Modeling Frac Hit Damage in ResFrac

Shale wells often experience huge production losses after a frac hit. For example, Figure 22 from King et al. (2017) shows a parent well in the Woodford that experienced a 65% reduction in oil production after it was hit by the fracturing of a neighboring child well. Production damage is not limited to parent wells – child wells tend to significantly underperform parent wells. What’s going on and what do we do about it?

A key question: why do parent wells lose so much production? There are reasonably well-understood explanations for why a parent well would lose production, but they appear to be insufficient to explain the severity of production loss at parent wells. The relatively well-understood explanations are: (a) depletion reduces the stress in the formation, and this tends to attract hydraulic fractures (Roussel et al., 2013); (b) fracture asymmetry causes inefficient drainage and loss of depletion efficiency (Cipolla et al., 2018); (c) the wells experience production interference as they compete to produce from the same rock. (Rimedio et al., 2015), and (d) proppant is remobilized and pushed into the wellbore. The severity of the parent/child impact has been found to depend on the age of the parent well (Elliott, 2019), which is consistent with these interpretations. But while these mechanisms are surely having an effect, they do not appear to be sufficient to explain the severity of frac hit production loss in most shales.

After the frac hit, the well may be clogged with proppant and water, and require cleanout with coiled tubing. But even after the cleanout and months of flowback, production may remain heavily depressed.

An important clue: frac hit production loss in parent wells is highly variable by formation. As reviewed by Miller et al. (2016), wells in the Bakken typically do not experience substantial production loss after frac hits. But wells in formations like the Marcellus or Woodford tend to experience significant production loss. Why should this depend so dramatically on formation?

It appears that chemical and/or multiphase flow effects play a major role in causing parent well damage and production loss. These effects depend on the mineralogy of the rock, and the composition of the formation fluid (both hydrocarbon and water phases), which explains why the effect is formation-specific.

ResFrac is a combined hydraulic fracturing and reservoir simulator. It can simulate the physics of frac hits, in a way that no other simulator can: multiphase flow as parent-well fractures mechanically reopen (and stress shadow each other), proppant is remobilized, and fluid cross-flows through the wellbore. In cases without large production losses at parent wells, we’ve been successful at history matching the full life-cycle of the well(s). In contrast, in datasets where parent wells have experienced large production losses after frac hits, we have had difficulty history matching. We simulate asymmetry and production interference, and these do have detrimental impact on production. But production interference isn’t enough to cause a well to lose 2/3 or more of production, as discussed by King et al. (2017) and many others.

This week, Fowler et al. (2020) is presenting a case study from the Bakken, written with Craig Cipolla at Hess, where we used ResFrac to history match a complex series of events in three Bakken wells, including frac hits, reinjection, and production, all in a single continuous simulation. However, in that dataset, the parent well experienced a substantial production increase after the frac hit (which we modeled in ResFrac), as is common in the Bakken.

When analyzing production data, we review the RTA plots, GOR trends, and water cut trends, and seek to identify systematic differences before and after production. These lines of evidence suggest not only production interference, but also either a loss of fracture conductivity and/or a skin along the fractures impeding production into the fractures.

All of this suggests that ‘additional’ frac hit damage mechanisms are at play. We need to understand what these damage mechanisms are and how they work. We need to integrate them into ResFrac so that we can successfully history match these types of datasets – and then optimize engineering solutions for our clients.

Damage mechanisms

In the SPE literature, there are a handful of pioneering papers that diagnose chemical/multiphase frac hit damage mechanisms in field data. I have been digging into these papers and having conversations with colleagues. Now, I am implementing processes into ResFrac to model the damage mechanisms that they describe. We run parent/child simulations with different mechanisms, compare the results with real data, and then use the comparison to help diagnose what is happening. Once we’ve done that, we use those history matched models to help design mitigation strategies.

Broadly, I see two types of frac hit damage – fracture conductivity damage (occluding flow through the proppant pack) and frac skin damage (blocking flow into the hydraulic fractures from the surrounding rock). Within those categories, damage might block flow of all three phases, or it might block flow of only the hydrocarbon phases, but not water.

Nieto et al. (2018) used swabbing operations to recover black, gunky solid particles from a damaged retrograde gas parent well in the Montney. Analysis indicated that the particles were composed of a mixture of silica (crushed proppant), formation fines, and high molecular weight hydrocarbons. The heavy hydrocarbon components were found to be soluble in an aromatic solvent, but not in a paraffin solvent, suggesting that they were asphaltenes. (Asphaltenes are large molecular weight hydrocarbon molecules with strong aromatic component. Because of the conjugated rings, they have some, but not unlimited, solubility in oil. See Pedersen and Christensen, 2006). Importantly, Nieto et al (2018) found that iron oxide formed from the reaction of formation iron with dissolved oxygen in the frac fluid. The iron oxide created nucleation points for asphaltene to agglomerate out of solution, and also cemented together the silica and fines into larger particles. The thick, gooey asphaltene semi-solid helped muck all this together. The parent well experienced more than a 50% reduction in production after the frac hit. Interference tests were performed, and only minor pressure communication was observed between the wells. This all points to a ‘fracture conductivity’ form of frac hit damage.

It surprises me that they encountered asphaltene in a retrograde condensate gas well. Typically, we would associate asphaltene with a heavier oil, not with the liquid dropped out from a retrograde condensate. Possibly, the asphaltene is coming from the TOC in the shale itself?

Regardless, this paper points to a (rather complicated) mechanism that leads to frac hit conductivity damage. While this is somewhat speculative, it appears that: (a) during production, as liquid drops out from the gas phase in the matrix and eventually becomes mobile through the formation, it dissolves small amounts of asphaltene from the TOC, (b) over time, as that liquid flows into the proppant pack, asphaltene accumulates in the liquid phase in the proppant pack, (c) when the frac job is performed, the oxygen dissolved in the frac fluid begins reacting with iron from the formation, making iron oxide, and (d) the iron oxide facilitates asphaltene coming out of solution and also cements together fines and ground proppant, creating the black sludge that blocks the proppant pack.

I don’t know whether or not this is exactly the right mechanism. But regardless, Nieto et al. (2018) show conclusively that: (a) the well experienced major frac hit damage, (b) it was not caused simply by well interference, and (c) it was associated with the formation of an iron oxide/asphaltene agglomeration with crushed proppant and fines. Whether or not we understand all the details, if we know this is happening, we can simulate it in ResFrac and engineer to mitigate it.

Rassenfoss (2020) reports on another mechanism that could cause conductivity damage – a so-called ‘gummy bear’ phenomenon in the Woodford. Again, iron from the formation is a culprit. When the cross-linking occurs, the result is a thick gooey mixture of cross-linked gel, ground up proppant, and formation fines. I did a quick informal lit review, and it does appear that the Woodford is unusually high in pyrite, compared to other shales.

The mechanism described by Rassenfoss could occur in a parent well, not just during a frac hit. And in fact, operators in the Woodford do report seeing this type of damage in parent wells. Chemical formulations can be helpful for operators to mitigate these issues.

In contrast to fracture conductivity damage, fracture skin damage is harder to conclusively diagnose. Unlike processes that create physical material that can be pulled out of a well, fracture skin damage involves blockage of flow occurring out in the reservoir. The most obvious potential mechanism for fracture skin damage is ‘water block.’ The idea is that water leaks off into the surrounding rock, and accumulates (rather than flowing back or flowing our further into the formation). The layer of accumulated water could block hydrocarbon flow as it tries to produce into the fracture.

Swanson et al. (2018) report very positive results from pumping chemical remediation treatments in damaged parent wells in the Woodford. They were uncertain about damage mechanism and so threw the kitchen sink – surfactant to reduce interfacial tension and mitigate water block, HCl to dissolve scale, and HF/HCl to dissolve silica/clay mineral scaling/fines, etc. We can’t be sure which mechanism was most important (and it might have been the gummy bear mechanism discussed by Rassenfoss). Whatever it was, they got good production recovery from these wells. Surfactant to prevent water block is certainly a reasonable fluid to include in a chemical treatment.

For all of these mechanisms, we might reasonably ask – why don’t these mechanisms also occur during the main frac job? What’s so special about frac hits?

For the gummy bear problem reported by Rassenfoss (2020), my understanding is that this may, in fact, be a substantial problem in parent wells in the Woodford. On the other hand, the mechanism described by Nieto et al. (2018) would not be expected to occur in parent fracs. It requires the presence of asphaltene in the fractures, and that will only occur as the asphaltene is drawn into the fracture over time by production.

For water block, we can reasonably hypothesize that this problem might be more severe in child fracs than in parent fracs. Elputranto et al. (2018) discuss how capillary end effect can cause water to accumulate along the walls of a hydraulic fracture, blocking flow. In general, capillary end effect occurs when a rock with capillary pressure is opened to a zero-capillary pressure interface (such as the side of core in the lab, or a fracture wall). Capillary end effect is closely related to the process of spontaneous imbibition of a wetting fluid into rock.

Elputranto et al. (2018) did not specifically discuss the topic of frac hits. But we can think through mechanistically why capillary end effect could be more severe after a frac hit than after the original parent frac (kudos to Joe Frantz for pointing this out to me). Elputranto et al. (2018) perform simulations to show that development of capillary end-effect depends on how pressure is available to drive fluid into the fracture. If the pressure gradient is strong enough, it can overcome the capillary end effect, fluid is drawn cleanly into the fracture, and the wetting phase is pulled further out into the formation by capillary pressure, instead of accumulating at the fracture walls. However, if the pressure gradient is weak, then flow still occurs, but capillary end effect is not overcome and water accumulates at the fracture wall.

When a parent frac is performed, the formation fluid pressure is still high, and so capillary end effect is less likely to develop. But, in a frac hit on a parent well, the formation pressure has been drawn down by prior depletion, and so there is less pressure available to overcome the capillary end effect. Thus, we might develop a ‘water block’ after a frac hit, but not in the original frac.

Finally – we might ask, what about preloads? Preloads are performed by injecting fluid into parent wells prior to an anticipated frac hit. They are, in fact, effective at reducing the amount of frac fluid that flows from the child well to the parent well. But why don’t they themselves cause damage – since they also involve injection of water into depleted fractures?

Perhaps because preloads are typically pumped with a different fluid chemistry than frac fluid. They are often pumped with surfactant, iron chelators, or other remediation chemicals that may not be found in a typical frac fluid.

To conclude, all of these lines of evidence suggest that chemical/multiphase flow effects are important in causing parent well damage, and that these effects can be mitigated and controlled by chemical and engineering decisions.

Modeling Frac Hit Damage in ResFrac

I implemented these different frac hit damage mechanisms into ResFrac. Right now, we don’t have a detailed understanding of most of these processes. I think that more detailed mechanistic study should be a high priority for future research in academia and industry. But in lieu of detailed understanding, we can still implement empirical/heuristic constitutive equations into ResFrac to describe them. Then, we can compare simulations with these mechanisms against field data, and determine which are most consistent with observations. We can history match datasets with frac hit damage, and then use those history matched models to quantitatively optimize mitigation strategies.

ResFrac also needs to be able to model chemical remediation. I implemented reactions that occur between injected solutes and the different forms of damage. The ResFrac user can define different reactions by saying which solutes react, with what type of damage, how quickly, and how potent these reactions are at reducing the damage.

Below, I run a series of simulations to demonstrate. First, I ran two simulations: (1) a simulation with only one well, (2) a simulation with a neighboring child well fractured and produced after one year. Next, I ran three additional simulations based on Simulation 2 that had different damage mechanisms turned on. Also, these simulations injected a remediation treatment into the parent well 200 days after the child frac job. The three damage simulations are: (3) development of water block from leakoff into depleted formations, (4) the ‘gummy’ mechanism based on Rassenfoss (2018), and (5) asphaltene/iron oxide gunk based on Nieto et al. (2018).

1. Single well simulation

In this baseline simulation with only one well, total gas production reaches 345 MMscf and water production reaches 6600 bbl of water.

2. Parent/child simulation without special damage mechanisms

When the child fracture is fractured, the injected fluid and proppant mostly heads back towards the parent well. Consequently, the child well severely underperforms the parent well. Relative to the simulation without a child well, the parent well production is only modestly impacted – a reduction from 345 to 311 MMscf at 5 years. The reduction is caused by production interference. Because of the water injected in the child well, there is a significant uptick in water produced at the parent well after the frac hit.

3. Water block from leakoff into depleted formation

This simulation mimics a mechanism is based on the capillary entry pressure mechanism discussed above (Elputranto et al., 2018). When the formation fluid pressure within a few ft of the fracture has dropped by a certain amount, leakoff is assumed to create a ‘fracture skin’ that represents water block. For example, the user may specify that once pressure has gone 20% below the initial pressure, leakoff begins to create skin. The user can specify that the fracture skin blocks the production of all three phases, or just oil and gas (and not water). The rate of skin development depends on the rate of leakoff and the amount of pressure depletion. The amount of skin that develops in each fracture-matrix element connection is quantified by tracking the ‘thickness’ of the skin region. The user specifies a relative amount that permeability is lower within the ‘thickness’ region than in the surrounding rock.

Below, we see the result. In this simulation, the water block reduces oil/gas production, but not water production.

Production is dramatically impacted in the parent well by the skin damage. Water production increases greatly, and gas production is almost totally eliminated. Because of the reduced interference from the parent well, the child well actually performs better than in the case with no damage mechanisms. The second figure below shows the distribution of water block in the parent well after the child frac. Damage primarily forms in the area near the parent well, where the frac hit occurred and fluid leaked off into depleted formation. The remediation treatment pumped after 200 days recovers some of the production of the parent well.