This paper gives details of an oil field reservoired in a turbidite sandstone fan in the North Sea. The majority of the field is characterised by a channelised, high density turbidite system of high net-to-gross sandstones. Associated with these sand-rich channel fairways are lower net-to-gross fairway margin and interchannel areas. A review of the conventional geological mapping originally used for the subject field has shown that this heterogeneity is not captured in the current deterministic geological model. The objective of the 3D modelling was to address this complexity early in the field life.
A hierarchical geostatistical model has been developed to create detailed 3D reservoir models for the reservoir interval. Initially the reservoir has been divided deterministically into the three different domains of channel axis, channel margin and interchannel by using seismic attribute data. Sequential indicator simulation was used to distribute reservoir facies within each domain. The reservoir facies were characterised using core, wireline log and sedimentological data. During model construction, each domain is conditioned to adjacent domains to ensure geological sensible continuity across any interchannel to channel margin to channel axis transect. Vertical variogram models were derived from well log and core data. Outcrop data were used to constrain the required variogram anisotropy for the horizontal variogram models. Sequential Gaussian simulation was used to populate the facies description with appropriate porosity and permeability values.
Following generation of the 3-D geological model, numerical pressure transient analysis was used to validate the reservoir model. By iteration of model parameters, the pressure transient description was honoured for a number of wells. Consequently, integration of pressure transient data has created a more robust reservoir model, and has highlighted unresolved uncertainties.
Stochastic reservoir models are an important tool for assessing the impact of geological heterogeneity on the performance and recovery from hydrocarbon reservoirs. Many techniques have been developed for generating geostatistical reservoir models; object-based methods, variogram-based sequential simulation algorithms, and Markov statistics to name some of the methods currently in the literature. Many reservoir models are constructed using a hierarchy of scale from genetic unit, through lithofacies to petrophysical characteristics, utilising the most appropriate technique at each level.
In the North Sea, much work has been done on the application of object-based methods to fluvial reservoirs. Several factors have led to this state of affairs; for example, sedimentary facies models of fluvial environments and the geometries of fluvial bodies are relatively well understood, and the importance of fluvial formations in the North Sea (e.g. Statfjord Formation, Ness Formation).
Modelling of deep-water clastic reservoirs is much less common, despite the importance of such reservoirs in the North Sea. This can be attributed in part to a lack of a clear understanding of the processes involved in deposition of deep water clastic reservoirs. In addition, the database of geometrical statistics for turbidite facies is limited, although this situation is improving.