DEVELOPMENT OF AN EXPERT SYSTEM TO IDENTIFY PHASE EQUILIBRIA AND ENHANCED OIL RECOVERY CHARACTERISTICS OF CRUDE OILS
With the increasing demand of oil and gas in the past decades, great endeavors in the oil industry have been devoted to develop and incorporate better techniques for production. Enhanced oil recovery (EOR) is put forward as an effective way to maintain the reservoir pressure and increase oil production. Because of the complexity of reservoir conditions and interpretation of case dependent production history, it is time consuming to run the reservoir numerical simulation applying EOR method such as gas flooding, thermal flooding and chemical flooding. In the literature, one can find tools which are developed to provide screening criteria for oil recovery methods and assessment of production scenarios. Some of those tools are qualitative utilizing experimental field applications knowledge and some are based on models which are quantitative.
Artificial neural network (ANN) based toolbox by Claudia Parada (2008) provides an alternative way for screening and designing enhanced oil recovery methods with four fluid types included to investigate EOR processes. The evaluation of EOR production scenarios or performance forecast for crude oils may not be easily obtained with high accuracy because in reality there are many crude oils with various physical and thermodynamic properties. In this research, four artificial neural networks are developed to represent four fluid types, respectively. Networks predictions are interpreted to categorize crude oils and the classification results are validated by numerical simulations. An expert system integrating all networks by Graphical User Interface is created to categorize a crude oil in a visualized manner.
Once the classification is achieved for a particular crude oil, investigation of appropriate recovery techniques and adequate design guidelines can be applied to a certain fluid type representing this crude oil in Parada’s toolbox with specific reservoir characteristics. Thus, the developed expert system assists in narrowing down the selection of a proper fluid type combined with feasible EOR processes, which subsequently can help to reduce time-consuming experiments.