Abstract
Intelligent reservoir management requires assessment of risks caused by geologic heterogeneity and examination of alternative development schemes. Numerical simulation is a powerful tool for integrating geologic and development models, but simulation is often too expensive or time-consuming if there are many sensitivities and scenarios to examine. In this paper, response surface methods are used to approximate the relationship between gas recovery responses and reservoir and production parameters. The response models are based on a set of numerical simulations selected using experimental design, which efficiently estimates effects of many factors across a broad range of uncertainty. The three-dimensional reservoir model examined in this study is from an outcrop study of the Permian-age Bell Canyon, which is interpreted as a channel-levee-lobe turbidite system.
Examination of realistic reservoir models is complicated by the large number of factors and by correlations between model factors. Principal component analysis can simplify the modeling process by using the correlations to reduce the dimensionality of the factor space. A large number of reservoir geologic parameters are reduced to a small set of principal components while most of the original geological information is still preserved. The principal component design also spans the scatter-cloud of correlated reservoir factors more efficiently. Based on these principal components and several engineering factors, reservoir simulations are selected using a central composite design. Response surface models relating the simulation responses and the design factors are estimated via linear regression. The validity of the models is verified by reservoir simulation of randomly selected points within the feasible region of geological parameters. Uncertainty analysis, including parameter sensitivity estimation and recovery uncertainty assessment, is done conveniently and inexpensively using response surface models and Monte-Carlo simulations. Quality maps identify optimal well locations within the complex depositional setting documented at Willow Mountain.
