The development of a New Enhanced Oil Recovery (EOR) screening method

Abstract

Selection of a suitable EOR process for a given reservoir is a challenge for Reservoir Engineers where he or she has to compare pros and cons of all the available EOR methods in the context of each selected reservoir. The task becomes even more complex if they have to go through the details of hundreds of mature and depleted reservoirs and come up with a short list of EOR candidates for more detailed investigation. Conducting such a screening study could be daunting unless it is approached in a systematic manner.This thesis describes development of an EOR screening method which will enable engineers to sieve through some commonly available reservoir parameters for dozens or hundreds of reservoirs at a time, in a relatively efficient manner without compromising quality of the outcome. An extensive theory and literature review has been conducted to develop a main IOR/EOR database and to narrow down on the essential few criteria which would be sufficient for judging a reservoirs candidacy for one or more known EOR techniques. The EOR techniques considered for screening in this study includes CO2 miscible flooding, steam injection and polymer flooding. Screening parameters proposed includes formation type (lithology), reservoir depth, permeability, porosity, oil gravity, oil viscosity, reservoir temperature and minimum miscibility pressure.Two types of classification methods have been selected to perform the screening based on the selected parameters. Multilayer feed-forward neural network and classification tree method have been utilized using data mining software called XLMiner. Sensitivity analyses on the training and validation percentages, input variables, data partitioning and training process have been carried out to examine the robustness of the developed screening system and produce the best neural network and classification tree for further screening purposes. It was found that there was disagreement of the screening results between neural network, classification tree and the actual cases for some fields. Classification procedure, the reference interval of input parameters within the training dataset, and missing parameter(s) are the possible reasons for this outcome.The screening test results show that a combination of the Multilayer Feed Forward Neural Network and the Classification Tree can provide a good EOR screening system. However, when there is disagreement in the screening results from these methods, a further investigation (manual screening) needs to be conducted for the disagreed cases.The success of the screening system using the Multilayer Feed Forward Neural Network and Classification Tree depends on the quality of the database, as well as the training and validation processes.