Prediction of Oil Density using Group Method of Data Handling (GMDH) Approach and the Effect of Reducing Correlating Parameters; A Comparative Study

Reservoir fluid or PVT properties are one of the most important elements in petroleum
engineering, especially in reservoir studies. It is required in material balance, reservoir
simulation, volumetric calculations and others. With the laboratory studies and the aids
of PVT correlations, PVT properties can be effectively obtained. PVT correlations
existed in the oil and gas industry is widely used when the experimental data cannot be
obtained or no fluid samples are available. However, some of the empirical correlations
in the literature are controversial in aspects of its accuracy, validity and range of
applicability. Recently, group method of data handling (GMDH) is introduced in the
petroleum industry as another alternative to improve the accuracy of existing PVT
correlations. This research proposes GMDH approach as a modeling tool for predicting
crude oil density at bubble-point pressure. The objective of this research is to study the
capability of GMDH in modeling oil density. The new oil density model incorporates
three (3) correlating parameters: (1) bubble-point oil formation volume factor, (2)
solution gas-oil ratio and (3) API gravity. A comparative study is carried out to compare
the performance of the new oil density model with other existing correlations. The
results obtained show that the oil density model with GMDH is more accurate and
outperforms other known correlations.