论文部分内容阅读
日本有一个万亿立方英尺的气田已持续生产了数十年。储积层由海下喷发的流纹熔岩叠加穹丘组成。每个穹丘发育孔隙网络,碎玻质熔岩遇海水迅速冷却并在穹丘上沉积。虽然该油田的碎玻质熔岩也是多孔隙的,但由于有泥质矿物及不渗透玄武薄层和泥岩层,其渗透率大幅下降。压力体系中的每个相都有其特定的作用。地层对比最初可识别多储积层。几乎毫不例外都是从最大的气田采气的,但经过10～20年的生产,人们发现未开发储积层的压力呈现出不同速度的递减,而且局部呈现出不同寻常的特征。地震资料是否可提供大量信息无以为证,因而,决定专门利用压力资料对整套体系进行重新模拟。采用多点地质统计和概率扰动理论相结合的方法。这个方法,在利用历史拟合阐述压力数据、进一步解释真实油田的非平稳性时,可成功捕获叠加熔岩穹丘弯曲相的边界。一般来说,多点法的一个难点是建立适宜的训练图形。针对火山岩油藏建立训练图形会遇到许多特殊的问题,文中,通过反复调整模型、获得有大量合理扰动并令人满意的历史拟合实现了图形的建立。通过植入预先确定的模型空间,用产层和非产层的象素可随机描绘油藏的模糊边界。模拟结果不仅精细模拟了测得的压力变化曲线,显示相分布和油藏边界的真实情况,表明相连通的弯流通道造成了不同单元间的不均匀压降,而且,还解释了观察到的小范围的压力异常变化。最终获得的训练图形表明相的空间变化比先前判定的更强。20个等概率值估算的原始气地质储量(OGIP)分散范围不超过平均值的±15%。经评估钻一口扩边井的采收率增量要比安装一台增压器的增量变化要大,这可对地质相关的风险性进行量化。
Japan has one trillion cubic feet of gas fields that have been in production for decades. The reservoir consists of ripple lava superimposed domes erupted offshore. Porous networks develop in each dome, where the crushed lava rapidly cools in seawater and deposits on domes. Although the fractured glassy lava is also porous in the field, its permeability drops dramatically due to the presence of argillaceous minerals and impermeable basaltic thin beds and mudstones. Each phase in a stress system has its own specific role. Stratigraphic correlation initially identifies multiple reservoirs. Almost without exception gas is collected from the largest field, but after 10 to 20 years of production, it has been found that the pressure of unexplored reservoirs shows a decreasing rate at different rates and presents some unusual features locally. Seismic data can provide a lot of information is not evidence, therefore, decided to use stress data to re-simulation of the entire system. Using multi-point geostatistics and probability perturbation theory combined method. This method successfully captures the boundary of the curved phase of superimposed lava domes by using history fitting to describe pressure data and further explaining the nonstationarity of real oil fields. In general, one of the difficulties of the multi-point method is the establishment of suitable training graphics. There are many special problems encountered in establishing training graphics for volcanic reservoirs. In this paper, a series of historical perturbations with a large number of reasonable perturbations are obtained by repeated adjustment of the model to achieve the establishment of graphics. By embedding a predetermined model space, the fuzzy boundaries of a reservoir can be depicted randomly using the pixels of the pay zone and the pay zone. The results of the simulations not only simulate the measured pressure curves, but also show the realities of the phase distribution and the reservoir boundaries, indicating that the connected flow channels cause an inhomogeneous pressure drop across different cells, and that the observed Small range of abnormal pressure changes. The resulting training pattern shows that the spatial variation of the phase is stronger than previously determined. The estimated distribution of OGIPs for 20 equi-probabilities does not exceed ± 15% of the mean. It has been assessed that the incremental recovery of a multilateral well is larger than that of a turbocharger, which quantifies the geologic related risks.