Difference between revisions of "Experiment 1"

From Open Energy Information

Line 5: Line 5:
[[Shift Reports|<span class="geoTab">Shift Reports</span>]]
[[Shift Reports|<span class="geoTab">Shift Reports</span>]]
[[Data|<span class="geoTab">Data</span>]]
[[Data|<span class="geoTab">Data</span>]]
[[Literature|<span class="geoTab">Literature</span>]]

Latest revision as of 19:27, 26 March 2020


Fig. 1. Schematic of wells for Experiment 1 along the West Drift on the 4850 level of SURF. The green line represents the stimulation (Injection) well (E1-I), the red line represents the Production well (E1-P), yellow lines represent monitoring wells, and orange lines represent kISMET wells.

In Experiment 1, we are stimulating and performing flow tests in ~1500 m deep (4850 feet) phyllite at the Sanford Underground Research Laboratory in Lead, SD. We are quantifying permeability enhancement, characteristics of the stimulated rock, determining the nature of stimulation in crystalline rock under reservoir-like stress conditions and generating high-quality, high-resolution, diverse data sets for model validation. In addition, we are testing and developing monitoring techniques under controlled conditions to allow selection of technologies appropriate for deeper full-scale EGS sites.

The EGS Collab project constructed an intermediate scale (~10-20 m) testbed at the 4850 level of the Stanford Underground Research Facility (SURF) in South Dakota for testing and validating fracture stimulation and flow/transport models. This testbed consists of eight ~200 ft (~60 m) HQ-diameter (9.6 cm) boreholes that are drilled into the crystalline rocks of the Poorman Formation from the West Access Drift tunnel. Of the eight boreholes, one borehole is used as an injection/stimulation well, while another sub-parallel borehole located about 10 m away from the injection well is used as a production well, and rest of the other boreholes are used as geophysical/fluid sampling monitoring wells. Hydraulic stimulation activities were conducted at three locations along the injection hole in an attempt to create direct fracture connections to the production hole. A flow system has been established between injection and production boreholes through a set of hydraulically stimulated fractures propagated from a notch located at 164 ft in the injection hole.

Although we planned for a single production well, a set of natural fractures in the testbed is believed to have intersected the stimulated hydraulic fractures and provided additional flow paths for water to be transported to the drift through multiple monitoring boreholes and weep zones. Since late October 2018, multiple tracers were injected into the fracture system at the 164 ft location that involves both stimulated and natural fractures, and tracers were injected into the fracture system from multiple locations in nearby wells and weeps.

Tracer Tests

Several tracer injection campaigns were completed at the testbed from October 2018 to November 2019. Non-reactive (conservative), reactive (sorbing), and thermally degrading tracers were injected, either exclusively or in combination, and their recovery rates from multiple producers were collected and analyzed. Each tracer test was injected at the 164-notch and targeted characterization of hydraulic fractures that might have connected to natural fractures providing flow paths to E1-P, several monitoring holes, and drift E1-I.

The activities to create hydraulic fracture(s) at 164-notch were commenced during May 22-25, 2018. Previously, we showed that the hydraulic fracture(s), two natural fracture zones, (OT-P Connector and PDT-OT Connector), E1-P, and at least 3 monitoring holes (E1-OT, E1-PST, and E1-PDT) were involved in defining dominant flow paths at one time or the other during the flow tests (Figure 2). Also, the Shallow fracture zone/weep depicted in Figure 2 indirectly served as a flow path during flow test at 164-notch. After the top section of the E1-OT was sealed, water leaking to this hole at depth moved up and diverted into the Shallow fracture zone, and ultimately seeped to the drift as a weep near E1-P.

All tracers were conducted with a nominal injection rate of 400 mL/min. Starting late October 2018, the testbed was under a constant flow regime with some interruptions. When there was an interruption in injection prior to a tracer test, a steady flow regime was established before injection of tracer solution. Starting early May 2019, the testbed was subjected to a chilled water injection although there may have been some brief durations where regular mine water was injected due to mechanical failures of the chiller(s).

Table 1. Tracer Injection Dates, Types, and Recovery Location
Tracer Injection.png

Injection Mechanisms

During early tracer tests, a large volume of relatively low-strength tracer solution was injected using a Quizix pump. This injection mechanism was initially modified by directly pulling a higher strength shear tracer solution into one of the cylinders of the Quizix pump or pulling the tracer solution in the injection line hose. Finally, a dedicated ISCO pump was installed and used to push tracer solutions directly into the main injection flow line. With the exception of the early October 2018 tests, a 5-minute injection pulse of 100 mL high-strength tracer solution was transmitted at a rate of 20 mL/min into the main flow line. Immediately after the injection of the tracer solution, 100 mL of clean water was injected using the same injection mechanism (20 mL/min) as chase water. During the tracer and chase water injection period, the main injection flow rate was 420 mL/min. The chase water was not injected during the last two tracer tests (October 22, 2019 and November 19, 2019). In these tests, the ISCO cylinder was rinsed and the rinse water was collected and analyzed for tracer that was stuck in the ISCO pump cylinder and connecting tubes. This mass was used in tracer injection-recovery mass balance equations.

Fig. 2. Natural fracture zones (or weeps) in the testbed. At least five fracture zones or weeps are identified in the testbed.

Tracer Cocktails

A suite of tracer compounds, including both conservative and sorbing tracers, was used for the tracer work in the testbed. Earlier tests included different strains of synthetic DNA; however, use of synthetic DNA was discontinued in later tests. In general, all tracer cocktail solutions were prepared so that they should include at least one fluorescing tracer for near real-time detection in the drift. Rhodamine-B, fluorescein, C-dot, and phenol acetate were used as fluorescing tracers. The ability to detect tracer in the drift not only provided real-time analysis of the tracer breakthrough data at multiple producers but also helped in modifying the sampling strategy and experiment operation in real time.

The C-dot is a nanoparticle (3-5 nm in diameter) tracer that consists of a carbon core decorated with a highly fluorescent polymer. The phenol acetate was used as a thermally degrading tracer, with phenol as a degrading product. In many tracer tests, the tracer cocktails also included solute tracers such as Cl, Br, and K (Table 1).

Sample Collection

Immediately after injection of tracer solutions, high frequency samples were collected from major producers (Table 2). The E1-P (production well) below the lower packer (PB) and E1-P interval (PI) flowed consistently during all tracer tests and was sampled for tracer recovery. However, some leaky monitoring wells (e.g., E1-OT, E1-PST, and E1-PDT) flowed intermittently. For example, a leak on E1-OT was significant during the 2018 tests. Repairs at the end of 2018 fixed this issue such that OT flow was not significant during the 2019 tests. All producers with significant flow were regularly sampled and analyzed for tracer recovery.

We adopted two methods to collect liquid samples from producers for analyses. The first method was manually collecting 'grab' samples in 10 mL amber sampling tubes. Typically, these samples were each collected in less than a minute, and were subsequently capped and labeled with the collection location and sampling time. Immediately after the collection of these samples, the fluorescing tracer included in that particular tracer injection cocktail was analyzed in the drift. When phenol acetate was also included in the tracer cocktail solution along with C-dot, the analysis of the thermally degrading (kinetically controlled) product was conducted first, followed by the analysis of the co-injected C-dot.

The second method of sampling was using fraction collectors to collect samples from the three locations. The fraction collectors were set to advance at a specified time interval and a peristaltic pump was used to control the rate of sampling such that the 10 mL sampling tube would be filled during the sampling interval. Unlike the grab samples, which represent a sample concentration at the moment of sampling, the fraction collector samples represent an integrated sample concentration over the sampling interval.

Table 2. Injection and Outflow Rates of Producers During 2019 Tracer Tests
Flow Rates.png

Tracer Analysis

All fluorescing tracers [C-dots, fluorescein, rhodamine-B, and phenyl acetate (phenol as degrading product)] were analyzed using an Ocean Optic spectrophotometer system (Ocean Optic FIA-SMA-FL-UTL cell, PX-2 pulsed xenon lamp, QEPRO spectrophotometer). For each fluorescing tracer, the optimum excitation and emission radiation wavelength was identified with a tracer solution prepared in background water (pre-tracer injection PI water). For each sample, three scans of average fluorescence count were recorded with a 5 to 30 second integration time for each scan. Background fluorescence values were established for each producer and corrected either during analysis (for PI samples) or during data processing. Prior to each analysis, a series of calibration standard solutions of the target analyte were prepared using the background (PI) water for construction of the calibration curve. Consequently, the recorded fluorescence count of each sample was converted to tracer concentration using the relationship between fluorescence counts and concentration as illustrated in Figure 3.

Table 3: Excitation and Emission Wavelengths
Tracer data.png

For solute tracers (Cl, Br, K, etc.), all time-stamped samples were shipped to a laboratory for analysis with ion chromatography (IC) and inductively coupled plasma optical emission spectroscopy (OCP-OES). Some of the samples were analyzed at the Center for Advanced Energy Studies (CAES) in Idaho Falls, ID whereas some additional samples are currently being analyzed at LLNL in Livermore, CA.

Fig. 3. C-dots calibration curve constructed for July 24, 2019

Tracer Arrival Time Adjustment

Initial tracer detection at the production wells at the EGS Collab testbed occurred in less than an hour. Such a short duration test allowed us to account for all possible time delays associated with the injection and sampling. Universally for tracer tests and producers, one common time delay was the time taken by tracer pulse to get to the injection point (164-notch interval). Given the flow rate (400 mL/min) and length of the tube from the drift to the injection target, we found this time delay to be about 3 minutes.

Similarly, we also found the travel time for water from where it left the rock/fracture and entered the outflow drain tubes at depths (e.g., for PB and PI) to the sampling location on the drift. Since the outflow rates of both PB and PI varied over time, the time delays for PB and PI samples were different for different tests. For PB and PI, the out bound time delays ranged about 9 to 32 min and 6 to 65 min, respectively, and we adjusted respective time delays for each sample from each test. On the other hand, for outflowing monitoring wells (e.g., E1-OT, E1-PDT, E1-PST, etc.) the out-bound time delays were unknown (water travelled through the grouted hole with unknown flow channel length/volume) and such time delays were not adjusted. Finally, when the fraction collector was used to collect samples, the time delays associated with the travel time for water from the producer outlet to the drip point were also adjusted in the reported time. Unlike the grab sample with instantaneous sampling at an associated time stamp, the fraction collector collected water samples over a duration (usually a 20 minute duration). For these samples, a mid-interval time stamp was used as sampling time.

Results and Discussion

Fig. 4. Injection pressure and rate, and production rates during the cold-water injection test (modified figure from Mark White, PNNL)

Much of the tracer characterization work of 2019 was to characterize changes in the fracture network due to the injection of water that was colder than the initial rock temperature. The chilled water experiment started May 8, 2019 (10:15 MDT, 17:15 UTC) with the circulation of chilled water through the down-borehole heat exchanger within E1-I, cooling the temperature of the water injected into the hydraulic fracture at the 164’ (50 m) notch for over 7 months of nearly uninterrupted circulation. Outside of system outages, the chilled water injection rate was maintained at 400 ml/min. Throughout this experiment a straddle packer was located over the intersection of the OT-P Connector fracture and E1-P at a depth of 121.75 ft (37.1 m) from the borehole collar, allowing for the recovery of water from the region within the straddle packer interval (E1-PI) and below the interval (E1-PB), where the hydraulic fracture intersects the E1-P borehole. Water recovery was recorded during the chilled water experiment. Water flows from all of the metered locations were noted throughout the course of the chilled water experiment (Figure 4). At the beginning of the cold-water injection, water production was dominated by PI and PDT. During the cold-water injection, water production increased in PI and PB and decreased in PDT. Near the end of the experiment water production was predominately from E1-PB and E1-PI (> 80%), with small amounts from E1-PST and E1-PDT. Volumetric recovery from all the producing zones approached 98% near the end of the experiment.

The Steady State Assumption

Tracer breakthrough curves for April 25th, May 1st (prior to thermal injection), July 24th and October 22nd are plotted below for wells PI, PB, and PDT as a function of volume of water produced during each tracer test. Plotting the C-dot concentration normalized to the injection concentration as a function of volume is more appropriate since the water recovery rate for each production location changed during this 7-month period. The tracer initial arrival and peak concentration as a function of produced water volume do not exhibit a consistent trend with time during the cold-water injection test. This is believed to be partially due the dynamic nature of the fracture system in response to the injection pressure potential changing flow pathways during this test. During this time, numerous sharp pressure drops were noted in the injection pressure which is believed to represent the creation of new fractures as the injection pressure reached a critical value. These new fractures appear to affect the production rates, both positively and negatively, of the two main production zones (PI and PB) and to a lesser extent PDT.

Fig. 5. C-dot tracer break through curves plotted for PI, PB, and PDT
C-dot breakthrough.png

Although there are overall positive and negative produced water rate changes over the duration of the cold-water injection, the numerous injection pressure drops are frequent enough such that individual production wells experience several episodic increases and decreases in their water production between tracer tests. As a result, comparison of individual tracer tests becomes much more complicated and the changes seen in the tracer break through curves are both due to the creation/change of fractures by pressure as well as the injection of cold-water creating thermoelastic effects.

Table 4. Volume of Produced Water (L) for the Initial Detection of the C-dot Tracer and the Peak Concentration for PI, PB, OT, and PDT
Produced water.png

Fig. 6. Injection Pressure Perturbations and Resulting Produced Water Flow Rate Changes as a Function of Time.
Red vertical lines indicate time when a C-dot tracer injection test was performed (modified firgure from Mark White, PNNL)
Inj pressure.png

Tracer Recovery Mass Balance

Nine tracers were conducted during the cold-water injection test up to November 19th, 2019. Two additional tracer tests were conducted in January 2020. April through July tests included a 100 mL "chase" clean water injection in an attempt to clean the pump and injection lines of residual tracer. October and November tests did not use the chase water injection protocol and instead back-flushed the injection line and pump into a graduated cylinder to record the volume and measure the tracer concentration, allowing for an adjustment of the tracer mass injected. Subsequent analysis of the tracer in the back flushed water suggests that 10 to 20 percent can be trapped in injection lines and the injection pump. It is not certain that a 100 mL chase water flush will inject this residual tracer into the injection interval. It is the authors' opinion that the back-flushing method produces a better pulse input than the chase water method.

Table 5 lists the tracer mass balance and water mass balance for each of the production wells. Whereas greater than 75% of the injected water is consistently recovered from the production zones, only approximately 33% of the tracer is recovered with this produced water. Several reasons could account for this discrepancy:

  1. The tracer mass is calculated from the tracer break through curves which do not encompass the complete concentration v. volume curve thereby biasing the calculated tracer mass.
  2. Delayed tracer break throughs after sampling is complete are not accounted in the total tracer mass recovery.
  3. Water is recovered in the production wells that did not originate from the injection well.
  4. The tracer is irreversibly absorbed or filtered in the system.
Table 5. Tracer Mass Balance Calculated form Integration of Break-through Curves and Water Mass Balance Calculated from Flow Rate.
Both values are expressed as a percent of mass of tracer injected and injection flow rate.
Tracer mass.png

The magnitude of incomplete breakthrough curves can be investigated by fitting equations to the tracer concentration v. volume data and re-calculating the mass balance. Sampling for longer periods of time can assess delayed break through due to longer pathways. Distinguishing between the collection of non-tracer laden water and filtering/decay/absorption is somewhat harder.

Fig. 7. Plot comparing the percent by mass of C-dot recovered of the initial injection mass to the percent of water recovered by flow rate to the injection rate (blue data = PI, orange data = PB, and grey data = PDT).

The figure to the left represents a method to determine the source of water produced at the three major water producing zones (PI, PB, and PDT) during the long-term cooling test. In this plot, the percent of C-dot mass recovered normalized to the total injected mass is plotted against the percent of water recovered normalized to the injection flow rate. As seen in the figure, the percent of tracer recovered to the percent of water produced of all three wells exhibit a linear relationship. The linear relationship between tracer recovery and water recovery can be interpreted to support two conceptual models. The first model assumes a constant inflow of water that does not include tracer. In this case, the remaining water would have a tracer concentration percent equal to the slope of the linear regression in the figure.

Examining the produced water electrical conductivity as a function of flow rate may provide additional information to distinguish between the two conceptual models described in the previous paragraphs. If the non-tracer water is proportional to the tracer laden water, then the electrical conductivity should be a constant regardless of the flow rate. If the non-tracer water is a constant (or near constant) and the flow rate is controlled by the tracer laden water, then the conductivity should change as a function of the flow rate. For this field experiment, the injected water has a conductivity of approximately 490 μS/m (measured on 01/28/2020) whereas the non-tracer water is believed to be higher.

The plot of electrical conductivity vs flow rate for PI suggests that the electrical conductivity of the PI produced water decreases with increasing flow rate. A mixing cell model fit of the day which optimizes the background water electrical conductivity and inflow rate suggests the following data: non-tracer water electrical conductivity 2000 mS/m with a flow rate equal to 27 ml/min. These values appear to be fairly reasonable to what we might expect in the field. The regression intercept of figure to the left would intersect zero tracer concentration at approximately 52 ml/min of input of non-tracer water. Based on these results we can consider the input of non-tracer water to be nearly constant; however, it is likely that this input is not a complete constant but exhibits a small amount of variability as a function of flow rate.

The mixing equation is expressed as: Mixing eq.png where ECtotal is the measured electrical conductivity, ECinj is the injection water electrical conductivity, Qinj is the injection rate, ECbkgnd is the non-tracer water electrical conductivity, Qbkgnd is the fitted non-tracer flow rate, and Qtotal is the measured production rate.

Fig. 8. Plot of the Production Well Interval Flow Rate as a Function of the Produced Water Electrical Conductivity

Recent sorption isotherm test results suggest that C-dots do not behave as a conservative tracer and exhibit a Langmuir type sorption to crushed Poorman Formation rocks. It is unkown at this time if this sorption is irreversible.

Based on geophysical monitoring, fiber optic temperature measurements, water flow production response, core descriptions and well bore analyses, a simple flow conceptual model of the flow to the three dominate producing wells is illustrated in the figure below. Injected water is believed to be flowing from the 164-notch through a single (or multiple) fracture(s) towards the production well. Some of the injected water (3 to 7%) is diverted to the PDT well through a natural fracture and then the water travels through a grout-filled borehole to the drift. The pressure at the intersection of the natural fracture and the PDT borehole is not known. The majority of the injected water (70 to 84%) continues through the hydraulic fracture until it bifurcates into two pathways. One pathway is along the hydraulic fracture that continues to intersect the production well below the lower packer (PB). The other water pathway flows through a second natural fracture to the production well between the packers (PI). For both these production zones, water is transmitted to the drift through ¼ inch stainless steel tubing resulting in a well pressure near atmospheric pressure. Certainly, the flow pathways are more complicated than this simple conceptual model but the model provides an easy layout of the water flow pathways that can explain most of the flow and tracer behavior.

Fig. 9. Major Outflow Patterns Observed During July 24, October 22, and November 19 Tracer Tests


Both fracture cooling and pressure perturbations appear to have affected the fracture flow pathways of the Experiment 1 test bed during the cold-water test. Overall the water production rates of the two production well zones have significantly increased during this time. However, the increase has not always been constant and due to sudden injection well pressure decreases, water produced from these two locations has exhibited temporary decreases in the water production rate.

PDT seems to be minimally affected by these changes during this test (see Table 4). One interpretation is that the lack of changes in the PDT tracer’s first and peak arrival volume parameters, despite 7 months of cold-water injection and numerous pressure step changes, suggests that any changes to the fracture system are not near the injection well hydraulic fracture and the natural fracture leading to PDT.

PI and PB water production rates appear to be generally inversely correlated and likely share a common flow pathway from the injection well. Although there is an overall increase in both PI and PB water production rates during the 7-month test, the two production zones compete for the same injected water.

Analysis of the C-dot recovery vs water recovery for the three main production zones suggests that there exists a linear relationship between the two parameters. This linear relationship could be due to a constant input of non-tracer laden water where any change in the total flow rate is due to injected tracer laden water that exhibits some tracer mass loss (e.g. irreversible absorption, filtering, degradation). A second model that would support this linear relationship is that there is no loss of tracer mass in the injected water recovery and any change in the total production rate is due to proportional rate changes in the tracer and non-tracer laden produced water. Examining the electrical conductivity vs flow rate for the PI production zone suggests the first hypothesis is correct.

A simplified fracture conceptual model has to be constructed illustrating potential flow pathways based on geophysical, temperature changes, borehole and core studies. These flow pathways can be used to better interpret the flow rate and tracer break through curves at the various production zones.