Changes

==History matching==

==History matching==

[[file:conducting-a-reservoir-simulation-study-an-overview_fig6.png|thumb|{{figure number|6}}Simulated water production for Initial and modified reservoir description. (From Mattax & Dalton,<ref name=pt10r22 /> © 1990 Society of Petroleum Engineers.)]]

[[file:conducting-a-reservoir-simulation-study-an-overview_fig6.png|300px|thumb|{{figure number|6}}Simulated water production for Initial and modified reservoir description. (From Mattax & Dalton,<ref name=pt10r22 /> © 1990 Society of Petroleum Engineers.)]]

Reservoir simulation models are normally calibrated using production history. During this process, significant modifications to the reservoir description may be required to match historical performance. [[:file:conducting-a-reservoir-simulation-study-an-overview_fig6.png|Figure 6]] shows the initial trial and the final match achieved in a well model. History matching is often a trial-and-error process requiring 20 model runs or more before a satisfactory match between the predicted and observed performance is realized. In new fields, little data exists and reservoir performance predictions may be made without history matching, but this produces unreliable performance predictions. Any data at all, even well tests from discovery and delineation wells, should be used to calibrate the reservoir model. Because of the relatively minuscule volume of data sampled by well logs, it is simple to construct a model that produces performance predictions that diverge dramatically from actual results. Any amount of data, unless it has been proven erroneous, is more credible than performance predictions from the most elaborately constructed simulator.

Reservoir simulation models are normally calibrated using production history. During this process, significant modifications to the reservoir description may be required to match historical performance. [[:file:conducting-a-reservoir-simulation-study-an-overview_fig6.png|Figure 6]] shows the initial trial and the final match achieved in a well model. History matching is often a trial-and-error process requiring 20 model runs or more before a satisfactory match between the predicted and observed performance is realized. In new fields, little data exists and reservoir performance predictions may be made without history matching, but this produces unreliable performance predictions. Any data at all, even well tests from discovery and delineation wells, should be used to calibrate the reservoir model. Because of the relatively minuscule volume of data sampled by well logs, it is simple to construct a model that produces performance predictions that diverge dramatically from actual results. Any amount of data, unless it has been proven erroneous, is more credible than performance predictions from the most elaborately constructed simulator.