Integration of Uncertainty Analysis and History Matching: Application in a Complex Case

Abstract

Abstract The large amount of uncertainties on reservoir modeling increases petroleum production forecast risks. Therefore, the history matching, which refines the simulation model to closely reproduce production data, is a vital procedure once it approximates numerical models to reality providing reliable predictions. As history matching is an inverse problem, the use of multiple models with good match is important to guarantee future analysis. Many methodologies were developed to integrate uncertainty analysis and history matching, in order to mitigate the reservoir uncertainties by using the observed data. In this context, the present article aims to evaluate the application of existing methods on a synthetic complex model (i.e. similar to a real field) and propose new methods with some improvements to be applied in real cases of the petroleum industry. The main characteristic of these methods is the use of differences between observed and simulated data to recalculate the probabilities distribution of uncertain parameters with the purpose of reducing reservoir uncertainties. The redistribution of occurrence probabilities is made with different formulas, representing the distinct four methods discussed: the first two are reviewed from literature and other two proposed by authors. The first proposed method combines the best practices of the two reviewed ones, making it robust to be used in real cases with a great number of wells and production functions to be adjusted such as water production and pressure. The second proposed method is only an iterative application of the first one with a redefinition of attribute values in order to refine the results. A comparison of the results of the four methods shows an evolution in the uncertainty reduction. Besides that, there is a decrease in the dispersion of the representative curves, which are centralized around the history data. Finally, this work presents a simple and practical way of integrating uncertainty analysis with history matching that provides reliable reservoir simulation models for production forecast and risk analysis of future projects.