Chapter 1 - Fundamentals Of FD-RTA

01-01 - What Is FD-RTA? (20 min.) Sample Lesson Quiz: 01-01 - What Is FD-RTA?
01-02 - Inside FD-RTA (24 min.) Quiz: 01-02 - Inside FD-RTA

Chapter 2 - Analytical FD-RTA

02-01 - Equally Spaced Fractured Wells (22 min.) Quiz: 02-01 - Equally Spaced Fractured Wells
02-02- Unequally Spaced Fracture Wells (28 min.) Quiz: 02-02- Unequally Spaced Fracture Wells
02-03 - Gas Wells (11 min.) Quiz: 02-03 - Gas Wells

Chapter 3 - Numerical FD-RTA

03-01 - Creating & Using Fast Numerical Models (27 min.) Quiz: 03-01 - Creating & Using Fast Numerical Models
03-02 - Condensate Wells (16 min.) Quiz: 03-02 - Condensate Wells
03-03 - Calibrating Multiphase Oil Wells (33 min.) Quiz: 03-03 - Calibrating Multiphase Oil Wells

Chapter 4 - Advanced Topics

04-01 - Application of FD-RTA To Parent-Child Interference (11 min.) Quiz: 04-01 - Application of FD-RTA To Parent-Child Interference

Fractional Dimension RTA / Chapter 1 - Fundamentals Of FD-RTA

Lesson 01-01 - What Is FD-RTA?

Access All SAGA Wisdom Training content Subscribe

Already a member? Sign in

Access All SAGA Wisdom Training content Subscribe

Already a member? Sign in

Transcript

01. Lesson 1.01: What is FD-RTA?

Hello, my name is Jorge Acuna. I used to work for Chevron for more than 25 years. My experience is in naturallys. I did a lot of geothermal reservoir engineering, modeling, and then I transitioned to unconventionals in 2012, also with Chevron. And then I started to develop a new option to analyze unconventional wells and develop the fractional dimension that I built on the experience I had while doing my PhD thesis at USC in the mid-90s, with respect to the use of fractals to characterize complex networks of fractures. What you see here was developed sometime between 2015 and 2020 at Chevron. And I have been developing it ever since. Currently, I work as an independent consultant.

02. Outline

So in this presentation we will go over that outline. The idea is to give you an overall view of what is this fractional dimension, and then how it compares to the RTA that we have been doing normally. This is a relatively new method so that's why I spend the time to try to explain it.

03. FD-RTA vs Classic RTA

What we call classic was developed for systems where the fractures are equally spaced. That's what has been used in the last decade or so. The FD-RTA is short for Fractional Dimension RTA. It was developed to deal with complex systems of fractures. As I mentioned before, it was developed between the years 2016 and 2020. And there are 2 versions: the analytical version and the numerical version of it. The power of this method is that it solves not only the case of equally spaced fractures, but it also solves the case where the fractures are not equally spaced. They are clustered in groups of closely spaced fractures calledswarms. That's why we call that the fracture swarm model. And then we can deal with even a system of very complex, like the one you see there to the left.

04. Rate Transient Analysis (RTA)

Now let's talk a little bit about rate transient analysis. Rate transient analysis is a method that is somewhat of the family of pressure transient analysis. It is known also by the name of production data analysis. The difference is that instead of shutting in the well, I'm looking at the pressure recovery. works with flowing intervals, in which we analyze the well when it flows.
Now it is used in conventional reservoirs, except that it's not that useful, mainly because the transient part (that is the part that contains the information) goes very quickly. They last hours or days. And then what you see when you try to apply RTA to a conventional well is boundary conditions. So some people use it to constrain boundary conditions.
In unconventional wells on the other hand, because of the very, very low permeability, to reach the boundary conditions it could take months or years. And this is why this becomes the method of choice to analyze unconventional wells.

05. Production History to Create Diagnostic Plot

What we do for rate transient analysis is to take the production data (pressure, cumulative production, flow rate), and then we plot that data in a particular way in which we create a normalized pressure, and use the time, and with that we create a diagnostic plot. What is interesting is that in this case, for example, where I have 1.5 years of production, I am able to create that diagnostic plot that captures the entire history. That graph that you see to the right is the log-log diagnostic plot. And it represents a virtual test where we flow the well at a constant flow rate. So if it were possible to flow this well up 1 bbl/day and then we plot the pressure, it will look like what you have there to the right. The upper curve is the pressure (it's just this normalized pressure) and the lower curve is the derivative.

06. Effect of Parameters on log-log Diagnostic Plots

Now the shape of this plot tells us alot. There are different variables and I want to show how the variables change the shape, for example if you change δ. And we will see what that is in the rest of the course. What we change is the slope and also the separation between the 2 curves.
What we change is Xmf. That is the total frac length, more or less. But we changed the vertical position. Higher values will bring the curves lower, smaller values will bring the curves higher.
And the skin, what it does is it affects only the pressure curve, and it makes that early part different. So no skin, they both look parallel; some skin, the curve starts to lift; more skin, it lifts even more.
The pressure is given by this equation. Now this is a very simplified equation of the generalized solution but it's useful to understand this form.

07. Recent Exploratory Core Observations on Fractures

Now, why is it that we are trying to solve for systems that are not equally spaced? Because many companies and consortia of companies, they have invested alot of money trying to get exploratory cores to try to explore what is the true of these fractures that you find in unconventional wells. And what they found is a totally unexpected level of complexity. There are many, many more fractures than perforation clusters, and the fractures are grouped in closely spaced sets that we callswarms. They also noted deflection and branching, what creates a complex geometry. They also used very interesting experiments. They took plugs next to theulic fractures and away from it, and they measured permeability in those blocks, and they found that there is not really a difference between the permeability that they found. What that means is they were looking for evidence of the enhanced permeability regions and they didn't find that ever. There are some references there that you could read if you are interested in it.

08. Do We Really Need a New RTA Method?

Now the question is, do we really need a new method?

09. Effects of Fracture Spacing on Well Performance

Well, let's make this a sample. This is a simulated example so we are not dealing with the uncertainty of not knowing exactly what will work with real wells. Here are 2 wells. One of them had equally spaced fractures and the other one doesn't. But they have the same number of fractures, the same total frac length, the same properties (of the fluid, permeability,... everything is the same, same volume. Now we produce that for a given pressure that we impose. And you see the blue curve represents the response of the equally spaced and the red curve represents the response of the unequally spaced. You immediately see that there is a huge difference.
Now, if we were to do the traditional √t plot for both sets of data, this is what we would see. Because we know the total frac length, we could calculate the correct slope that goes with that frac length, and it is this one that we see here. So you see when the fractures are equally spaced, you correctly recover the properties of thatmade. But when they are not, what you see is that you can only match the very early time, and then the curve starts to bend upwards and deviates significantly from what you are expecting.

10. Log-Log Diagnostic Plot Always Works

Now if you were to do the log-log diagnostic plot for the 2 runs, this is what you see for the equally spaced fractures (that is how it looks), and to the right you see for the unequally spaced fractures. Notice that the slope is different for the 2. The slope is for the early part of the equally spaced is 0.5. That is what you will always get in the log-log plot when you have equally spaced fractures. But for the one to the right you get the slope of 0.70, that the slope is equal to (1 - δ), δ is 0.3 for that particular case. The position for both of those graphs will be total frac length (Xmf) × k ^ δ. Notice that when δ is 0.5 (that corresponds to equally spaced fractures) you get X√k, that is the familiar term that everybody uses. The onset of the boundary condition shows up as a deviation from the, let's call it linear channel, though it doesn't represent. It's linear because it shows like a straight line in the log-log plot. But you are able to see that. And with that time at the end of the non-linear flow you are able to calculate the volume, provided that you see it in your well. That is kind of hard to see sometimes.

11. Analyzing Equal Fracture Spacing Wells

Now let's say that I don't know what type of network I have. I only have the production data. And I know my data because this is a model. So if I apply the classic, I get the correct results. If I apply the FD-RTA I also get the correct results. OK, no problem.

12. Analyzing Fracture Swarm Model

Now aswarm model. I also know the data. If I apply the classic, I asked this software to do the best possible match with a slope of 0.5 into something that doesn't quite have this slope. And then if I calculate the frac length, the frac area, the volume, I notice that it is way off. If I apply FD-RTA with the results, I am able to recover the correct parameters. So that tells you what you find. If your well happens to be something like this (that, of course, wells like this doesn't exist in reality, but this is just a model to represent fragments that are not equal size fractures that are not equally spaced), you try to analyze that. What you find is, yeah, you match the early part with, but then it deviates very quickly from what should be an interaction between fractures that is related to reaching the size of the boundary. But you reach that very, very quickly in time and that doesn't make sense. And that's the problem. That's why you end up normally underestimating your frac length, and with that you underestimate the volume. That's why you have to apply other methods to calculate volumes.

13. Effect of Long-Term Predictions

Now, even if you have a relatively good history match in this case (for example you have up to 6 years), you do the best you can with the classic and you match with early data, but you project that to the future and you see that there is a very important difference.

14. How does FD-RTA Work?

Now let's explain how this FD-RTA works.

15. Complex Systems and Multi-Zone Fractional Dimension Models

Let's say that we have a very complex network of fractures. What we are going to do is measure the reservoir volume as it changes with distance from the fractures.
You were able to draw these contours. These contours are drawn at the same distance from the closest fracture. Now if I calculate the area inside each one of those contours, I end up with a curve that looks like that. Now that curve is not a straight line by any means. That is a curve of volume vs. distance. What we do for FD-RTA is to apply a set of straight lines for the different parts of the curve and try to match it to 3 segments. With that I'm able to create what is called a butterfly model (that is something that I will explain later), and the regions have different slope, they have different δ.

16. We can get CFV from Production Data

Now the nice thing about this is that even though those 2 models look so different, they share the same variation of volume with distance from the fracture. Notice that there are discontinuities in the butterfly model. Even though there are discontinuities in the cross-section, they are continuous in terms of volume. You integrate the volume from distance. With respect to distance from thethe curve never breaks. The curve you get is the one that the you see here. So you see it's smooth here and it's smooth here in the active transitions.
Now I simulate my pressure transient response to work in both systems, and this is how the 2 responses compare. And it is saying, even if I impose my constant pressure and calculate cumulative production and flow rate, this is how the 2 compare. What that tells me is that there is a way to simulate a complex system with something that is much simpler. And you will see what is important. Now something important that we will explain later on, is the shape of the volume curve and the cumulative production curve is the same. That is important because that opens the door to a whole new way of analysis in which you don't need to know the. All you need to do is process your production data to get a curve that will give you the geometry of an object that behaves exactly like you want, without having to go through knowing the actual geometry.

17. The Reason Why the Method Works

Why this system works, this FD-RTA, is because different systems with the same variation of volume with distance to the fractures perform the same. And this is true for systems where the conductivity of the fractures is high. What that means, and this is very important, is that we do not have to know the geometry of the to be able to build a model for it. Only from production data we are able to construct a model that will have the same variation of reservoir volume with distance from the fractures.

18. Butterfly vs Fracture Swarm Models in Multiphase

For example, once you have those models there are 2 alternative models that you could create. Knowing the parameters you could create aswarm model or a butterfly model. This is a comparison of a multiphase run for both of them. And you see that the results are very similar, except that gas may be in the fracture swarm model. Now they both run in multiphase and give comparable results. But the butterfly model runs in seconds. The fracture swarm model could take from more than 30 minutes to a few hours, depending on how many fractures you have. So the butterfly model, the advantage that it has is that it allows you to run multiphase models so quickly that you could do automatic calibration of the different parameters.

19. Workflow for Multiphase Well

So what does this FD-RTA, how does it work? Well, what we do is we take our production data and create the log-log plot. We fit it with the model. And then we get the parameters. And once I have the parameters, I have 2 options. I could create theswarm model or I could create a butterfly model. And with any of those I am able to run and calibrate multiphase model or a single-phase model, whatever you want to do.

20. Learnings

So what are the learnings that we have so far? If we have high conductivity fractures, any system with the same variation of volume with distance from those fractures has the sameulic response.
The butterfly and theswarm models are just that. They are just models for complex networks of fractures. We don't claim that any of those geometries have the actual of the system. We don't know what the geometry is and we will never know unless we expose it somehow (that is unthinkable). In any case, we can use our production records and be able to create models that will give me the same response that's possibly all I need for engineering purposes. Because with that model I could create predictions, I could do alot of things.
In real wells we can calculate the CFV (we don't show it but it is included in the method) by analyzing production and pressure data. And again, actual fracture geometry will be different.

21. Differences Between Classic RTA and FD-RTA

So what are the difference between classic and FD-RTA?
Remember that we said classic RTA is for equally spaced fractures. FD-RTA solves equally spaced fractures and equally spaced fractures in complex systems. Classic RTA gives shorter totallength for non-equally spaced systems. Here in the FD-RTA it gives a total frac length that is actually longer. For matching performance in classic RTA, sometimes you need EPRs. Here you do not require that. If you have a high decline, it usually requires pressure-dependent permeability. Here you don't need that. Usually just by lowering δ you get that, high initial production with high decline. Classic RTA deviates quickly from if the RTA matches the complete history of the well.

22. Summary

So we already mentioned that with this method you can analyze simple and complex networks. Results are physically consistent and can be validated with numerical simulation. It provides your new parameters to characterizes. That is very interesting because we are always trying to find ways of how to do this in reality, how to compare completions, how close this new well I am going to drill with respect to the wells that I already have, etc., etc.
This is already available in several commercial software that possibly you have seen.
Acuña, Jorge A. "Pressure and rate transient analysis in fracture networks in tight reservoirs using characteristic flow volume." In URTeC, Austin, Texas, 24-26 July 2017, pp. 338-356.Lems, Willem F., and Hennie G. Raterman. "Critical issues and current challenges in osteoporosis and fracture prevention. An overview of unmet needs." Therapeutic advances in musculoskeletal disease 9, no. 12 (2017): 299-316.Ciezobka, Jordan, James Courtier, and Joe Wicker. "Hydraulic fracturing test site (HFTS)-project overview and summary of results." In SPE/AAPG/SEG unconventional resources technology conference, p. D023S023R002. URTeC, 2018.