There may be many reasons for the success geoscientists recently reported at Salt Creek, Wyo. – but a big one was the result of data collected from a 4-D seismic survey, which revealed fine interval resolution as well as identified flow patterns that cannot be seen from well data alone.
That’s the report of Houston geophysicist John O’Brien, a geophysical adviser and team leader at Anadarko Petroleum Corp., who earlier this year discussed the findings at the annual 3-D Seismic Symposium in Denver.
There, O’Brien presented a time-lapse study that monitored a CO2 flood with fine time steps in Salt Creek, a mature oilfield located in Wyoming’s Powder River Basin.
The first well was drilled there in 1908. Since then the field has been produced under primary production, waterflood and, most recently, a tertiary CO2 flood.
“At Salt Creek, it was a shallow reservoir, and a lot of oil wells had been drilled there over the years,” he said. “There was a small injection pattern and really tight well spacing.”
All of which made the 4-D seismic survey results of note: Features were consistent from survey to survey and also resulted in fine interval resolution.
“We’re convinced we’re seeing strong time lapse responses associated with CO2,” he said.
The study also resulted in more details and higher confidence in observations and interpretation. By characterizing fluid flow in the reservoir, the team was able to see flow patterns that cannot be seen from well data alone, he said.
Study Details
The baseline study was acquired in March 2008, O’Brien explained. Then CO2 was injected and five surveys were conducted, with the final one completed in July 2009.
“We wanted to progressively watch it move over time and used multiple monitors,” he said. “We’re all trying to get the most out of our seismic data as possible.
“How can we improve the time lapse seismic on reservoirs?” he asked. “We took multiple flash shots so we could see the fluid flow in the reservoir.”
To manage the CO2 flood, a time-lapse seismic program was designed over a portion of the field, and a semi-permanent monitoring system with geophones was set up at a shallow depth.
The CO2 flood was designed with a small injector pattern of 20 acres. The separation between wells in a pattern was 600 to 700 feet, requiring high spatial resolution for time-lapse monitoring on the scale of the interwell spacing, he said.
The CO2 was expected to advance rapidly over these distances, calling for short time periods between monitor surveys.
For the program, geophones were deployed to a depth of 18 inches. Once deployed, the geophones remained in place for the complete monitoring program.
“Then five monitor surveys were recorded at intervals of approximately three months,” he said. “Each survey was recorded in approximately four days.”
The survey imaged 45 contiguous injection patterns consisting of a central injection well for each surrounded by four production wells covering about 20 acres.
“We were looking for high quality seismic data at a fairly shallow target,” O’Brien said.
Observations
With multiple images recorded at different times, the team was able to observe the initial flood at each injector, follow it as it expanded over time and characterize it spatially.
“Data can be acquired repeatedly in a very efficient manner,” O’Brien said. “We also can evaluate the consistency of trends observed on multiple surveys and assess the level of confidence in our observations.”
This design led to a true time lapse monitoring of progressive changes in the reservoir rather than the common practice of a couple or a limited number of surveys taken over a longer time period to obtain a before-and-after comparison, he said.
Indeed, if there was only a baseline and a final study recorded 16 months later, “we could identify areas with higher or lower sweep efficiency, but would not gain the same insights into the mechanisms driving these differences,” he said.
O’Brien noted the study’s results showed a breakthrough at some wells with the CO2 produced, but not at all of them.
“There is a significant amount of variability,” he said. “The pattern response was not uniform and there were differences from pattern to pattern.
“Injected fluids do not flow predominantly in a radial direction from injector to adjacent producers,” he concluded. “Instead, time-lapse imaging demonstrates a significant degree of asymmetry indicating a strong flow component in the up-dip direction with lesser sweep in the strike and down-dip directions, most likely related to CO2 buoyancy.”
He said fluid flow cannot be characterized as a symmetric five-spot pattern that serves as a unit cell to describe the flood.
“We now recognize the relationships between adjacent patterns and fluid flow across pattern boundaries,” he said.