Replies: 4 comments 3 replies
-
What is black oil? |
Beta Was this translation helpful? Give feedback.
-
It is a simplified model that consider hypothetically a mixture of oil and gas (maybe water) to describe fluids properties. It is extensively used in the Petroleum industry, because it does not require a lot of informations from the fluid, such the composition of each component. To describe a fluid using black-oil model we require basic inputs such the solubility ratio and bubble point pressure (in a specified temperature), and as a result we can get the fluid density and viscosity, for example. These models are very simple, the pro is that it does not require computational power, but the con is that they are not so powerful as the compositional model (cubic EOS, for example). Also, they are mainly experimental correlations, which do not give us a thermodynamic approach such as EOSs. I'm trying to compile different correlations to put them together. These models can be used in Reservoir simulators or Multiphase flow simulators to enhance the numerical solver performance, because in these simulators we need to calculate fluid properties in each control volume in the discretized domain. I hope I explained a little what it is. Reference -> A paper from IntechOpen: PVT Properties of Black Crude Oil (https://www.intechopen.com/chapters/64816) |
Beta Was this translation helpful? Give feedback.
-
I am not entirely sure this is a good fit for CoolProp. It doesn't exactly fit into the CoolProp modeling methodology, and I think cubic EOS are a better fit. |
Beta Was this translation helpful? Give feedback.
-
@Gualcar I'm a petroleum engineer and I've been using CoolProp for about half a year now to. Black oil correlations are totally inappropriate for CoolProp. In fact, I would go as far as to say that unless you are doing a manual calculation as a sense check, these days you shouldn't be using correlations. For hydrocarbon mixtures cubic EOS are a better fit, particularly recent variants like EPPR78. That said, CoolProp is very useful for pure fluids like CO2, particularly when considering applications like CCS, which is what brought me to this project originally. CoolProp does also already have implementations for SRK and PR cubics which you might find interesting. Now, specifically talking Black Oil vs Compositional... yes Black Oil is faster, but not because it uses correlations. Black Oil is faster because it uses pre-computed look up tables (LUT) e.g., PVTO and PVTG keywords in an Eclipse simulation deck. Those tables are actually usually generated using a cubic EOS (at least if you're doing it properly). You can use correlations, but as I said before, you shouldn't be doing that these days. By having a LUT you don't have to solve the cubic EOS flash calculation each and every time you want to find a value. You just look up an interpolated result from the table. You might be interested to know that CoolProp has a similar functionality as it can tabulate the EOS results and then use a bicubic spline lookup which speeds things up. Compositional models, on the other hand, perform a flash calculation for each and every cell, on each and every time step. Bazillions of iterations. This is why they take a long time. So to try to reduce the computation times, engineers have been employing lumping schemes etc. to reduce the number of components which just strikes me as somewhat backwards. What's the point in using a more sophisticated approach if you are then going to simplify your input data just to make it work? I've often wondered if there wasn't a better approach. Take dry gas recycling for example, a prime candidate for compositional modelling. The injected dry gas composition is known. The in-situ wet gas composition is also known. So in any given cell, at any given time, there is a composition that will be close to X% original and (1 - X)% injected gas. X can be determined easily enough using Rv (or tracers if you must). So this means we could simply generate a series of LUTs for our PVT properties for the range of X from 0% to 100%. Our pseudo-compositional model should now run at a similar speed to a Black Oil model as it is just doing look up interpolations to get the fluid viscosities, expansion factors, dewpoints etc. However, it would have close to the improved sophistication as a compositional model, potentially better as you are not compromising the composition by lumping it -- you can pre-compute the LUTs using the full composition. I was going to write a simulator to test this idea out, but then I discovered that OPM Flow has already implemented it using their PVTSOL keyword. I strongly encourage you to look into it. The only snag you'll run into is that, to my knowledge, you won't easily find tools to generate the PVTSOL keyword input needed. I've had a personal hobby project going for a while now to implement cubic EOS, including EPPR78, in a tool that allows generation of PVT tables for ECLIPSE simulator etc. I'm close to releasing a version of this, which will in all likelihood initially be free to download and use. It's probably still a few months away. If you're interested I can let you know when it's out. |
Beta Was this translation helpful? Give feedback.
-
Hello everyone.
Do you think Black-oil correlations for Oil & Gas applications such multiphase-flow modeling, reservoir simulations, etc, is a good feature here in CoolProp?
With this feature I think CoolProp could be used by these people from the Petroleum Engineering. I don't know if this is the main idea of CoolProp to be a "hub" of fluid properties calculations, but I think it should be nice to have a unique library where you could perform black-oil or compositional analysis.
Beta Was this translation helpful? Give feedback.
All reactions