Evaluation Of Below Bubble Point Viscosity Correlations & Construction of a New Neural Network Model
- Mohammed Abdalla Ayoub | Anwar Ibrahim Raja (UTP) | Muhammad Almarhoun (KFUPM)
- Document ID
- Society of Petroleum Engineers
- Asia Pacific Oil and Gas Conference and Exhibition, 30 October-1 November, Jakarta, Indonesia
- Publication Date
- Document Type
- Conference Paper
- 2007. Society of Petroleum Engineers
- 4.2 Pipelines, Flowlines and Risers, 4.6 Natural Gas, 5.5 Reservoir Simulation, 6.1.5 Human Resources, Competence and Training, 4.1.2 Separation and Treating, 5.2 Reservoir Fluid Dynamics, 5.6.4 Drillstem/Well Testing, 3.3.6 Integrated Modeling, 4.2.2 Pipeline Transient Behavior, 5.4.1 Waterflooding, 4.1.5 Processing Equipment, 7.6.6 Artificial Intelligence, 5.2.1 Phase Behavior and PVT Measurements, 5.3.2 Multiphase Flow, 3.1.6 Gas Lift, 5.4.2 Gas Injection Methods
- 2 in the last 30 days
- 507 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 8.50|
|SPE Non-Member Price:||USD 25.00|
This paper, precisely, evaluates two famous below bubble point viscosity correlations and tries to create a new Neural Network model for estimating this property. The new created model outperforms the two investigated correlations namely Khan Model (1987) and Labedi Model (1992). The new technique (Artificial neural network) found to be successful in developing a model for predicting viscosity below bubble point with an outstanding correlation coefficient of 99.3%. A limited number of data points have been collected from Pakistani fields in order to construct, test, and validate the model. Viscosity from 99 sets of differential liberation data covering a wide range of pressure, temperature, and oil density were used to validate the correlations and to develop the new model. A series of statistical and graphical analysis were conducted also to show the superiority of the model that has been formulated through an Artificial Neural Network technique. A thorough literature review is also made to check the applicability of the existing correlations and their drawbacks.
The main objective of this paper is to propose a simple procedure to predict black oil viscosity at the region below bubble point pressure as a function of easily determined physical properties. Based on thorough and critical literature survey of available technical and published papers, only two models that dealt with viscosity below bubble point were chosen; Khan et al model (1987) (1) and Labedi model (1992) (2).
Utilizing Matlab statistical toolbox, programs have been generated using regression analysis for both correlations. Neural Network toolbox was used for creation of a new successful model. Back propagation/feed forward scheme has been followed to generate a model. Statistical analysis was used to test the validity of the new model.
Viscosity is the measure of the resistance to flow exerted by a fluid; the lower the viscosity of a fluid, the more easily it flows. Like other fluid properties viscosity is mainly affected by temperature and pressure. An increase in temperature causes a decrease in viscosity. A decrease in pressure causes a decrease in viscosity, provided that the only effect of pressure is to compress the liquid. In addition, in the case of reservoir liquids, there is a third parameter which affects viscosity, which is the reduction in the amount of gas in solution in the liquid. It causes a decrease in viscosity; hence, the amount of gas in solution is a direct function of pressure.
However, as reservoir pressure reduces below the bubble point, the liquid undergoes a change in composition. The gas that evolves takes the smaller molecules from the liquid, leaving the remaining reservoir liquid with relatively more molecules with large complex shapes. This changing liquid composition causes large increase in viscosity of the oil in the reservoir as pressure decreases below the bubble point (as illustrated in Figure (1). Crude oil viscosity is needed in reservoir engineering as well as many petroleum applications such as calculation of two-phase flow, gas-liquid flowing pressure traverse, gas-lift and pipeline design, calculation of oil recovery either from natural depletion or from recovery techniques such as waterflooding and gas-injection processes. Besides, these correlations are also needed for the calculation of multiphase flowing pressure gradients in pipes at different temperature and pressures. Live oil viscosity is a strong function of pressure, temperature, oil gravity, gas gravity, gas solubility, molecular sizes, and composition of the oil mixture. The variation of viscosity with molecular structure is not well known due to the complexity of crude oil systems. It is not a quite trusted procedure to estimate viscosity depending on the crude oil composition.
When it is hardly to be measured in the laboratory, viscosity might need to be specified with high degree of accuracy in order to be involved in a series of highly sensitive calculations. The ordinary way to measure viscosity is definitely through laboratory equipment. The alternative way is through empirical correlations in case of lack of PVT information, which were found to be easily applied if they gained a wider range of confidence in industry.
|File Size||162 KB||Number of Pages||10|