Full text loading...
-
Optimization of WAG in Real Geological Field Using Machine Learning and Nature-Inspired Algorithms
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, ECMOR XVII, Sep 2020, Volume 2020, p.1 - 14
Abstract
Maximizing oil recovery is a challenging task for the oil industry worldwide, mainly in the presence of dynamic technical and economical constraints. To achieve this target, a number of enhanced oil recovery technologies are being applied, and one of the most successful and used methods is water alternating gas injection (WAG). The estimation of the optimal operating parameters of the WAG process is a complex problem which requires considerable number of time-consuming runs. Therefore, developing a faster alternative tool without scarifying the precision of the numerical simulators becomes essential. Proxy models that are user-friendly mathematical models based on machine learning and pattern recognition, have a noticeable ability to deal with highly complex problems, such as the outcomes of the numerical simulators in reasonable time.
The present work aims at establishing various dynamic proxy models for optimizing a constrained WAG project applied to real field data from “Gullfaks” in the North Sea. Two types of artificial neural network (ANN), namely multi-layer perceptron (MLP) and radial basis function neural network (RBFNN) were taught for predicting all the needed parameters for the formulated optimization problem. Levenberg–Marquardt (LM) algorithm was applied for optimizing the MLP model, while genetic algorithm (GA) and ant colony optimization (ACO) were applied for the proper selection of the RBFNN control parameters. Furthermore, the best proxy model found was coupled with GA and ACO for resolving the WAG optimization problems.
The results showed that the established proxies are robust, practical and effective in mimicking the performance of numerical reservoir model. In addition, the results demonstrated the effectiveness of GA and ACO in optimizing the parameters of WAG process for the real field data used in this study. The findings of this investigation contribute to the knowledge of the mathematics of oil recovery in various perspectives, namely the establishment of cheap and accurate time-dependent proxy models for real cases, the optimization of WAG process in the presence of various types of constraints and also the robustness of nature-inspired algorithms for resolving the optimization problems related to enhanced oil recovery.