Automated Well Test Analysis I

Abstract

Due to the continuous search for cost effective and less time consuming means of obtaining reservoir and well parameters(k, S, etc.), well test analysts have sought for other means of automating the well test interpretation process. Although nonlinear regression is central to the automation process, its use is limited by the subjective selection of the regression reservoir model. This is due to the difficulty in distinguishing the characteristic behavior/features of the pressure (or pressure derivative) response hidden behind noise or of ambiguous plots and this usually leads to wrong parameter estimation. To forestall this problem of model selection, an Artificial Intelligence (AI) approach has been developed to identify the features necessary to discriminate these different models.

This approach completely automates the well test interpretation process and involves the generation of a representative dimensionless pressure derivative data and the extraction of symbolic data from the pressure transient data. This symbolic data is matched with the generated dimensionless pressure derivative data and subsequently used by the AI system to chose the reservoir model and make initial model parameter estimates. Nonlinear regression is then used to refine these estimates. The part 2 of this paper presents the analysis of the results of this approach.