A detailed simulation study of three-dimensional laboratory experiments, representing steamflooding in a tar sand reservoir underlain by a communicating water zone was conducted to validate a thermal simulator and to delineate important process effects. Four experiments which included effects of injection rate and unconfined patterns were simulated. The same input properties were used in all runs.

Calculated results agree well with the measured data for oil production rate, cumulative oil recovery, produced water/oil ratio, temperature profiles in the model at different times, and the oil saturation profile at the end of the flood for all experiments. The linear model for three-phase oil relative permeability predicts steamflood residual oil saturation more accurately than the Stone's II model. For unconfined patterns, a significant portion of displaced oil can escape through the bottomwater zone.

References and illustrations at end of paper.


Computer simulations are widely used for field-scale performance calculations of steam injection processes. However, some major differences have been observed between the field data and the simulator-predicted performance.1 These differences are generally attributed to the lack of detailed reservoir geology, relative permeability data, three-phase oil relative permeability models, and/or migration of fluid across the reservoir boundaries. Therefore, to lower the uncertainty in the predictions, a (non-unique) history match of the observed field data is often performed as a prelude to process variation or parametric studies.2,3

Three-dimensional, scaled, physical models have also been extensively used to investigate and predict steamflood performance.4,5 High-pressure models employ actual field crude oil and reservoir pressures and temperatures. Thus, they represent real compositional and flow behaviors of the crude oil. However, all parameters are not scaled in physical models. Numerical models, on the other hand, sometimes employ simplifying assumptions or semi-empirical models to describe complex processes; for example, three-phase oil relative permeability and high-temperature rock-fluid interaction. Therefore, physical and numerical models can complement each other as a means of investigating process mechanisms.

Physical models are ideal for validating numerical simulators and for evaluating modeling and input parameters. They have simple geology, well-defined boundary conditions, and known initial saturations. In addition, detailed injection and production data, temperature profiles, and residual saturation profiles are generally obtained in a laboratory model study. However, little published information is available on detailed comparisons of three-dimensional laboratory vs. numerical model studies.

Shutler 6 simulated one-eighth of a five-spot laboratory experiment using a two-dimensional, wedge-shaped, cross-sectional (x-z) model. Shutler showed good comparison between the calculated and measured cumulative recovery curves. He did not compare any other measured and calculated data.

Coates et. al.7 compared Intercomp's three-dimensional model against Shutler's6 data and Getty's laboratory experimental data for a quarter of a five-spot model. They compared the cumulative recovery curves and temperature contours for one experiment and showed reasonable agreement. The discrepancy between the model and observed data was speculated to be caused by the inhomogeneities in the sand pack, non-uniform initial oil distributions, and relative permeability curves and their temperature dependence.

This content is only available via PDF.
You can access this article if you purchase or spend a download.