Different modeling approaches for predicting bubble point pressure of Yemeni crude oil

Abstract EN

Pressure-Volume-Temperature (PVT) properties are important reservoir parameters.

Two common approaches are available to obtain PVT properties empirical correlations and laboratory experiment.

Correlations are used whenever experimental PVT data are not available.

No universal PVT correlation is available and data from local regions are expected to give better approximation to estimated PVT values.

Available PVT data from 35 oil fields from different regions of Yemen were used to obtain different models for calculating bubble point pressure.

These models include artificial neural networks (ANNs), neuro-fuzzy model, neuro-genetic model and empirical correlation equation for bubble point pressure.

Statistical error tools were used to evaluate the performance and the accuracy of empirical correlations and different models.

It was found that standard artificial neural networks outperform when genetically optimized or used with hybrid neural fuzzy inference system.

New developed empirical correlation shows good agreement with the actual data comparing with the commonly used correlations.