Ranking the appropriate enhanced oil recovery (EOR) methods for a specific oil field characteristic is a difficult and challenging task due to the large number of related parameters and financial risks. However, an intelligent ranking tool enables making important decisions about potential EOR strategies by using previous reservoirs experiences. In this research, first, a new production rate consisting of natural production and EOR production called EOR-PR is introduced, which resolves the shortcomings of the proposed rate of previous research. Then, a new machine learning approach for ranking EOR methods based on predicting the EOR-PR values is proposed. In this approach, the efficiency of methods on specific conditions of a reservoir including rock and fluid characteristics is calculated numerically, which is simply comparable. In this regard, Multi-Gene Genetic Programming (MGGP), which is a powerful machine learning method for modeling engineering problems, has been employed to predict EOR-PR values and then its performance has been compared with artificial neural network method. The results show that MGGP with an average of 0.982 for R 2 correlation has a significant performance in this issue. Also, this method, unlike conventional machine learning methods, provides a definite function as the output that allows further analysis for reservoir experts.
Read full abstract