Methodology to Assimilate Multi-Objective Data Probabilistically Applied to an Offshore Field in the Campos Basin, Brazil

Abstract

This work applies a new methodology to assimilate multi-objective data (production, injection and, pressure of all wells) based on five of the twelve steps described by Schiozer et al. (2015) using reservoir simulation and uncertainty reduction for a brown offshore field in the Campos Basin, Brazil. We use probabilistic techniques to assimilate all data simultaneously, improving the performance of the process. The 12-step methodology by Schiozer et al. is based on a closed-loop reservoir development and management process. Steps 1 and 2 construct the model under uncertainties and select the numerical model. Steps 3 to 5 assimilate history data in an iterative process proposed by Maschio and Schiozer (2016). At each iteration, a set of best-matched models is selected to update the probability distributions of the reservoir properties (parameters) based on a correlation matrix. Steps 6 to 12, comprising the decision analysis, were not included in this work. The results reflect a practical application of the methodology, considering a real reservoir with two zones and complex behavior that was captured during reservoir characterization using an uncertainty reduction algorithm. The reservoir was characterized through the probabilistic combination of uncertain variables, based on well logs and seismic data. The probabilistic characterization highlighted the geological variability under uncertainty. A set of three hundred geological realizations with associated porosity, net-to-gross ratio, and permeability distributions was generated for further combination with uncertain dynamic parameters. The method DLHG (Discretized Latin Hypercube combined with Geostatistics) was used during the entire process to build approximately 1000 uncertain scenarios allowing the review of reservoir parameters in any iteration. The data assimilation process was used to update the probability density function for each parameter according to the data match indicators. We significantly reduced the uncertainty and improved production forecast reliability. This paper integrated different areas including reservoir characterization, reservoir simulation and history matching with the associated uncertainty reduction. The methodology was successfully applied in a practical case with several uncertainties, indicating good potential for application in other fields. The matching quality was better than in previous approaches.