Candidate selection process for polymer gel application by using artificial neural networks

Örs, Oytun
Rapid increase in water-oil ratio has been one of the most important challenges of the oil companies that are producing from the mature oil fields. For this reason, various studies have been conducted to diminish the excessive water production. One of the most widely used chemical shutoff technique, polymer gel treatment, involves injection of different polymers into the production wells in order to plug the easy flow pathways of the excessive water. For the successful application of the polymer gel treatment, selection of the candidate wells is crucial. Consequently, this study aims to find a methodology that would be helpful during the candidate selection process and investigate the parameters dominating the successful applications. In this study, MATLAB® codes are developed in order to get benefit from the Artificial Neural Network technique for the candidate selection process and production data of the 60 wells are used for the simulations. Throughout this study, data pre-processing is applied to the available dataset in order to obtain more utilizable dataset. Here, some inconsistent field data is eliminated. Then, by using the dataset some success criteria that represent the success rate of the treatments are determined and calculated. Before performing a neural network analysis, Principal Component Analysis (PCA) is performed and dominating well parameters in the data set are identified. Then correlation analysis is performed and some of the irrelevant parameters are excluded from the neural network analysis. Finally, by combining the PCA and correlation analysis different scenarios are developed for each success criterion and 450,000 different neural network structures are developed for the entire study. For each success criterion, the most successful neural networks are selected and their error levels are also presented.  
Citation Formats
O. Örs, “Candidate selection process for polymer gel application by using artificial neural networks,” Ph.D. - Doctoral Program, Middle East Technical University, 2017.