Designing An Expert System For EOR Screening Based On Machine Learning Classifiers

Enhanced oil recovery(EOR)provides an opportunity to increase the oil production.There are various types of EOR methods,and it is too crucial to choose the right method for the right reservoir.EOR projects are usually go through different steps.The first step is EOR screening which is defined as a process of choosing the right EOR method for a target reservoir.Different studies have been done to develop new EOR screening criteria.These are usually involved traditional methods or more advanced methods such as artificial intelligence.In this study,new machine learning classifiers have been used such as ensemble classification method.New concept was defined for choosing the right number of clusters by using silhouette value.In addition,an innovative approach for EOR screening was provided which considers the category of the EOR method,and only after that by using the expert system and reservoir engineering knowledge,EOR method can be selected.At the end,the results show that it is better to keep the database intact for supervised machine learning classifiers as inputs.Ensemble classification(Bagged trees)was selected as the most accurate classification method since it combines all machine learning classifiers.Final framework can give recommendations up to seven different EOR method from the most probable to the least probable.It has the high accuracy of 86% when only first recommendation is considered and accuracy of combined 95% when the first two recommendations are concerned.The designed framework can be considered as one of the most accurate frameworks in the area of EOR screening.At the end,an innovative commercial software was designed(Screen Pro)to make friendly graphical user interface for engineers and operators which will be introduced at the end of this dissertation.