Fracture models provide the ability to predict how changes to a fracture treatment should alter fracture geometry, but suffer from a tenuous and generally unknown relationship with
 
Figure 1. Calibrated model matches microseimic mapping results. The east frac wing was not fully observed — fracture is assumed to be symmetric. (All graphics courtesy of Pinnacle Technologies)  
reality. Fracture mapping provides a direct measurement of fracture geometry from a given treatment but cannot be used to predict what might happen under a different set of conditions. By combining direct measurements with models, we can create calibrated fracture models with superior predictive capabilities. Calibrated models have been developed for various regions and formations, and the improvements in predictive modeling capabilities have led to a proliferation of calibrated fracture models throughout the industry. This has provided improved insight into fracture growth behavior in a diverse set of environments throughout North America.

Overton field
The first example discusses the integration of microseismic fracture mapping with fracture modeling, well production results and offset well interference data in the Overton field of East Texas operated by Southwestern Energy (SPE Paper 95508). Target formation is the low-permeability gas-producing Taylor Cotton Valley at depths of about 11,500 ft (3,708 m). The fracturing program in this field included different types of waterfrac and linear gel hybrid frac treatments.

Microseismic fracture mapping indicated that created fractures in this area are very long, with half-lengths of about 1,500 ft (458 m). Figure 1 shows the mapping results along with the calibrated fracture model. Due to observation well position it was not possible to see the entire east wing of the fracture, so based on other mapping data in the Cotton Valley the fracture was assumed to be symmetric. Fracture modeling was performed using FracproPT, a calibrated 3-D fracture model. The model was calibrated with the microseismic mapping results by matching both the measured fracture geometry and the treating net pressures.

After determining the created fracture length, it is important to evaluate how much of the created length is actually effectively producing. There is compelling evidence in this area that
 
  Figure 2. Example of production interference between two wells along the fracture orientation (N71E) indicates long conductive fracture lengths.
effective hydraulic fracture lengths are also very long, as immediate well interference was detected in wells that are located along the fracture orientation with very large inter-well distances ranging up to 3,500 ft (1,068 m). Figure 2 shows an example of an interfering well pair. The insert figure shows the Well C gas production rate on a log-log graph. This graph is of particular interest as it shows clear production interference from offset Well D completed 6 months later (red arrow). Offset Well D was drilled along the fracture azimuth (as determined from microseismic fracture mapping) at an offset distance of about 1,850 ft (564 m). The interference is seen both as lower measured pore pressure in Well D and sudden slope changes (or abrupt loss of rate) in the older vintage Well C as the new Well D starts producing. The effect is almost immediate, within a few days, and indicates communication between wells through a long and conductive hydraulic fracture system (fractures from both wells link up). One well in this study area was temporarily killed by a fracture treatment in an offset well along the fracture direction, but production was restored within 1 day. Offset wells that were not positioned in the fracture direction did not see production interference. There was no evidence for a wide fracture network (as has been reported in the Barnett shale), but there may be a small-scale fracture network within a narrow area around the hydraulic fracture. Production modeling indicates that the permeability feeding the long hydraulic fractures (perpendicular to frac) is on average very low. The low permeability and long fractures will create slim “cigar-like” drainage areas that should be taken into consideration for well placement and spacing strategies. In this case, the operator repositioned future drilling locations to avoid the observed production interference. The results of this work illustrate the application of microseismic mapping to calibrate fracture models and, with the subsequent integration of production and interference data, to improve well placement and field development strategies.

Copper Ridge
The second example is from a coal reservoir in Wyoming and shows how calibrated fracture modeling can help with on-site fracture treatment design changes and optimization of future
fracture treatments (SPE Paper 84490). The Copper Ridge field is in southwest Wyoming near
 
Figure 3. Fracture height coverage and length for a two-stage completion strategy shows the final result of fracture optimization with the calibrated fracture model.  
the town of Rock Springs. Anadarko Petroleum was evaluating the development of the Almond coals in this field with a 16-well pilot program. The coals in the Copper Ridge field are usually present as multiple 2 to 20 ft (0.61 to 6.1 m) stringers, which cover 100 ft to 300 ft (30.5 to 92 m) of gross section at depths of about 2,600 to 3,000 ft (792.7 to 914.6 m). The most prolific coal(s) are usually perforated in 20-ft (6.1-m) intervals and fracture-treated with one to two stages using crosslinked gel with a combination of 16/30 and 20/40 mesh proppant. Underlying this section is the Ericsson sand, which is used for water disposal. Hydraulic fracturing must avoid the Ericsson sand under all circumstances. Overlying the coal sections are higher permeability shore-face water sands, which also need to be avoided.
In this project, measurements from treatment well tiltmeters which were deployed in the treatment well during the fracture treatment were integrated to build a calibrated fracture model. Treatment well tiltmeter mapping (TWTM) was performed on six wells, and fracture engineering and modeling was performed on all 16 wells in the pilot program. In addition to a general assessment of how these coals treat, the objective of the study was to optimize treatment designs in order to: 1) stay out of prolific water sands, 2) provide payzone coverage of all coal stringers, and 3) create long fractures with adequate conductivity. Fracture growth was monitored in real-time, and frac volume and injection rates were adjusted on site to avoid fracturing into any water sands.

After performing TWTM in the first well, calibrated model predictions indicated the potential danger of significant frac height growth into the overlying shore-face water sands with the planned 150,000-lb frac job. The 150,000-lb design was subsequently reduced by 50% on location after the minifrac. In addition, height growth also accelerated as pump rate was increased from 25 bbl/minute to 35 bbl/minute (original designs had planned to use a 40-bbl/minute pump rate). Therefore, pump rates were reduced to 25 bbl/minute for all subsequent treatments. Subsequently, the resulting calibrated model was generally able to predict fracture growth in future treatments. Variations from the design predictions were usually associated with variations in frac complexity and increased leakoff throughout the treatment, which is typical for complex coal formations with intermixed sands and shales. The fracture complexity used for future fracture designs was based on the average complexity observed in previous wells, since complexity can vary from well to well. Based on the initial calibrated modeling results, the completion strategy was modified on subsequent wells to include two-stage treatments in order to provide coverage of all prospective coal stringers while staying out of the overlying water sand. Figure 3 shows the height coverage of a two-stage completion, indicating that all coal stringers can be stimulated while avoiding the overlying shore-face water sand.