致密气储层弱反射特征的地震信号高分辨率相空间表征及其智能化解释

The Sulige tight gas field has become the top one of gas production in China. The reservoir characterization with 3D seismic data is significant for efficient development. However, the response of a tight reservoir is very weak in the non-stationary seismic data, because the difference between the impedance of the tight sand and shale is very small in this study area with complex surface and deep buried condition. Therefore, we will propose a novel approach including a phase-space representation and an assistant interpreted artificial intelligence. Firstly, analyzing the attenuation of seismic propagation in the tight gas reservoir, we build a basic function family in a multidimensional phase space representation with time, frequency and attenuation. This phase-space representation can enhance the weak features of tight gas reservoirs. Then, these features in the multidimensional phase-space can be interpreted using the prior knowledge of sedimentology and transfer learning theory. Finally, we apply the proposed method on the seismic data to describe the three-dimensional space distribution of tight gas sandstones, which helps to guide the horizontal well placement and economical gas recovery in the Ordos Basin.

苏里格致密气田是我国第一大气田，其高效勘探开发的一个关键是利用三维地震数据实现精细的储层描述。然而，由于致密储层和围岩间的物性差异小，加之复杂地表和埋藏深，导致非平稳地震信号中有效储层的响应特征非常弱，很难直接揭示储层分布。因此，本项目拟从非平稳地震信号的时频特征分析出发，在高维相空间中寻找弱储层反射特征的最佳多分辨率表示，再结合沉积背景约束和新一代机器学习辅助，智能化的给出定量地质解释。首先，研究含气储层中地震子波等效衰减过程的表示，尝试构建出由时间、频率和衰减量等组成的更高分辨率的、更高维的多参数相空间表示，实现对弱储层反射结构的高维提取。其次在多维相空间中，基于迁移学习理论，设计和训练符合先验沉积知识的深度学习网络结构，实现从高维特征上解释出储层空间展布及物性变化。最后，在靶区应用，得到致密甜点的空间分布，优化水平井部署。

内容获取失败，请点击重试