The SPE Hydraulic Fracturing Technology Conference was last week, and as usual, there were some outstanding papers presented. Here are a few highlights.
In SPE-204140-MS, Brinkley et al. (Devon) gave results from their sealed wellbore pressure monitoring (SWPM) technology. When hydraulic fractures intersect cemented, not-yet-completed offset wells, they cause a small pressure response. By tracking the timing of these responses across multiple wells, it is possible to infer the geometry of the hydraulic fractures. SPE-204140-MS is not the first paper on this topic. However, it was notable because they gave a lot of detail about their results from applying the concept. They aggregated a large number of hits from the Eagle Ford, and derived detailed calibrations of fracture length and height growth in the Upper and Lower Eagle Ford and the Austin Chalk (their Figure 14). In their Figure 17, they report that Eagle Ford fractures tend to be much shorter than Midland Basin fractures, which tend to be much shorter than SCOOP/STACK fractures. Keep in mind – these are the total fracture length, and not the propped or producing fracture length. Their findings are concordant with our philosophy in ResFrac that fracture length in shale is highly variable between formations and needs to be measured and calibrated against. However, once you have calibrated data within a play, the model can be used reasonably across the basin.
In SPE-204185-MS, Huckabee et al. (Shell) provide an excellent overview of their results measuring and optimizing limited entry completion using fiber. They provide an excellent perspective on the current state of the art in the industry, and summarize their own findings from a large number of fiber installations. In their Figure 2, they show the distribution of breakdown pressure within a stage in the Delaware Basin. It shows variability over a range of 2300 psi, suggesting that significant limited entry is needed in this formation to achieve high perf efficiency. I was excited about this finding, because in our work, we have sometimes found that we need to assume random variability in breakdown pressure in order to match measured perf efficiencies. Note – breakdown pressure is not the only reason we need limited entry. We also need to maintain a uniform distribution of flow between the clusters that have broken down. In their Figure 14, they show that the variability in breakdown pressures in other basins is quite a bit lower than in the Delaware Basin. In their Figure 4, they provide estimates for fracture net pressure in the Delaware Basin. In their Figure 5, they provide estimates for near-wellbore tortuosity derived from step-down tests. As a modeler, I am loving these figures because I like to see the numbers that others are coming up with, so that I can check for consistency and gain perspective.
In SPE-204203-MS, Synder et al. (COP) investigated the effect of phasing on perf erosion. They observed that perfs along the top of the hole tend to be lower diameter, even prior to injection, because of imperfect centering of the guns. In addition, the bottom perfs tend to erode at a substantially greater rate because of the dynamics of the proppant slurry in the wellbore. Their ‘best practice’ recommendation is that you should orient your perfs at 0’ phasing along the top of the well in order to maximize uniformity of flow.
In SPE-204177-MS, Horton (Ovintiv) provides an interesting perspective on the perf phasing question. He notes that in a strike-slip faulting regime, longitudinal fractures will tend to form along the sides of the well, rather than along the top and the bottom. He trialed 0/180’ phasing and 90/270’ phasing, and found that the 90/270’ phasing led to significantly lower pressure drop from near-wellbore tortuosity, evidently because the perfs align with the longitudinal fractures. This is advantageous because it lowers pumping cost. The implication is that, in strike-slip faulting regimes, the ‘best practice’ may be to use 90/270’ phasing. In normal faulting regimes, we would still rely on the advice from Synder et al. and use 0’ phasing only along the top.
In SPE-204165-MS, Sakardande and Devegowda (University of Oklahoma) described a clever statistical method for assessing the impact of parent/child effects. The problem with all statistical studies using production data is that they are ‘observational’ studies. Correlated inputs, confounding covariates, and spurious correlations caused by uncontrolled ‘false positive rate’ have the potential to create apparent relationships that are not actually meaningful. Sakardande and Devegowda draw on the mathematical theory of causality to develop a method based on pairing up wells that are assessed as ‘similar,’ which allows them to make cleaner apples to apples comparisons. There is no perfect way to resolve these problems, but I think that their approach is intriguing. They show an example where a naïve search for correlation leads to conclusions that are opposite the findings from their more advanced technique.
In SPE-204199-MS, Shelley et al. (Liberty Oilfield Services) summarize results from an analysis of their database of frac jobs. As discussed above, I think that all statistical studies should be taken with a grain of salt. Nevertheless, it’s useful to keep an eye on then, and focus on identifying their high level trends. Their Figure 4 shows that in their dataset, the Wolfcamp B operators were pumping jobs in the range of 1000-2000 lbs/ft proppant, with very few above or below that range. Fluid volume varied from 20-60 bbl/ft. There appears to be a modest correlation between fluid volume and production, and a weak correlation between proppant volume and production. But again – these results should not be taken too literally because nonrandom selection may obscure the true relationships. In the paper, they bin wells by frac design and reservoir quality and estimate optimal designs as a function of reservoir quality.
In SPE-204205-MS, Wu et al. (Silixa and Apache) provide a detailed look at the amazing fracture mapping work that is possible with offset fiber. This paper is a companion paper to a ResFrac/Apache paper, SPE-204172-MS (Shahri et al.). As discussed in a recent blog post, the fiber data shows fractures propagating in relatively narrow corridors, with subparallel strands of opening-mode hydraulic fractures. Not zig-zagging, branching fracture networks. I love this kind of data because it provides excellent calibration data for numerical models, similar to the SWPM, but with even more detail.
In SPE-204151-MS, Dontsov et al. (von Gonten and BPX) demonstrate the application of the ultrafast frac simulator developed by Egor Dontsov. It’s impressive to see the model in action.