Asphaltene precipitation (AP) is recognized as a complicated occurrence that results in weakening reservoir characteristics and subsequent considerable decline in oil production rate. Asphaltene instability occurs due to variations in thermodynamic properties such pressure, temperature, and mixture composition. AP prediction is an important design factor in implementation of any enhanced oil recovery (EOR) process. In this study, experiments were conducted using some light oil samples to measure important phase behavior properties such as the bubble point pressure (BPP) and the amount of precipitated asphaltenes. A thermodynamics model was also developed to determine equilibrium compositions of the oil samples, considering AP. Then, potential application of a feed-forward artificial neural network (ANN) model, optimized by the imperialist competitive algorithm (ICA), was proposed to estimate BPP and the amount of AP. Comparison between the ICA-ANN predictions and the experimental data shows that the average absolute error between data originated from these two different approaches is less than 5%. In addition, it was found that temperature and pressure have the greatest impacts on AP during natural depletion. Employing laboratory PVT data, the thermodynamics framework resulted in construction of an asphaltene precipitation envelope. This study implies that utilization of an appropriate PVT model along with the ICA-ANN approach in the investigation of AP leads to more reliable predictions compared to the conventional ANN and also a scaling model. The outcomes of this study appear to be useful in the design stage of more-efficient EOR processes.
کلید واژگان :asphaltene genetic algorithm particle swarm optimization neural network natural depletion porous media
ارزش ریالی : 600000 ریال
با پرداخت الکترونیک