秦皇岛32 ̄6油田北区三维相控地质统计建模与反演

2015年 22卷 第03期
阅读:134
查看详情
3D geo-statistical modeling and inversion constrained by geo-facies in north area
of QHD32-6 Oilfield
闵小刚 康安 周守信
中海石油研究中心,北京 100027 中海石油伊拉克有限公司,北京 100010
Research Center of CNOOC Ltd., Beijing 100027, China CNOOC Iraq Limited, Beijing 100010, China
 秦皇岛32 ̄6油田主要含油层系是被断层复杂化了的明化镇组下段河流相储层,由于薄夹层规模小,砂体薄厚极不稳定且横向变化大,井间可对比性差。常规确定性反演方法受地震频带限制,对目的层段出现频次最多的3~5 m砂体无法准确表征,造成钻前、钻后砂体的变化比较大。为了提高河流相储层预测的分辨率和精度,文中发展应用了相控地质统计学反演方法。具体实现过程是,首先反演出高分辨率的砂泥岩性模型,将它与基于目标的方法模拟出的沉积微相模型合并,然后在新的相模型约束下,按照双重变差函数分析思想,实现对薄储层物性的高精度反演。实际应用表明,井点预测砂体厚度误差较小,井间预测砂体连通、接触关系符合地震趋势和地质统计规律。由此得到的阻抗、优势相和物性三维地质模型,其垂向网格间距可达到1 ms,既能在平面上很好地表现地质认识,又能在纵向上合理划分出隔、夹层。
 The main oil-bearing series in QHD32-6 Oilfield is fluvial reservoir of lower Minghuazhen Formation which is destroyed by faults. Considering the small-scale of thin inter-bed, the instability of sand-body thickness, and the lateral variation of reservoir, the stratigraphic correlation between wells is very difficult. Limited by the seismic effective frequency band, the ordinary inversion method based on CSSI is difficult in recognizing the major sand bodies with 3-5m thickness in target layer, these lead to the large change of sand body numbers after drilling. However the geo-statistics inversion method based on Markov chain Monte-Carlo algorithm (MCMC) is more powerful in improving resolution and accuracy of reservoir prediction. The specific method is: invert some high-resolution lithology realizations using MCMC geo-statistic inversion firstly, and model the channel and non-channel sedimentary microfacies using object-based methods simultaneously, then merge these two models; under the controlling of the new facies model, in accordance with the concept of double-variogram analysis, three physical property models which can predict thinner reservoir more accurately have been inverted by using the MCMC algorithm again. The practical application at north area of QHD32-6 Oilfield shows that the thickness errors between predicted and the actual at every well are less than 2 m, and reservoir property between wells follows the geo-statistical rules. The P-impedance, dominant reservoir facies and physical properties geo-model inverted with 1 ms grid interval in vertical can not only deliver the geology knowledge in lateral but also separate the main sand and shale barrier-bed/inter-bed clearly, these have broken through the deadline of ordinary seismic resolution.
明化镇组; 河流相储层; MCMC算法; 变差函数; 相控地质统计学反演;
Minghuazhen Formation; fluvial reservoir; MCMC algorithm; variogram; facies controlled geo-statistic inversion
;
10.6056/dkyqt201503012