algorithm was utilized to predict the TOC content of the entire reservoir,so as to identify the most promising exploration areas.The results showed that the proposed method could provide TOC content estimates that were consistent (within a relatively small margin of error) with those obtained from core logging data.Moreover,the in-line and planar distributions of TOC content values predicted from seismic data were consistent with the well-side analysis.It was concluded that the proposed method can effectively improve the TOC content prediction accuracy in shale gas reservoirs with strong heterogeneity (such as in the Yuxi area of the Sichuan Basin),thereby aiding in the exploration and evaluation of their development potential.

国家科技重大专项(2017ZX05035003)资助。