致密裂隙介质波致流机理及物性甜点检测关键方法研究

Bearing rich resources, tight oil and gas reservoirs in our country are the most feasible unconventional oil and gas fields for exploration and production. Due to the complicated lithology and pore throat structure existed in tight reservoir, it is necessary to perform fundamental research on the wave induced fluid flow discipline and the innovation of petrophysical sweet spot detection methods for them. This is of great significance for guiding the fracturing design and capacity construction in tight reservoirs.. This subject is going to study the attenuation and dispersion regularity in the full frequency band of wave, which is induced by fluid exchange between cracks in different scales. The quasi-static equivalent model in the seismic frequency band will be constructed, and the variation pattern between the snail gather and the fractural parameters of tight reservoir will be deduced. Dual transformation methods based on multi- dictionary learning will be researched to perform dual sparse transformation for snail gathers. Deep learning algorithms based on hybrid convolution and inner product operators in tight support sets will be studied to create robust data driven detection methods for petrophysical and fractural properties of tight reservoirs. The recognition resolution of petrophysical sweet spots will break through the limitation of seismic sampling rate. This subject will provide theoretical support and method innovation for the effective product of tight reservoir.

我国致密油气藏资源量丰富，是最可行的非常规油气勘探开发领域。由于致密储层岩性组合和孔喉结构复杂，针对致密砂砾岩特殊地质特征，开展致密裂隙储层波致流基础研究和物性甜点检测方法创新，对指导致密储层压裂设计、规模有效建产具有重要意义。. 本项目拟综合研究不同尺度下裂隙之间流体交换引起的波在整个频带上的衰减频散规律，构建地震频带内的准静态等效模型，进而分析其地震波场的动力学和运动学特征，推演地震蜗牛道集与致密储层可压裂性及物性参数之间的变化规律；研究多字典学习对偶稀疏变换方法，实现蜗牛道集的紧支集对偶变换；研究紧支集内基于褶积、内积混合算子的机器深度学习算法，构建稳健的、数据驱动的致密储层物性与可压裂性检测方法，替代反演方法，物性甜点的识别精度突破地震采样率限制，为致密储层有效开发提供理论支持和关键方法创新。

本研究针对致密裂隙介质频散衰减规律不系统、物性甜点预测精度低等难题，开展了致密裂隙介质波致流机理及地震波运动学动力学理论研究，并以此为基础建立了创新性的致密储层物性甜点检测方法。. 首先，基于多尺度波致流理论，通过构建符合实际致密裂隙储层主要特征的准静态各向异性等效模型，研究了地震波传播的运动学动力学特征，建立了地震蜗牛道集与致密储层物性参数之间的理论关系；在此基础上，以对偶稀疏表示理论为基础，提出了基于多字典学习的对偶稀疏变换方法，实现了蜗牛道集的紧支集对偶变换和规则化重建，并利用紧支集地震数据开展了深度学习致密储层甜点检测方法研究，有针对性地创建了适用于致密储层物性甜点预测的褶积+内积混合算子深度学习结构。该深度学习结构能自动利用测井数据的高频信息，得到高分辨致密甜点物性参数预测结果，突破了地震数据采样率限制，并适用小样本的条件，促进了甜点识别由模型反演型到数据驱动型的转变。相关研究成果在地球物理学报、Geophysics、Petroleum Science等本学科权威期刊发表SCI文章14篇，授权国家发明专利5项，并在鄂尔多斯盆地致密气、四川盆地深层气等实际工区得到推广应用，为地质甜点的高质、高效勘探提供了理论支撑和关键方法创新。

内容获取失败，请点击重试