Modeling of Two-Phase Gas Deviation Factor for Gas-Condensate Reservoir using Artificial Neural Network
Advances in Research,
In petroleum engineering, reservoir fluid characterization is of great importance. Accurate determination of the two-phase gas deviation factor is essential in modeling gas-condensate and gas reservoirs, pipeline flow and reserve estimation, this is because the reservoir fluid is in a two-phase state at pressures below the dew-point pressure. Correlations are replete for predicting single-phase gas deviation factor using different Equation of State (EOS), but no correlation have been found to accurately predict the two-phase gas deviation factor.
Traditionally, the two-phase gas deviation factor for a gas-condensate fluid is determined experimentally in the laboratory, however, this laboratory experiments are quite expensive, though quite reliable. Hence, a need for simple but less expensive methods of determining the two-phase gas deviation factor. Thus, this present study modeled the two-phase gas deviation factor of a gas-condensate fluid using Artificial Neural Network (ANN), a biologically inspired non-algorithmic, non-digital, massively, parallel distributive and adaptive information processing system. Its ability to perform non-linear, multi-dimensional interpolations makes it unique and fit for this work.
The results obtained were compared to existing empirical and analytical correlations. Average absolute deviation (AAD), root mean square errors (RMSE) and correlation of determination (COD) between the ANN output and other correlations gave 1.343%, 1.344% and 61.6% respectively.
On the basis of the results, it was discovered that ANN approach is an improved, simple, less expensive and more accurate method of determining the two-phase gas deviation factor. ANN approach gives the closest value to the observed two-phase gas deviation factor from experimental work.
- Artificial neural network
- gas-condensate reservoir
- two-phase Z-factor
How to Cite
Abstract View: 483 times
PDF Download: 642 times