Refraction Monitoring Shows Promise In Heavy Oil Field

?The Peace River heavy oil field in northern Alberta is one of the largest undeveloped hydrocarbon resources in Canada. Shell has lease holdings of 147 sq miles (381 sq km) and operates thermal enhanced oil recovery at several pilot production pads to mature the technologies needed for full-scale field development.

Efficient use of steam is critical for the economics of this project, and geological heterogeneities inside the reservoir that block or divert the passage of steam are a key risk. Tracking steam as it moves from injector to producer creates opportunities to mitigate this risk by controlling injection intervals and rates, well by well, to target bypassed oil, increase oil production, and improve the oil-steam ratio.

New approach

A promising new seismic technique that could provide low-cost on-demand information about steam conformance has been developed and tested in a recent field trial at Peace River. By measuring time shifts on first arrival head waves from a refracting layer below the reservoir, the method aims to produce high-resolution areal maps of reservoir time shifts at a fraction of the cost of existing time-lapse seismic techniques.

Conventional 4-D seismic in this area often is hampered by ground roll and multiples that can be suppressed only through expensive high-fold acquisitions and careful processing efforts. The refraction method is less susceptible to these common forms of noise and requires far fewer sources to monitor a large area.

The time-lapse refraction method uses a seismic wave incident at the Debolt carbonate at the critical angle, which refracts along this high-velocity layer. Locations where the downgoing and upgoing waves intersect the reservoir are denoted entry and exit points, respectively. Debolt refractions can be seen as first arrivals in the offset range from approximately 5,544 to 13,044 ft (1,700 to 4,000 m). (Images courtesy of Shell)

During a three-month field trial in the summer of 2009, six refraction seismic surveys were acquired over a producing pad. Preliminary results show plausible one-way time-lapse time shifts in the reservoir of the order of 2 ms. However, while these time shifts are of similar magnitude to modeled time shifts, their accuracy is not guaranteed due to possible contamination by near-surface effects. New data shot in December 2010 aim to reduce the impact of this near-surface noise.

Refraction seismic surveillance methods previously have been suggested for high-velocity carbonate reservoirs acting as the refracting layer and for monitoring levees. The method uses a fast formation underlying a lower velocity reservoir and introduces a migration processing step critical for improving the lateral resolution.

The method

The proposed time-lapse refraction method works by measuring seismic time shifts on first arrival head waves from the high-velocity Debolt carbonate formation immediately beneath the Bluesky reservoir in Peace River.

Unmigrated (left) and migrated (right) synthetic refraction data were pulled from a model based on the Peace River field trial geometry.

These first arrivals essentially are free from interference with surface waves and multiples, the most severe forms of seismic noise in the area. The method assumes that the dominant time-lapse effects are confined to the reservoir. Travel time changes between a baseline and monitor survey then reflect changes in the reservoir at the locations where the head wave intercepts the reservoir layer.

In general, time shifts are composed of an entry-point time shift and an exit-point time shift that must be separated. When recorded at the surface, however, diffraction and wave propagation degrade spatial resolution. This can be improved by numerically redatuming the wavefield recorded at surface to a datum just above the reservoir. Following this approach, excellent resolution has been confirmed on synthetic data, both for 2-D and 3-D geometries. An implementation based on Berryhill’s algorithm was used for redatuming, and time shifts were measured on the imaged data using cross correlation. Vertical resolution currently is not achievable with the refraction method.

The Debolt refractions are visible as first arrivals over a distance of more than 1.2 miles (2 km), starting approximately one mile (1.7 km) away from the shot. The area that can be probed with a single source is more than 30 times larger than that probed with conventional reflection seismic methods, where only offsets up to the critical angle are used. This allows for use of far fewer sources to monitor large areas, resulting in substantial cost reduction and a smaller environmental footprint.

Well-repeated surveys and robust processing methods to mitigate any time-lapse effects originating outside of the reservoir are essential for successful application of the refraction method. Burying the receivers and using cased shot holes or permanently installed seismic sources can reduce the impact of near-surface seasonal variations. Time shifts due to geomechanical effects in the overburden can be significantly more challenging to mitigate but are not expected at the pad chosen for the field trial.

Field trial

The production pad where the field trial took place comprises eight rows of horizontal wells being produced via the mega-row steam injection process, whereby one row of wells injects steam for one to two months while all other rows are either soaking or producing. This pattern then moves along one row until the cycle is complete. At the start of the field trial, the pad was in its third cycle, and some uncertainty was associated with the expected magnitude of time-lapse signals given the lack of a presteam baseline survey.

During a three-month period, while steam was injected in the northernmost row 7 and row 8 wells, a baseline and five repeat refraction surveys were acquired. The acquisition geometry showed the nominal refraction entry and exit points as determined by ray tracing through a velocity model derived from well logs. Each survey consisted of 205 dynamite charges fired in cased shot holes and recorded into 782 buried receivers. The geophones were encased in bentonite at 39 ft (12 m) and supported additional reflection seismic reference surveys and other geophysical experiments. Shot locations were chosen along existing line cuts to minimize environmental impact.

The geometry of the refraction seismic field trial in Peace River is shown on the left, and a shot gather is shown on the right. The green points indicate the approximate location of entry and exit points. A road, a stream, and the well pad create holes in the data.

Shot hole deterioration was found to cause significant changes in source wavelet from survey to survey. If not carefully handled, this effect easily could overshadow time-lapse signals of several milliseconds. To compensate, each shot record in each monitor survey was matched to the corresponding baseline shot record prior to time shift migration and measurement. Matching filters were derived on a strong direct arrival propagating through the overburden and not affected by reservoir changes. Matching direct arrivals rather than refracted arrivals can impact results due to their differing kinematic properties.

Maps of imaged reservoir time shifts from the field trial and corresponding reservoir predictions show a weak pattern of seismic slowdown emerging over an area in the northwest three weeks after injection in row 8 (left panels). Significant slowdown is observed in the vicinity of row 8 (Aug. 25) near the end of the row 8 injection cycle. Dark blue colors indicate speedup, perhaps due to pressure depletion relative to the baseline survey.

Results

Preliminary maps of reservoir time shifts relative to the June 25 baseline showed two monitor surveys. When the baseline was acquired, row 7 was near the end of its two-month injection cycle. The July 23 monitor was acquired three weeks after the injection had moved up to row 8 and showed a faint pattern of positive one-way time shifts (i.e., seismic slowdown) emerging in the northwestern part of the field. On Aug. 25, near the end of the row 8 injection cycle, these one-way time shifts had grown to nearly 2 ms, while negative time shifts indicative of increased reservoir velocities dominated in the south and east.

A preliminary comparison with synthetic time shifts from the reservoir model suggested the large-scale patterns might be caused predominantly by changes in reservoir pressure, while more localized effects are interpreted as changes in gas saturation. A detailed interpretation of the time shifts carried uncertainty due to near-surface effects. New data have been acquired with receiver stations outside of producing areas, and Shell hopes that an improved matching can be performed with these new data.