Quantitative ranking of different operating areas and assessment of uncertainty due to reservoir heterogeneities are crucial elements in optimization of production and development strategies in oil sands operations. Although detailed compositional simulators are available for recovery performance evaluation for SAGD, the simulation process is usually deterministic, cumbersome, expensive (manpower and time consuming), and not quite suitable for real-time decision making and forecasting.
In this paper, Artificial Neural Network (ANN) is employed as a data-driven modeling alternative to predict SAGD recovery performance in heterogeneous reservoirs, an important application that is lacking in existing literature. In this study, numerical flow simulations are performed to construct a training data set consists of various attributes describing characteristics associated with reservoir heterogeneities and relevant production/injection parameters with the corresponding recovery factor as output. The network is trained using the data set to identify all significant patterns and relationships that exist between these attributes and the output parameters. The model is then tested using a verification data set (cases that have not been used at the training stage). Sensitivity studies on network configurations are also investigated. In addition, new modifications are proposed to identify and reduce extrapolations in predictions, which are often considered as major drawbacks in most data-driven modeling approaches.
The approach described in this paper can be integrated directly into most existing reservoir management routines. In addition, the technique can be used as a viable tool for analyzing large amount of competitor data efficiently. Given that robust forecasting and optimization of heavy oil recovery processes is a major challenge faced by the industry, the proposed research has great potential to be applied in other recovery projects such as solvent-additive steam injection.