Application of Advanced History Matching Techniques to an Integrated Field Case Study

Abstract

Abstract This paper presents an innovative integrated methodology for constraining 3-D stochastic reservoir models to well data and production history as well as a successful application to a real field case. The proposed approach allows to history match complex reservoir models in a consistent way by updating the entire simulation workflow. Advanced parameterization techniques are used to modify either the geostatistical model directly or the fluid flow simulation parameters in the same inversion loop. In a first step, the relevant inversion parameters are selected according to a sensitivity study based on the experimental design technique. In a second step, history matching is performed with the most significant parameters using an automated inversion procedure. In this step, the Gradual Deformation Method (GDM) is used to constrain the geostatistical model while respecting the global model properties. This technique may be combined with gradient based inversion methods in order to history match other deterministic parameters simultaneously. A successful application to a real field case located offshore Brazil, named PBR, is presented. The lithofacies reservoir model was built using geostatistical simulations with the non-stationary truncated Gaussian method. The fluid flow model includes about 40 wells and 15 years of production history. The entire simulation workflow was integrated in one history matching loop in order to update the geostatistical lithofacies model and the fluid flow model simultaneously. The mismatch between simulated and real production data was quantified by the calculation of an Objective Function (OF). A sensitivity study based on the experimental design technique was performed to determine the most influential parameters within the workflow and to capture the possible reasons of the mismatch. Based on these observations, an automatic history matching was performed with the key parameters. A set of deterministic parameters, including the facies permeabilities, vertical anisotropy ratios, correlation lengths, relative permeability end points and Corey exponents was used first to improve the global match. Then, the GDM was applied to modify the geostatistical model itself. The GDM proved efficient in history matching to modify the spatial facies distribution of the geostatistical model. Moreover, the repetition of this history matching procedure with several model realizations confirmed the reduction of uncertainty on production forecasts. Results on this real oil field case showed that this innovative approach, which combines both the experimental design technique and the GDM, enables detailed analysis of the mismatch and significantly increases the predictive quality of reservoir models, using a limited number of simulations.