多源井震数据驱动下渝西地区Y区块页岩裂缝预测研究

Multi-Source Well-Seismic Data Driven Shale Fracture Prediction in Block Y, Western Chongqing

摘要: 渝西地区Y区块五峰组–龙马溪组深层页岩受多期构造挤压改造，天然裂缝空间非均质性强，裂缝发育等级是控制页岩气保存条件与储层压裂改造效果的关键地质因素。传统单一地震属性方法难以有效区分裂缝发育等级，而岩心与成像测井仅能获取井点信息，无法实现工区尺度裂缝展布的精细刻画。针对渝西深层页岩工区样本类别不均衡的现实数据条件，本文开展面向该特定地质背景的深度学习裂缝预测工作流程优化与验证研究；融合叠前、叠后地震及成像测井多源井震数据，将研究区裂缝划分为无裂缝、弱裂缝与强裂缝三个发育等级；以成像测井解释结果为标定依据，建立分类评价指标体系，多维度校验井震融合预测方案的可靠性，并剖析裂缝发育主控因素。结果表明，五折交叉验证与留井验证证实模型泛化性能稳定，预测结果与地质认识吻合度高。裂缝发育受断裂体系与构造部位双重控制，强裂缝沿主干断裂带呈条带状集中分布，背斜核部及构造转折带裂缝发育程度显著增高。研究成果可为Y区块水平井轨迹优化及压裂漏失风险预判提供地质依据。

Abstract: The deep shale of the Wufeng-Longmaxi Formation in Block Y, western Chongqing, has undergone multi-stage tectonic compression and reworking. The spatial heterogeneity of natural fractures is strong, and the fracture development grade is a key geological factor controlling shale gas preservation conditions and reservoir fracturing stimulation performance. However, conventional single seismic attribute approaches are insufficient to hierarchically characterize complex fracture development, and core and borehole image logging data are limited to well-point scale, failing to achieve continuous and fine fracture distribution prediction across the entire study area. Targeting the class‑imbalanced dataset under this specific deep shale geological setting, this paper optimizes and validates a deep‑learning‑based fracture prediction workflow. To address this technical bottleneck, this study integrates pre-stack seismic, post-stack seismic and borehole image logging data to construct a multi-source well-seismic fusion prediction framework. According to logging-calibrated fracture characteristics, natural fractures are classified into three grades: non-fracture, weak fracture and strong fracture. A multi-index evaluation system is established to quantitatively constrain fracture development, and the reliability and generalization ability of the prediction model are systematically verified. The five-fold cross-validation and well-leave-out validation demonstrate that the proposed method achieves stable performance and high consistency with regional geological laws. Fracture development is jointly controlled by fault systems and structural positions. Zones of intense fracturing are distributed in bands along major fault zones, and fracture intensity increases markedly in anticlinal cores and structural transition zones. This study provides a geological basis for horizontal well trajectory optimization and fracturing fluid loss risk prediction in Block Y.