岩相骨架的地质统计模拟

Geostatistical modeling for lithofacies architecture.

储集层非均质性是碎屑岩储集层描述中的重要问题。与流体性质分布的不确定性相比，岩相骨架的不确定性对流体预测影响更大。近十年来，地质统计方法大量地应用于岩相骨架的模拟中，其模拟方法主要有离散及连续两类。用数学语言对４个主要的离散模拟方法（序贯指示模拟、截断高斯模拟、模拟退火及示性点过程模拟）进行概述，展示其工作过程，讨论各自的优缺点，特别强调如何合并次生数据以及在现场应用的可能性。此外，还提出了该领域未来的研究方向。

Reservoir heterogeneity is an important issue in most clastic reservoir characterization studies. Uncertainty in lithofacies architecture is often considered to have a larger impact on flow predictions, compared to uncertainties in the distribution of lithofaciesspecific hydraulic properties. Hence, generation of detailed geological model is the first priority in building reservoir simulation model. It provides and understanding of the spatial distribution of various lithofacies and their connectivity. The geological model then serves as a guidance for subsequent assignment of rock and fluid properties within each lithofacies. Geostatistical methods are increasingly popular in simulating facies architecture in the past decades. These methods can generate realistic threedimensional geological models. They also give multiple or equallyprobable reservoir images which honour the available data and possess the nearlyidentical spatial correlation structure. Hence, the impact of geological uncertainty on flow prediction can be evaluated. Geostatistical methods can be classified into two main categories: discrete and continuous modeling methods. The former deals with nonnumerical (binary) data, and the latter deals with numerical (continuous) data. The objective of this paper is to present the stateoftheart review of four major discrete modeling models or methods which are appropriate for building facies models: sequential indicator simulation (conventional and Bayesian); truncated Gaussian field; simulated annealing; marked point processes. This paper presents the above simulation methods in a simple mathematical language which attempts to provide an insight on how the algorithms work. The pros and cons of each technique will also be discussed with a special emphasis on how secondary data (such as seismic attributes) can be incorporated and availability of field applications. A future research direction and a comprehensive reference list will be provided.