CORE-LOG-GEOMODEL INTEGRATION: ADVANCED CLASSIFICATION AND PROPAGATION WORKFLOWS FOR THE CONSISTENT, RIGOROUS, AND PRACTICAL UPSCALING OF PETROPHYSICAL PROPERTIES

摘要

A consistent framework for reservoir characterisation is presented that focuses on the integration of core and wireline log data for the development of geocellular models. Concepts that clarify the characterisation problem are explained and the scales that are most important in developing a tractable solution are defined. Workflows are presented to show how both basic (static) and saturation-dependent (dynamic) petrophysical properties may be moved from one scale to the other. The similarities and the differences in the procedures at each scale are described. In addition, an explanation is given as to why a rigorous multi-scale approach involves far more than just 'upscaling', as is often assumed. Illustrations are given at various scales of what a comprehensive multi-scale approach entails. The need for five major reservoir characterisation steps at every scale change of the multi-scale characterisation workflow is outlined. These steps are: Classification, Selection, Evaluation, Propagation, and Upscaling (CSEPU). An explanation is provided as to why each step is vital in securing a robust suite of petrophysical properties at the succeedingly larger scale. A workflow is developed which is applicable at the core plug scale - where petrophysical properties are derived - and is then carried through the wireline log scale to the geomodelling scale (and potentially to the simulation scale). The difficulty of ensuring a rigorous reservoir characterisation at the 1D wireline log scale, especially of saturation-dependent properties, is explained and a consistent and robust solution is provided. The adopted workflow is based on several major components, the first of which is the adoption of the CSEPU concept to permit rigorous scale changes. Consistent Classification within CSEPU is provided by employing Bayesian-based Probabilistic Multivariate Clustering Analysis (PMVCA). Another important component is the assumption of an equivalent homogeneous medium at any scale. These prior concepts are then coupled with the principles of model-prototype hydraulic similitude to develop a unique, scale-independent parameterisation by combining key petrophysical properties into three Characteristic Length Variables (CLVs). These CLVs are then used in a PMVCA model, which is thus also scale-independent. Following the Upscaling of the basic petrophysical properties, the PMVCA is used to implement the Propagation (prediction and distribution) at the coarser scale of Upscaled saturation-dependent properties. An onshore, predominantly siliciclastic, conventional gas reservoir is used as a case study to illustrate the workflows. Examples of what constitutes effective Classification are provided and are used to illustrate why a flexible probabilistic (Bayesian-based) multivariate classification procedure is necessary at all scales and vital at those scales with sparse data. The critical step of the Propagation of the petrophysical properties derived at the fine scale into the coarse scale volume using the Bayesian probabilistic model is emphasised as it is essential to the success of the workflows. Finally, the use of, and results from, appropriate Upscaling at each scale are stressed. The workflows are comprehensive, consistent, and rigorous in their specification and implementation, but are also simple enough in their design and application to permit them to be embraced by all disciplines involved in the reservoir characterisation process.