Reliable reservoir performance forecasts with as little uncertainty as possible are key information for optimal reservoir management tasks. These production forecasts are directly related to the reservoir size and internal porous media properties. There is a need for improved techniques for production data integration to construct realistic reservoir models by using geostatistical techniques.

A methodology is proposed that integrates production data into reservoir models by the local updating of porosity and permeability fields. The focus is on conditioning a proposed initial model to injection/production rate and pressure history in an iterative fashion. For each pass, a perturbation location is selected and master point locations are defined and used as reference to calculate the pressure and flow rate sensitivity coefficients subject to changes in porosity and permeability. The optimal changes of porosity and permeability at the master point locations are propagated to the whole grid by kriging. Integrating flow simulation and kriging algorithms within an optimization process constitutes the proposed methodology.

This method makes it possible to condition the permeability/porosity distributions to injection/production rate and pressure history data from large reservoirs with complex heterogeneities and changes of well system with time. A field case application demonstrates that the proposed methodology is efficient and practical for large reservoir models.


Many people are working on production data integration and several methods have been proposed. However, there is a challenge to condition reservoir property models to production data for large scale fields accounting for realistic field conditions. Direct calculation schemes are avoided considering that they are often limited to 2-D single-phase flow. Stochastic approaches such as simulated annealing or genetic algorithms require a lot of simulation runs, making them practically unfeasible for large scale application.(1,)(2)(3) Algorithms and software for production data integration based on hydrogeological developments such as sequential self calibration and pilot point methods have not proven applicable in complex reservoir settings with multiphase flow, 3-D structure and changing well conditions.(4) Streamline simulation based methods suffer the same limitations although some papers show that it has been used in large reservoirs.(5) The convergence of results for gradual deformation methods is very slow so that lots of iterations are needed for large 3-D models (6)(7).

All the production data integration methods relay on flow simulation. Streamline simulation is commonly proposed as a solution to be used in large 3-D reservoir models due to its computational efficiency and analytical sensitivity coefficient calculation. However, the simplification of streamline simulation may promote computational efficiency at the expenses of accuracy reduction in cases of high heterogeneity with multiphase flow, 3-D structure and changing well conditions.

There is a need for a novel computational efficient production data integration method that can be used in large complex 3-D reservoir models with many wells and long production and injection history.

Basic Idea and General Procedure of the Proposed Methodology

Our basic idea consists on the numerical calculation of the sensitivity coefficients on the basis of two flow simulations - an initial base case and a single sensitivity case.

This content is only available via PDF.
You can access this article if you purchase or spend a download.