Reconciling high-resolution geologic models to production history is a very time-consuming process in reservoir modeling. Current practice still involves a tedious history-matching process that is highly subjective and often employs ad-hoc property multipliers. Recently streamline models have shown significant promise in improving the history matching process. In particular, the streamline-based "assisted history-matching" utilizes the streamline trajectories to identify and limit changes only to the regions contributing to the well production history. It is now a well-established procedure and has been applied successfully to numerous field cases.

In this paper, we enhance the streamline-based assisted history matching in two important aspects that can significantly improve its efficiency and effectiveness. First, we utilize streamline-derived analytic sensitivities to determine the spatial distribution and magnitude of the changes needed to improve the history match. Second, we use a "generalized travel time inversion (GTTI)" for model updating via an iterative minimization procedure. Using this approach, we can account for the full coupling of the streamlines rather than changing individual or bundles of streamlines at a time. The approach is more akin to automatic history matching. However, by intervening at every step in the iterative model updating, we can retain control over the process as in assisted history matching. Our approach leads to significant savings in time and manpower during field-scale history matching.

We demonstrate the power of our method using two field examples with model sizes ranging from 105 to 10 6 grid blocks and with over one hundred wells. The reservoir models include faults, aquifer support and several horizontal/high angle wells. History matching is performed using both assisted history matching and the GTTI. Whereas the general trends in permeability changes are similar for both the methods, the GTTI seems to significantly improve the water cut history matching on a well-by-well basis within a few iterations. Our experience indicates that the GTTI can also be used very effectively to improve the quality of history match derived from the assisted history matching. The changes to the reservoir model from GTTI are found reasonable with no artificial discontinuities or apparent loss of geologic realism.


Geostatistical reservoir models are widely used to model the heterogeneity of reservoir petrophysical properties, such as permeability and porosity. These geostatistical reservoir models are usually upscaled from fine-scale geologic/geocellular models to coarser reservoir simulation models for field development studies and performance predictions.

It is imperative that geostatistical reservoir models incorporate as much available, site-specific information as possible in order to reduce the uncertainty in the subsurface characterization. Available information on reservoir heterogeneity can be broadly categorized into two major types: static and dynamic. Static data are time-invariant direct or indirect measurements of reservoir properties, such as core measurements, well logs, and seismic data. These data can, relatively easily, be integrated into geostatistical models using the traditional geostatistical algorithms.1 Dynamic data are the time dependent measurements of flow responses that are related to the reservoir properties through the flow equations, such as pressure, flow rate, fractional flow rate, or saturation. Integration of dynamic data generally leads to an inverse problem and requires the solution of the flow equations several times using an iterative procedure.2–3 The process is generally referred to as "history matching" and is usually the most tedious and time-consuming process of a reservoir simulation study.

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