摘要：

渤海湾盆地济阳坳陷沾化凹陷埕岛—桩海地区侏罗系三台组近年来成为油气勘探重点层系,但受构造演化复杂、沉积类型多样、岩性多变的影响,储集层非均质性极强,储集层质量预测难度大,而成岩相类型的精确识别与划分对储集层的评价起至关重要的作用。传统利用有监督学习识别全井成岩相的方法在学习样本数目较少的情况下实用性有限,本研究开展基于单因素约束的无监督全井成岩相测井识别方法研究。结合视压实率、胶结物含量、面孔率和裂缝率将研究区成岩相划分为致密压实相、碳酸盐胶结相、溶蚀与裂缝相和不稳定成分弱溶蚀相。确定了GR(自然伽马)、AC(声波时差)、DEN(密度)、RD(深侧向电阻率)4条对成岩作用敏感的测井曲线作为成岩相测井识别的依据。分别对4条曲线聚类范围进行约束,确保了无监督学习方法与成岩相良好的映射关系。基于无监督学习划分的测井相种类,利用铸体薄片等资料实现测井相与成岩相的标定,完成地区成岩相的识别与划分。划分结果表明,溶蚀与裂缝相、不稳定成分弱溶蚀相为研究区储集层有利成岩相,2类成岩相对应较高的孔渗度,溶蚀与裂缝主要分布在含砾砂岩等粗粒岩中,不稳定成分弱溶蚀相主要分布在粉砂岩等细粒岩中。通过盲井成岩相对比,验证学习方法的准确性,进而为缺乏取心井段的储集层成岩相识别提出新的预测方法,对有利储集层的评价及预测具有一定意义。

关键词: 济阳坳陷, 沾化凹陷, 三台组, 成岩相测井识别, 单因素约束, 无监督学习

Abstract:

The intricate tectonic evolution,coupled with diverse sedimentary environments and highly variable lithologies,results in strongly heterogeneous reservoirs in the Santai Formation of the Zhanhua sag,making reservoir quality prediction particularly challenging. This highlights the critical importance of accurately identifying and classifying diagenetic facies for effective reservoir evaluation. Traditional supervised learning methods for whole-well diagenetic facies identification are limited by the availability of training samples and thus often impractical in data-scarce scenarios. To address this,an unsupervised learning approach constrained by single-factor analysis is proposed for the logging-based identification of diagenetic facies. By integrating parameters such as apparent compaction rate,cement content,porosity,and fracture density,four diagenetic facies were identified: dense compaction facies,carbonate-cemented facies,dissolution-fracture facies,and weakly dissolved facies with unstable components. Four diagenesis-sensitive logging curves—GR(natural gamma),AC(acoustic travel time),DEN(density),and RD(deep lateral resistivity)—were selected as inputs for the unsupervised clustering algorithm. The clustering ranges of these curves were individually constrained to establish a reliable correlation between logging responses and diagenetic facies. Calibration of logging facies with diagenetic facies was conducted using core data,including thin-section analysis of cast samples,allowing for regional-scale identification and classification of diagenetic facies. The results show that the dissolution-fracture facies and weakly dissolved facies with unstable components represent the most favorable diagenetic facies for reservoir development. These are associated with relatively high porosity and permeability and are mainly distributed in coarse-grained rocks such as conglomeratic sandstones and,while the weakly dissolved facies with unstable components are mainly distributed in fine-grained rocks like siltstone. The accuracy of the proposed method was validated through comparisons with blind wells,demonstrating its effectiveness in non-cored intervals. This approach provides a novel method for predicting reservoir diagenetic facies and offers practical implications for the evaluation and prediction of high-quality reservoir zones in data-limited settings.

Key words: Jiyang Depression, Zhanhua sag, Santai Formation, diagenetic facies logging identification, single factor constraints, unsupervised learning

中图分类号: 

• TE122.2