摘要
该研究以泸州地区五峰组—龙马溪组已钻井的测井资料、岩心测试数据和全岩衍射数据为基础，通过数据集人工智能训练、多参数优选和多元非线性拟合，探讨了计算矿物体积分数的敏感参数，利用常规测井GR，KTH，CNL，AC，DEN等5种数据建立了伊利石、石英、方解石和黄铁矿等矿物体积分数的预测解释模型。其中：伊利石与方解石体积分数与KTH，DEN，CNL相关性强；石英体积分数与CNL，KTH，AC相关性强；黄铁矿体积分数与GR，CNL，AC相关性强。对比发现，利用多元非线性矿物解释模型解释的伊利石、方解石、石英和黄铁矿等4种矿物体积分数与元素测井矿物体积分数吻合度好，标准偏差和残差均小于10%。在该模型基础上进行岩性的识别，可以解决深水陆棚沉积储层复杂岩相划分的问题，为识别沉积微相储层划分提供了快捷高效的方法，对于页岩气开发具有重要意义。
Abstract
Based on well logging data, core test data and whole rock diffraction data of drilled wells in Longmaxi-Wufeng Formation in Luzhou Area in this paper, the sensitive parameters that affect minerals content are discussed by AI training of data set multi-parameter optimization and multivariate nonlinear fitting, and prediction models of minerals content including illite, quartz, calcite and pyrite are established by using conventional logging GR, KTH, CNL, AC and DEN curves. Among them, the illite and calcite content show strong correlation with KTH, DEN and CNL curves, the quartz content shows strong correlation with CNL, KTH and AC curves, and the pyrite content shows strong correlation with GR, CNL and AC curves. It is found by comparison that illite, calcite, quartz and pyrite contents interpreted by conventional logging are consistent with that of element captured logging, and the relative error of each mineral content is less than 10%. The identification of lithology based on the models can be used to solve the classification of complex lithologies in deep-water shelf sedimentary reservoirs, and provides a fast and efficient method for division of sedimentary microfacies, which is of great significance for shale gas development.
论文详情