Automated Parameter Estimation From Well Test Data Using the Signal Theory

摘要

This paper presents a new methodology for the automatedrnparameter estimation from well test data, based on type curvernmatching using the Signal Theory. This procedure solves therninverse problem faster than the conventional techniques, withrnthe additional advantage that the results are not affected byrnnoisy data.rnTrnhe new procedure has been proved with hundred ofrnsynthetic and field cases, and it can be used for therninterpretation of all the different tests currently of commonrnuse in the field. In the present paper this technique is appliedrnto three tests, all already published in the literature.rnThe methodology derived in the study surpasses the currentlyrnavailable matching theory based on non linear regression,rnsince it requires a smaller computing time and its assuredrnconvergence (selection) of the correct (best) model thatrndescribes the physical conditions of the formation on thernvicinity of the well. On the other hand, the conventionalrnregression methods require large computing times for thernsolution of the system of non linear equations; in addition,rnthey are affected by the presence of noise in the recordedrnpressure response and usually present convergence problemsrnwhen the initial solution for the unknown parameters is notrnclose enough to the searched solution.rnTrnhe conventional type curve matching procedure inherentlyrnintroduces interpretation subjectivity and the possibility ofrnerrors, because of the close similarity of the different pressurernresponses, and to its visual solution approach, whichrncurrently is not well understood; thus, it is not possible torndevelop fully efficient codes that could emulate this humanrnfunction.Based on a comprehensive set of type curves, the techniquernproposed in this paper allows an automated match of thernpressure response, without the need of the visual effort of thernanalyst. The Signal Theory has been modified for thernautomated interpretation of well test data. The new matchingrncorrelation derived in this work, based on rules of shift,rnmultiply and sum, solves the matching of pressure data in anrnimproved way.