Facies modeling described by probabilistic patterns using Multi-point statistics an application to the k-field, Libya

Abstract

The main objective of this study is to illustrate the stochastic modeling of facies bodies using a multi-point statistical methodology (MPS) based on probability patterns derived from the integration of facies interpretation at well locations, seismic data, and the conceptual geological model. The proposed methodology was applied to real data in the K-Field of Libya. The reservoir unit, the Mamuniyat Formation of Ordovician age, is interpreted as a glacially influenced setting ranging from shoreface to base of slope fan-channel environments. The input data was classified in two main groups: 1) hard data corresponding to facies description, derived from core, conventional logs, and especially image logs (FMI), from wellbores, and 2) soft data used as conditioning, derived from a scaled conceptual model constructed on the basis of seismic horizon-slice interpretation and analogues. The final results illustrate the benefits of using multi-point statistics for facies distribution in complex settings. The resulting model reproduces the facies heterogeneity and, to a significant degree, the conceptual model, leading to the conclusion that multi-point statistics offer a significantly improved representation of the geological heterogeneity. reservoir heterogeneities whilst modeling facies distributions. In order to deal with these issues, it is important to use stochastic methods for modeling, using, for example, soft data conditioned to hard data. Typically, facies modeling of a reservoir is carried out using abundant hard data (cores and logs), but where such data is limited in scope or quality, the task becomes more difficult, as typically occurs in the exploration phase. In such cases, additional soft data is useful as a constraint on facies distribution, based on conceptual models, regional geology, analogues, and paleogeography, etc. Both kinds of input data, hard and soft, are essential for geological modeling and contribute to understanding the conceptual depositional model, notably in terms of the shape, size, and orientation of the different sediment bodies. Needless to say, the output geological model should always honor all input data. Whereas outcrop models and seismic interpretation provide important information on reservoir architecture, heterogeneity, and spatial distribution, it is not fully clear how such information can be used in geostatistical reservoir modelling. Multi-point simulation (MPS) methods allow the combination of conceptual models in training images (Strebelle and Cavelius, 2013). After creation of the training image, MPS generates patterns, then anchors them to subsurface well logs and seismic. In this way, complex and heterogeneous facies can be distributed in a reliable and more realistic model (Caers and Zhang, 2004).