EGS Collab Experiment 1 Long Term Circulation Test
From Open Energy Information
Contents
Introduction and Objectives
A long-term chilled water circulation test was performed within EGS Collab Exp. 1 from early 2019 to early 2020. Previously, the testbed had been created by stimulations in 2018, which created flow pathways among several wells, particularly the injection well (E1-I) and production well (E1-P). The primary objective of the flow test was to provide data to validate computer models concerned with flow processes and heat exchange processes in EGS.
Because the rock body in the test bed is not “hot”, approximately in the low 30’s °C, the test used chilled water to achieve a temperature difference between the host rock and the circulated fluid.
The circulation rate was targeted at 400 ml/min. In the various tests performed in 2018, particularly the test in late October 2018, the team found that a higher flow rate, such as 800 ml/min, could alter the testbed by extending or creating fractures. It was therefore hypothesized that a circulation rate of 400 ml/min can achieve a pseudo-static state test. The test data seem to support this hypothesis as no microseismic activities were observed and the flow pathway seemed to be stable.
Pre-test modeling work assuming a penny-shaped fracture connecting the injection and production wells indicated that it could take years to observe a significant thermal breakthrough at the production well. However, the team recognized that the penny-shape fracture model tends to be a best-case scenario in terms of delaying thermal breakthrough. The actual system is likely affected by flow channeling which tends to accelerate thermal breakthrough and could make it observable in a manageable time window.
Design-phase modeling results from the LLNL modeling team. Left, temperature distribution on a penny-shape fracture after 124 days of fluid circulation at 400 ml/min. The injection well is placed at the center of the fracture. Injection temperature was assumed to be 5°C. Right, production temperature prediction for three different circulation rates.
Preparatory tests
A viable flow pathway was not established in the 2018 tests. Three intervals in the injection well had been stimulated in 2018. The 164 ft (nominal depth) interval, stimulated in May 2018, was the most promising candidate. However, a preliminary flow test in late October to late November 2018 suffered from seemingly unbounded pressure increase and low fluid recovery ratio. It was later speculated that certain geochemical or biological processes had caused clogging in the flow pathway, which could be mitigated by proper treatment of the injected water, giving hope to viable circulation testing.
This flow path, starting from the 164 ft interval in E1-I, was again tested in February 2019. In the first round of testing, from Feb. 4 to Feb. 7, the injection pressure appeared stable, but the fluid recovery ratio was disappointing. This round of the test ended with exploratory testing such as targeting other intervals, moderate stimulation, and injection into the production well, but the results were still not satisfactory.
Several short-term injection tests were performed from Feb. 12 to late March. Consistent, repeatable outcomes started to emerge. Although the fluid recovery ratio, mainly from well E1-P and E1-PDT, was still less than 50%, the repeatability of the response was still encouraging. A discussion of these “repeatability tests” and interpretations of some of the observations are available in https://pangea.stanford.edu/ERE/db/IGAstandard/record_detail.php?id=29250
Injection and fluid recovery in the series of tests beginning in late February 2019 and leading to the long term circulation test. The upper panel shows the injection pressure (blue) and injection rate (green). The lower panel shows the fluid recovery rates from several channels. Because the part of the purpose was to test the system, some repurposing of the recovery rate channels took place in this time window, of the labels are only nominal.
Overview of observed behavior
Largely continuous circulation started on March 27, 2019 and was wrapped up in February 2020. Injection was into the 164-ft interval in E1-I. Fluid recovery was mainly from two intervals in E1-P (one connected to the OT-P Connector and one to the main hydraulic fracture), E1-PDT, and E1-PST. Flow from E1-PDT was initially significant but diminished to a low level. Flow from E1-PST remained at a low level. Water recovery from E1-P (two intervals combined) continuously increased and eventually reached more than 90% of the injection rate. Note that water also leaked out of the testbed from several other places. There was flow from E1-OT but it was mostly dripping. Sometimes the sensor registered high values at OT but those were mostly caused by equipment malfunctions such as tube clogging. Water was also dripping from certain locations on the drift wall.
The packer system (two packers) in E1-P straddled across the intersection with OT-P Connector (approximately 122 ft depth). The intersection(s) of E1-P with the main hydraulic fracture(s) was at 127 to 129 ft, below the lower packer. Therefore, the test system made measurements and collected fluid from two locations. The interval between the two packers was customarily known as PI (Production-Interval). The segment below the lower packer was customarily known as PB (Production-Bottom). Fluid in each interval, PI and IB, was conveyed to the well collar through tubing. Temperatures were measured by thermistors placed near the intersections with the fractures.
Illustration (made by Pengcheng Fu of LLNL) of packer locations in E1-P during the long term circulation test. The packer interval was increased to 45.5 inches according to the daily shift report from April 1, 2019. https://openei.org/wiki/EGS_Collab_Daily_Shift_Reports_April_1_2019
An illustration of geometrical relationships between the main hydraulic fracture, a natural fracture named OT-P connector and various wells in the circulation system. The conceptual model and the illustration are from Wu et al. (2021). Color in the background represent the initial temperature field on two slice planes cutting the testbed.
A chiller to reduce the injection water’s temperature was started on May 8, 2019. Injected water thereafter was mostly chilled except for a few short interruptions due to hardware or power issues. The chiller circulated chilled water (not the water to be injected) in a closed loop. The closed-loop includes tubing running down well E1-I and turning back near the packer. Therefore, the part of E1-I between the well collar and the packer housed the “heat exchanger”. This configuration caused this segment of E1-I, approximately 50 m in length, to have been chilled as well. As the wellbore served as a “negative heat source”, radial heat conduction around E1-I manifested into cooling signals in several wells as observed by DTS. See detailed description and modeling in Wu et al. (2021). Injection fluid temperature, as measured by a thermistor in the injection interval, was maintained at approximately 12°C most of the time. The designer of the flow system had expected better chilling performance (i.e. lower injection temperature) and made additional improvements to the system, but was unable to reduce the temperature further.
Injection pressure under the constant injection rate slowly increased, increasing by nearly 1000 psi over the 10 months of circulation. This increase has been attributed to the poro-mechanics effect by the team, meaning that the pressure diffusion into the rock matrix, through Darcy flow and/or flow through natura fractures, caused the total in situ stress to increase, thereby demanding higher fracture-opening pressure. The injection of chilled water caused a visible quick pressure decrease and then the slow pressure increase pursued. A peculiar observation was that when the pump was stopped momentarily and restarted soon, the injection pressure usually experienced a small decrease. The reason for this decrease is still unclear.
Temperature measures in E1-P
As mentioned earlier, fluid temperatures at two locations in E1-P were monitored by two thermistors. A thermal breakthrough was expected to manifest into measured temperature decrease in these two locations.
When the two thermistors were put in place, the measurements aligned with expectations: around 30°C and PB temperature higher than PI due to the deeper location. After an interruption to injection from April 4 to April 5, the temperatures at PI displayed rather large changes, increasing to 33°C and then decreasing to 32°C on April 11. Largely continuous temperature measurements at these two locations were made between April 17 and November 11. During this period, measured temperature at PB gradually decreased, which was interpreted to be thermal breakthrough, whereas measured E1-PI temperature gradually increased, which was speculated to reflect flow path evolution within the OT-P Connector. The observed “thermal breakthrough” at PB was much sooner than expected so the Collab team did various analyses and modeling to reconcile with the observation, including invoking extreme flow channeling.
PI and PB temperature measurements when the two thermistors were initially placed (left) and when the thermistors were replaced (right).
However, an inspection of the two thermistors in early November 2019 revealed that they might have been damaged. A new thermistor with an improved design was installed at PI in early November 2019 and a new thermistor for PB was only available in mid-December 2019. Based on an analysis of the data, the team concluded that temperature measurements at both locations were likely to be questionable between April 4 and the replacements of the thermistors, as indicated by the dotted line segments in the attached. The measured temperatures at E1-PB and E1-PI after December 2019 were slightly higher than those measured on April 3, meaning that thermal breakthrough at the production well were not observed during the water circulation test. Thermal breakthrough modeling in Wu et al. (2021) indicates that the outflow temperature at PB could have decreased by a fraction of 1 °C, which was likely obscured by the temperature increase caused by the Joule-Thomson effect at the fracture-well intersection.
Tracer tests
Several tracer tests were performed during the circulation tests.
Endgame for EGS Collab Experiment 1
Injection was finally shut-in on February 5, 2020, marking the end of the long term circulation test. Various other tests were performed afterwards, including circulation with backpressure application and attempts to circulate at higher rates. Those were beyond the scope of this wiki page.
Data release and reduction
Raw data and useful documents for the low term circulation test were released through
https://gdr.openei.org/submissions/1254
Reduced data, as explained below, as well various Python scripts to process/reduce the data
Despite the unique value, the data generated by the long term circulation test are difficult to use.
First, the circulation test generated a large amount of data. The flow system collected data from ~100 channels at an output rate of 10 s (occasionally 1 s), generating approximately 3 million rows of data, stored in more than 100 CSV files with a total size of 3 GB.
Second, the correspondence between physical quantities and data collection channels is very complicated for several reasons: (1) Two pumps were used: a Quizix pump and triplex pump; (2) Some channels were added in the middle of the test; (3) Occasionally, mistakes in connecting the sensors were made after system repair or reconfiguration; (4) Some sensors were later found to be faulty.
For these reasons, the experiment team (mostly by Pengcheng Fu of LLNL) created several products in an attempt to reduce the data to easier-to-consume forms.
Raw channel plots
Data streams from selected channels are plotted in a multi-page PDF file. Each page includes one month of data. The nominal meanings of these channels were annotated on the plots while the meanings were not always accurate due to the reasons mentioned above. The Python script used to generate these plots is included in the GDR release. Note that the channels are groups based on similarities, namely, flow rates are grouped together and temperatures are grouped together. The plots are visually messy due to the lack of further processing.
One page in the raw channel plot.
Single-quantity streams
Nine quantities were selected to generate single-quantity stream files. These nine quantities are:
1. Injection Pressure
2. Injection Rate
3. Flow rate out of E1-P Interval
4. Flow rate out of E1-P Bottom
5. Flow rate out of E1-PDT
6. Flow rate out of E1-PST
7. Injection Interval Temp
8. Temperature at E1-P Interval
9. Temperature at E1-P Bottom
Data stream for each quantity is organized into a single CSV file. The user would not need to worry about or even understand which sensors were used to collect the data. All the embedded notes (stored in the “metadata” column of the original CSV files) in the original CSV files are included in every stream file, regardless of whether they are useful for analyzing the current stream or not. The name of the original files and the channel names are also included as notes. Because the PI and PB temperature thermistors had issues as explained above, we store the readings before the thermistor replacement and after the replacement into separate files, named as “_Questionable” and “_Upgraded”, respectively
Notes in one of the stream files.
The Python script used to generate these steam files is included in the GDR release. The channel switching and skipping of meaningless segments (from diagnostic testing or interruptions) are explicitly “documented” in the script as follows.
A segment of Python script noting the channel switching for injection rate.
Hour-interval streams
The single-quantity streams, in 10 s or 1 s intervals, still resulted in large files, with up to more than two million lines of record each. The data have been processed to generate hour-interval data streams for easy handling by prospective users. To resample the data, each stream is first divided into separate, continuous segments. Within each segment, the first data point and the last data point are preserved. The data in-between are resampled by averaging values within hour-length intervals. After reduction, each stream only includes several thousands of lines. The script to resample the streams is also included in the GDR release.
Scrips as documentation
In the data processing mentioned above, all scripts were released along with the processed data. This reflects the philosophy of “scripts as documentation”. The two main benefits of this approach are: (1) The prospective users would know what have been done to the data. Some subjective judgements were exercised and ad-hoc treatments applied in processing the data. The scripts serve as permanent records of these treatments. (2) The scripts serve as examples and instructions on how to use the raw data.