The purpose of this blog post is to cover the recently developed ResFrac capability that allows investigation of the effect of natural fractures on hydraulic fracture propagation. While this option has always been available for the ‘discrete’ propagation algorithm, now it also has become available for the ‘continuous’ algorithm. There are some noticeable changes compared to the previous implementation and they are covered next.
Typically, we don’t explicitly model natural fractures in hydraulic fracturing simulations since there is a wealth of field evidence from fiber optic data suggesting that hydraulic fractures are predominantly planar and are oriented in the SHmax direction. Nevertheless, there are many applications where it can be useful to explicitly represent preexisting fractures. For example, natural fractures may be more important for some geothermal reservoirs.
To model the effect of existing fracture stages in multi-well developments, we often represent hydraulic fractures on a parent well using preexisting fractures. This brings us to a very important distinction among the preexisting fractures, namely, preexisting natural fractures and preexisting hydraulic fractures. From now on, we define preexisting hydraulic fractures as the preexisting fractures that are aligned with the SHmax direction. At the same time, preexisting natural fractures are the ones that are not aligned with the SHmax direction. Within the paradigm of planar hydraulic fractures that is currently adopted in ResFrac, preexisting hydraulic fractures propagate in mode I, similarly to the injection-induced hydraulic fractures. While the propagation from natural fractures occurs either in mode II or in mode III. Thus, this introduces a strong qualitative difference between the two types of preexisting fractures.
First, let’s cover the topic of preexisting hydraulic fractures. Fig.1 shows a simulation, in which a preexisting hydraulic fracture is being hit by a hydraulic fracture from a different well. One obvious feature of this simulation is that the preexisting fracture has a complex shape. How is it possible? Well, ResFrac allows users to specify rectangular preexisting fractures. But nothing precludes users to create multiple fractures nearby such that they overlap and effectively create a single fracture with a complex shape. Also notice that the use of continuous propagation algorithm with partially filled elements allows us to represent the prescribed fracture length and height precisely. In contrast, for the ‘discrete’ approach the fracture dimensions are always rounded to be an integer number of elements. As Fig.1 demonstrates, the preexisting hydraulic fractures behave similarly to regular hydraulic fractures, they can collide with other fractures with the same collision logic as when two hydraulic fractures collide, they can propagate in mode I if pressurized, they can also be depleted or fluid can be injected into them.
Next, let’s review mode II propagation from preexisting natural fractures. Fig.2 shows an example of a height-contained hydraulic fracture colliding with two symmetrically placed natural fractures. Right now ResFrac uses a relatively simple crossing logic, which can be specified by the user. Hydraulic fractures can either always cross natural fractures, never cross, cross or not cross depending on the natural fracture orientation angle, or have a certain probability to cross. In this example, to promote mode II hydraulic fractures, the ‘never cross’ option is selected. As can be seen from Fig. 2, once the natural fractures are pressurized, mode II fractures are created at the edges of the preexisting fractures. Note that the fractures are initiated on both edges, but propagate only from one due to stress shadow. The use of ‘continuous’ propagation allows us to more accurately represent the geometry of all the fractures, so that the hydraulic fracture stops precisely at the fracture intersection and the new fractures are generated precisely at the fracture edges. This is all thanks to partially filled elements.
A more complex example showing interaction with natural fractures is shown in Fig. 3. Here multiple hydraulic fractures interact with a system of randomly generated natural fractures. The well is then put into production and the pressure field is shown in the figure.
Finally, the last example demonstrates mode III propagation from a preexisting fracture. This example is motivated by experimental results of Ruiting Wu (2006). Fig. 4(a) shows these experimental results. A radial hydraulic fracture is first generated, then a twist is applied, and the pumping resumed. As can be seen, this creates petal-like fractures, each of which is inclined relative to the original radial fracture. Such a configuration presents a great challenge from the modeling perspective and yet is very interesting from the fundamental side. While we cannot model it precisely in ResFrac, it is possible to make an approximation. Fig. 4(b) shows a preexisting natural fracture that is inclined 10 degrees relative to the Shmax direction. This fracture is supposed to resemble the initial radial fracture. Then, fluid is injected into this fracture and only mode III propagation is allowed. Fig. 4(c) shows the final result, where multiple mode III fractures are generated from the natural fracture. Note that the visualization is stretched 10x for a better view. In contrast to the experimental results, all the newly induced fractures are parallel to each other, but yet still have a small offset, so have similar stress shadow interactions. Results qualitatively agree with the experiments in the sense that multiple petal-like fractures are created.
To further explore this topic, Fig. 5 presents sensitivity of the result to fracture propagation regime. In particular, panels (a) and (b) show the results for toughness-dominated propagation, while panels (c) and (d) show the corresponding results when viscosity dominates. Here (a) and (c) correspond to early time, while (b) and (d) to late time. Results are qualitatively close to previously observed behavior in ‘Understanding fracture morphology’. Namely, when toughness dominates, fractures tend to avoid having an overlap and develop distinct petals. At the same time, fracture shapes are more regular and have overlaps if viscosity dominates. These observations also bring insights to what could have happened in experiments if injection continued and also how the degree of overlap depends on input parameters.
To summarize, the capability of fracture propagation from preexisting fractures opens intriguing possibilities in different ways. As is demonstrated in this blog post, it can be used to quickly account for the effect of parent well fractures, it can be used in field applications to investigate the effect of natural fractures on fracture propagation and production, or it can be used to address fundamental mechanisms of hydraulic fracture mechanics.
Donstsov, Egor. 2021. “Understanding fracture morphology.” ResFrac Blog Post.
Wu, Ruiting (2006). Some Fundamental Mechanisms of Hydraulic Fracturing, Ph. D. thesis, Georgia Institute of Technology.