Uncertainty History Matching and Forecasting, a Field Case Application

Abstract

Abstract Geological, reservoir, economical and technological uncertainties have an effect on decision making and consequently on reserves development plans. Quantifying the impact of these uncertainties can make this process more reliable. A great difficulty to achieve this in practice is the variability and complexity of workflows available to manage uncertainty using numerical simulation. The inaccuracy, high uncertainty or lack of reliable data yields risk to the forecasting process, making the calibration of the dynamical model with the field production data indispensable. History matching is an inverse problem and, in general, different combinations of reservoir attributes can lead to acceptable solutions, especially due to the high degree of uncertainty of these attributes. A set of solutions that respect the observed data may lead to different prediction scenarios. The objective of this work is the integration of history matching with probabilistic analysis of representative scenarios. A methodology that allows the recognition of well-calibrated models within an acceptable deviation is used. This procedure helps to identify the critical uncertain parameters and their possible variation in order to estimate the representative reserve range. The goal is not to find the best deterministic match, but rather to show how the calibration process allows a mitigation of identified uncertainties. A real case based on a reservoir from Campos Basin in Brazil was used. A 14 year historical period followed by a 12 year forecast period was considered, allowing verification and validation, at a global level, of the proposed procedure in a complex dynamic model. Two different commercial softwares were used, in order to demonstrate the advantages and restrictions of each approach. Distribution variations of the responses in time were evaluated by Latin Hypercube sampling and Monte Carlo propagation on validated proxy models. The proposed methodology allows: (1) to reduce the range of possible models taking into account the observed data; (2) to identify the existent uncertainty as a function of observed data; (3) to reduce the uncertainty range of critical reservoir parameters; (4) to increase confidence in production forecast. One contribution of this work is to present a quantitative approach for increasing the reliability of reservoir simulation as an auxiliary tool in decision making processes in order to reduce the associates risk and to maximize development opportunities.