Corradi A, Ruffo P, Corrao A, Visentin C. 

Modelling hydrocarbon migration in complex depositional environments is a difficult task. This paper illustrates a workflow that has been developed in order to make use of information at seismic resolution for migration modelling purposes. After a regional identification of the main seismic sequence boundaries and systems tracts, the first step of the workflow consists in a semi-automatic classification of the sand and shale prone seismic facies that are interpretable in the seismic volume. A software based on supervised neural network techniques and textural seismic attributes supports volumetric classification. Each class represents a different facies with specific petrophysical properties (threshold pressure, porosity, irreducible water saturation, etc.), assigned according to concurrent sedimentological studies; the result is a 3D lithofacies model. Petroleum system modelling is performed on a 3D regional model up to the calculation of hydrocarbon quantities expelled from the source rock. Then migration modelling is performed on the smaller area covered by the 3D lithofacies model. Invasion percolation approach has been chosen because it allows to perform migration modelling at a scale that is small enough to maintain a sufficient detail of the channels' shape and of their connections. A probabilistic approach allows to estimate charge risk in new prospects and to check modelling consistency. This workflow has been applied in a deep water setting in West Africa, characterized by an interconnected channel system, where a number of prospects had to be evaluated. A well was drilled in the area prior to this study, resulting dry in spite of being located in a position where the source rock is present and is deemed to have expelled interesting amounts of hydrocarbons. This study has shown that hydrocarbons expelled from the source appear to be deviated laterally by the presence of a local seal and therefore do not reach the channel system in the well location.

 

Note:

常见的油气运移算法有：达西流，流线，混合法。流线法不是严格意义的三维运移，应该是2(1/2)维的油气运移。达西流真正实现了三维运移，但是其CPU耗费时间过长直接影响这种算法在高精度地质模型中的应用，如地震相体。混合运移是一种折中的方法，即获得了较接近三维的运移效果，也缩短的CPU耗时。本文所介绍的方法，首先利用神经网络将地震体划分为不同岩性的体积模型，再利用渗流技术（Invasion percolation,IP）进行三维油气运移。方法的优点在于可以保持地质体足够的精度（细节），进行油气运移模拟的快速运算。