An intuitive, trial-and-error approach to history-matching can be costly and time-consuming. In an attempt to reduce these factors, considerable efforts have been made to automate history-matching procedure for implementation on high-speed computers; and several methods based on ex post facto techniques have been developed for this purpose.
This study examines presently available automatic history-matching algorithms reported; discusses the viability of each; and explores the fundamental weaknesses of two of the more promising methods. In addition, the existence of alternative approaches to solving the inverse problem for analogous, but less complex, situations is noted. Most published methods address the problem of single phase flow, although many would appear to be satisfactory for multiphase flow.
The most serious problem encountered in automatic history-matching is the tendency to construct ill-conditioned systems of equations Ax=b. By the very nature of the history-matching problem, inherent uncertainties exist in both A and b, because they are based on a measured performance history. This history-matching problem can therefore be treated only by incorporating some a priori information about the solution vector x. But due to the scarcity of physical data, even this requirement cannot be adequately satisfied.
It would seem that the current automatic history-matching methods are less than satisfactory. Such methods are simply not competitive with manual techniques which an experienced reservoir engineer might employ. It would appear, then, that the determination of effective and reliable solutions to the inverse problem requires development of improved methods of geological parameters, and adaptation of mathematical techniques to allow for a more theoretical treatment.
