Viscosity Modeling

After having developed a tuned EOS model, the final step is to viscosity tune the model. Provided you have viscosity data. If you do not have viscosity data, you would have to add it to the sample(s) selected, or add another sample with that to the project and at least run the EOS Tuning again.

General Overview

In the picture below, we see the options available to us.

Here, there are many of the same features as in the EOS Tuning tab, where we can again ignore experiments, add weight factors and set boundaries for our parameters. One thing to note is that the viscosity modeling utilizes the LBC model, but can utilize the Z_c,vis to ensure for monotonic component viscosity. The Pedersen model will also be available in the next whitson^PVT update.

LBC Parameters

If we take a look at the LBC parameters, we observe that only the P4 parameter has been selected for tuning.

As with all whitson^PVT calculations, the default tuning parameters set, is what is recommended for the inital run. This means it is advisable to generate results with as little tuning as possible, and if desired adding further parameters to the tuning process.

Z_c,vis

The first option available to us, is to select the Z_c,vis regression method, which is split between the 'scaling' and 'ramping' methods.

The scaling method will adjust all component properties up or down equally, while the ramping method will tune heavier components more than lighter ones. Further described in the section modules/Viscosity Modeling

In the graph below, we see the impact of using scaling vs ramping for the Z_c,vis tuning for a two-sample EOS with one viscosity experiment.

It can be observed that the two methods deviate at opposite sides of the measured oil viscosity trend. If we increase the max range (Parameter Tuning Bounds) of Z_c,vis and P4, allowing these to be altered by 50% we will see a drastic difference in the ramping method. The scaling method yields negligable difference in this case.

There are still some deviations at higher viscosities (low pressure), but we could play around with the boundaries to align this value. It is always a good idea to compare the actual experimental data, so we see the pressures we are actually dealing with.

In the oil viscosity experiment comparison below, it becomes apparent that by playing around with the boundaries to ensure higher viscosity match, we have simply matched our model to atmospheric conditions.

While this may be relevant in specific instanses, we generally would like to have a good match at reservoir conditions, to accurately capture subsurface fluid flow.

Practical Workflow

Having gone through the impacts of the various properties, we could try to optimize our viscosity modeling.

Try generating cases for each of the following steps:

1. Run a regression with only P4 tuning (default case)
2. Unlock the Z<sub>c,vis</sub> Multiplier tuning
3. Unlock P3 Multiplier tuning
4. Unlock P2 Multiplier tuning

Utilizing this workflow, see that it yiels a lower Root Mean Square (RMS) than for our previous examples, meaning the overall match to experimental data is better.

Having generated a few different cases, the next step would be to compare results and selecting the case you're most happy with. You can then click 'Store Fluid Model', which will transport the page to 'Fluid Models' where you can see the model you have now created!

Next Steps

The next step in the fluid model development is to Quality Check your model!