The literature on numerical simulation is very extensive. However, there are very few papers which deal with the process of history matching.
The approach to history matching recommended by most textbooks on numerical simulation is to match the pressure first, and then match the saturation dependent performance. The parameters to be varied should be selected on the basis of the uncertainty in the value, i.e. parameters with greater uncertainty should be adjusted first.
In this paper, an alternative approach for oil reservoirs is proposed, where the emphasis is placed on matching the cumulative produced fluids as a function of time.
During each stage, the field data should be examined to identify the sources of the produced fluids and the principal reservoir mechanisms. When the mechanisms are identified, the parameters which control each mechanism should be varied to produce the match, taking care that the changes are consistent with the uncertainty in the parameter.
The approach is illustrated with 5 examples of matching early water production under primary recovery.
In most numerical simulation studies, the most time consuming part is calibrating the model to match the historical performance of the reservoir. This process is commonly called "history-matching". Table 1 shows a typical plan for a simulation study - the history matching phase is allocated xx% of the total time.
In spite of the importance of history-matching, there is very little in the oil industry literature about this process. Most papers either present a new mathematical /computational technique or present the results of a reservoir study with emphasis on the reservoir description and predicted performance under different operating conditions.
The situation in textbooks is similar. Table 1 shows the fraction of pages in several common textbooks(1–5) which are devoted to the different aspects of numerical simulation. In all of these, history-matching is a minor component, though it could be argued that the "model building" or "input data" sections cover many aspects of history-matching.
Courses on numerical simulation come in three basic types:
Physics of reservoir behaviour and mathematical development.
Construction of input data set.
Use of specific set of simulation software.
These can be labelled as: "math", "data collection" and "point and click". None deal with history-matching in any detail.
In spite of the lack of formal avenues for acquiring the skills for effective history matching, many good matches are produced by numerous practicing "simulation engineers". How is this achieved? In most cases, the courses, textbooks and literature provide enough clues to start. Tips from more experienced simulation engineers are passed down, and as the engineer gains experience, he develops his own set of "ricks" to identify reservoir situations and apply an appropriate solution.
The net result of this process is that history matching has become an art, with each practitioner having his own style, with strengths and weaknesses depending on experience (in both time and geography/geology) and talent (intelligence and creativity). The disadvantage of this process is that each simulation engineer has to rediscover items which are known to others - i.e. each engineer has to re-invent the wheel for himself.