Probabilistic Forecasting for Mature Fields with Significant Production History: A Nemba Field Case Study

Abstract

AbstractThe Nemba Field is the third largest oilfield in terms of daily production in the Block 0 Concession in the Republic of Angola. Oil production from the field began in January 1996 through an Early Production System (EPS). The field reached a peak production rate of approximately 110 MBOPD in August 2002 and is currently undergoing crestal gas injection at an instantaneous Voidage Replacement Ratio (VRR) of approximately 0.85. The field is being evaluated for the installation of a new compression train to allow for gas injection at or above a VRR of 1.0.To support this effort, along with other reservoir management activities, a reservoir characterization and simulation study was performed in 2003/2004. The coarse-grid, reservoir simulation model generated from this study was a six component equation-of-state (EOS) model containing approximately 246,906 total cells (136,675 active) with approximately 33,794 non-neighbor connections to account for complex faulting and truncation across stratigraphic intervals. The workflow for this study consisted of data integration, fine-grid model development, scale-up, coarse-grid model development, history matching, and deterministic/probabilistic forecasting. What distinguishes the probabilistic forecasting of this brown-field development from that of comparable green-field developments is the available production history and the need to retain the integrity of the history match while fully assessing the impact of uncertainty levels in key geologic and project variables (uncertain parameters).The methodology employed in this study is similar to the framework proposed by Castellini et. al.1, but was adapted to use more off-the-shelf technologies. This methodology uses both single variable input modifications (sensitivity analysis) and simultaneous, multi-variable input modifications (experimental design). The sensitivity analysis were used to determine the appropriate uncertainty ranges for individual uncertain factors while the experimental design was used to determine the interactions of these uncertainty factors.During the experimental design portion of the analysis, modifications were made to the history matched input data and the simulator was run through both the historical and prediction periods of the project. In order to retain the integrity of the history match, a Quality of History Match (QoHM) variable was defined to quantify and rank any degradation to the history match caused by perturbation of uncertain data. Proxy equations for the simulation response were then generated for the QoHM variable and the project result variables: Estimated Ultimate Recovery (EUR) and discounted cumulative oil production. Finally, a two-step Monte Carlo simulation approach was used to develop the cumulative distribution functions (CDFs) for all alternatives considered for the gas compressor project.In this two-step Monte Carlo simulation, the uncertain factors were sampled from their individual CDFs for use in a Monte Carlo trial. The sampled data were first input into the proxy equation for the QoHM variable to determine the impact on the History Match. If the selected input data were found to meet predetermined acceptance criteria for the QoHM variable, then they were passed to the second step: input to the proxy equations for the project result variables. Input data combinations that did not meet the acceptance criteria were discarded. This process was repeated for the desired number of Monte Carlo trials to complete the simulation.Topics to be discussed in detail in this paper include: development of the reservoir simulation model, application of the proposed methodology for probabilistic forecasting, development of the QoHM variable, selection of the QoHM acceptance criteria, the two-step Monte Carlo simulation approach, final project results, and lessons learned / best practices in probabilistic forecasting of brown-field development projects.