Viscosity is the most important parameter influencing heavy oil production and development. While heavy oil viscosities can be measured in the lab from core and wellhead samples, it would be very useful to have a method to reliably estimate heavy oil viscosity directly from well logs.
Multi-attribute analysis enables a target attribute (viscosity) to be predicted from other known attributes (the well logs). The viscosity measurements were generously provided by Donor Company, which allowed viscosity prediction equations to be trained.
Once the best method of training the prediction was determined, viscosity was successfully predicted from resistivity, gamma-ray, NMR porosity, spontaneous potential, and the sonic logs. The predictions modelled vertical viscosity variations throughout the reservoir interval, while matching the true measurements with a 0.76 correlation.
Another set of viscosity predictions were generated using log-derived seismic properties. The top viscosity-predicting seismic properties were found to be P-wave velocity and acoustic impedance. They predicted viscosity with an average validation error less than one standard deviation, however the predictions were less detailed with a correlation of only 0.35.
Also explored in this thesis was the effect of including depth as a viscosity predictor, predicting viscosity from acoustic logs scaled to seismic frequencies, and bitumen-water contact detection from acoustic logs.
View full article as PDF (8.62 Mb)