DEVELOPMENT OF COMPOSITIONAL MODEL FOR PREDICTING VISCOSITY OF CRUDE OILS USING POLYNOMIAL NEURAL NETWORKS (PNN) INDUCED BY GROUP METHOD OF DATA HANDLING (GMDH)

Viscosity or the intemal resistance of the fluids to flow is the most important
transport property that controls and influences the flow of oil through porous media
and pipes. Accurate predictions of reservoir fluids are required in equation of state
(EOS) based reservoir simulators. Due to time and money spent of experimental
viscosity measurements, reliable viscosity models are developed for predicting crude
oils viscosity. Throughout the years, although many of the common correlations
were developed, laboratory measurements still cannot be replaced due to the
complexities, varied composition and reservoir characteristics difference from
different reservoirs. This study estimates crude oil viscosity by using a group method
of data handling (GMDH) based on polynomial neural network (PNN). GMDH is an
inductive algorithm for computer-based mathematical modeling using neural
network with active neurons that optimizes model coefficients for predetermine
mathematical equation and selects the optimal model complexity. The new model
was built and tested using experimental measurements collected through literature
search. The database consists of crude oils composition, viscosity, temperature and
pressure from Middle East, North Sea and the others. Overall, the proposed model
improved the prediction as compared to other viscosity model.
111