PROCEEDINGS, Thirty-Eighth Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, February 11-13, 2013 SGP-TR-198

A LABORATORY INVESTIGATION OF REVERSIBLE PERMEABILITY DECLINE EFFECTS DURING FLOW THROUGH STIMULATED FRACTURE NETWORKS: IMPLICATIONS FOR

IMPROVED EGS RESERVOIR OPERATION METHODOLOGIES

Luke Frash, Marte Gutierrez, and Jesse Hampton

Colorado School of Mines

1610 Illinois St. Golden, CO, 80401, USA e-mail: lfrash@mines.edu

ABSTRACT

It is well known that permeability decline with time is a harsh and inevitable reality when producing fluids from or injecting sub-fracturing pressure fluids into fractured geological networks. Currently, it is often considered that these decline effects are mostly irreversible processes such as particle crushing, fines migration, particle embedment, thermal effects, and chemical precipitation. However, recent laboratory testing performed at the Colorado School of Mines has indicated that a portion of these effects may be reversible with potentially significant implications for engineered improvement of long-term production potential from fractured rock reservoirs. This paper details and compares the results obtained from recent laboratory pressure-flow testing performed on a complex multi-fractured unconfined granite reservoir, a simple bi-wing fractured simulated Enhanced Geothermal Systems (EGS) granite reservoir, and homogeneous unconfined single-wing fractured acrylic glass samples with and without proppant. For this analysis, repeatable pressure-dependent flow response was observed through an extensive series of constant pressure, constant flow, stepped constant pressure, and stepped constant flow testing. For the resulting data, it was discovered that a significant portion of the permeability decline with time was reversible and a qualitative theoretical explanation for this observed behavior is proposed. Additionally, with the benefits of controllable boundary conditions and known stable fracture geometry available with laboratory testing, it has been found that some common interpretations for field pressure data could be inaccurate, especially with respect to the interpretation of the fracture extension condition and estimation of fracture reopening pressure. Ultimately, several potential reservoir improvement techniques are suggested with the hope that they may be used to improve long term production levels for geothermal, petroleum, or even water wells.

INTRODUCTION

The flow of fluids through fractures in geologic materials is a complex process and an active research area. These complexities arise from the dynamic nature of geologic formations which may be described as heterogeneous fractured porous media existing in a transient thermodynamic, mechanical, and chemical state. The transient state is known to exist through naturally occurring tectonic processes as well as the relatively rapid artificial processes induced by well production and injection operations. In general, it may be assumed that the behavior of flow through fractured porous media includes effects from:

Combined fracture flow and Darcy flow effects

Thermal and chemical instability between the flowing fluid, residual pore fluid, and soluble geologic minerals.

Particle flow and transport phenomena

Thermo-hydro-mechanical response of the rock to natural and artificial flow

Channeling and short-circuiting flow effects through fractures and joints (Ishibashi, 2012)

Near wellbore effects Typically, it is expected that the permeability of a reservoir decreases with time once a completed well is put into production due to the interaction of a combination of these fracture flow effects. This behavior has been consistently observed with petroleum well production decline curves (Fetkovich et al., 1987) and during the lab characterization of proppant pack conductivity (API, 1989) as applied to artificially stimulated fracture networks. However, the application of stimulation treatments has been observed to increase the production of wells even when previous stimulation treatments may have already been performed (Ennis, 1989).

In an attempt to better understand some of these flow processes and why multiple stimulation treatments are so effective, laboratory experimentation has recently been performed using hydraulic fracture stimulated samples in a controlled laboratory setting. For these experiments, specimens of heterogeneous natural rock and relatively homogenous acrylic glass were subjected to controlled thermo-mechanical boundary conditions then subsequently drilled and injected with fluids to create stimulated fracture networks and characterize the flow of fluids through these networks. The focus of this testing was to utilize the benefits of known and stable boundary conditions to investigate the long term behavior and pressure-flow response of fractured geologic reservoirs. With the intentional extension of hydraulically stimulated fractures to the edges of the sample, it was possible to achieve effectively stable fracture geometries such that the dynamic effects of fracture growth with high pressure fluid injection could be minimized. This stable geometry allowed for the injection of fluids at pressures and rates greater than the fracture reopening limit such that a large range of the fracture flow behavior could be investigated and repeatable treatments could be performed within a given sample. Performing a comparable repeatable injection test in the field would be comparatively difficult as the critical injection rates and pressures for fracture reopening could not be exceeded without risking unwanted fracture propagation which would likely change fracture-flow response between consecutive tests. For this particular study, a unique experimental system was used to perform four tests. The results from each of these tests are summarized and presented to validate the repeatability of the flow testing experiments, the stability of the stimulated fracture networks, and to investigate the fracture flow phenomenon itself. A preliminary analysis of this data suggests that some of the expected permeability decline effects with production flow are actually reversible and suggestions on how this phenomenon may be utilized to achieve improved long term well production potential are presented.

EXPERIMENTAL PROCEDURE AND SETUP

For each test performed, an intact material specimen was subjected to a desired thermo-mechanical state and then drilled to install a sealed high-pressure injection borehole having an uncased injection interval. A diagram of the sealed borehole geometry is detailed in Figure 1. The specimen was then stimulated with hydraulic fracturing using constant flow rate injection to achieve critical fracturing pressures. Once the fracture was extended to the edge

of the test specimen, a series of computer controlled injection tests were performed to characterize the pressure-flow response of the reservoir. These injection tests included constant flow, constant pressure, stepped constant pressure, and stepped constant flow methods. Both pressure and flow controlled tests were performed to investigate any changes in reservoir response between these two injection control methods.

Figure 1: Diagram of the borehole sealing system

used for the injection experiments. The threaded casing was found to be critical for achieving a reliable hydraulic seal.

Injection was performed using a dual Teledyne Isco 65DM syringe pump system and an Autoclave Engineers remote control valve system. This system, as diagrammed in Figure 2, was operated using a custom LabVIEW program which allowed for programmed pulseless continuous injection of fluid with either directly controlled flow rate or proportional-integral-derivative (PID) controlled pressure. This system was capable of providing injection of fluids at pressures up to 70 MPa with corresponding flow rates in the range of 10 nL/min to 60 mL/min. Using the LabVIEW control program and an additional National Instruments Compact-DAQ data acquisition system it was possible to continuously monitor the following system parameters:

Injection pressure

Injection flow rate

Injection fluid temperature

Test specimen temperature

Bottom hole temperature Also, a Physical Acoustics Corporation (PAC) acoustic emission (AE) monitoring system was installed with piezoelectric sensors directly contacting the faces of the test specimen. For the granite tests, six sensors were used such that source

location and moment tensor analysis methods could be applied to the collected data. Typically, this system was setup to passively capture compression-wave arrival times and amplitudes as produced by micro-seismic events during injection.

Figure 2: Diagram of the hydraulic injection system

showing (a) the inflow fluid reservoir, (b) the automated valve system, (c) the Teledyne Isco syringe pumps, (d) the mixed slurry accumulator, (e) the injection wellhead, (f) the test specimen, (g) the optional production borehole, and (h) the optional outflow reservoir.

A heated true-triaxial apparatus was used when simulating elevated temperature and mechanically stressed conditions. This recently developed equipment was capable of applying independently controlled minimum, intermediate, and maximum confining stresses up to 13 MPa as well as simulated reservoir temperatures up to 180 ºC. This equipment allowed for percussive oriented borehole drilling into a simulated 30x30x30 cm

3 reservoir while the target

environmental conditions were applied. Additional details on this equipment are provided in the literature (Frash and Gutierrez, 2012, Frash, 2012, and Frash et al., 2012). The fluids used for testing included tap water and Valvoline

® Durablend

® SAE 80W90 Gear Oil. In all

four tests, the oil was used for the first hydraulic fracture stimulation treatment and water was used for all subsequent pressure-flow injection testing. Oil was preferred for stimulation as it has been found that higher viscosity fluids produce more ideal planar fractures and are able to cut through heterogeneities and discontinuities easier than lower viscosity fluids such as water (Ishida et al, 2004). High viscosity fluid for laboratory scale stimulation has also been found to be preferred when a comparison to field

scale data is desired (Johnson et al, 1991). After stimulation, water injection was used as the lower viscosity was expected to be more favorable for minimizing additional and unwanted hydraulic fracture propagation with high pressure injection and reopening treatments. It is also important to note that tap water was used in preference over distilled water as it was expected to be less chemically active with the natural minerals in the granite specimens so chemical dissolution and precipitation effects could be minimized. For tests using proppant to enhance the permeability of stimulated fractures, 170-325 mesh Ballotini

® was

added to the injection fluid. Typically, the fracturing treatment duration was short enough that the proppant was placed to fill the entire uncased hydraulic fracturing interval prior pressurization for stimulation. This method ensured that proppant was in contact with the opening fracture during stimulation such that it could enter the fracture before the treatment was completed. The hydraulic system was built to allow for additional proppant to be added with the injection stream using a rotationally mixed horizontal accumulator. Ballotini

® was selected as the

proppant due the superior sphericity of the material which was expected to allow for improved flow and placement characteristics.

TEST DETAILS AND RESULTS

Four different tests were performed using two granite specimens and two acrylic glass specimens. Both granite specimens were Colorado Rose Red Granite obtained with water jet cutting and diamond wire sawing methods in order to attain high quality intact blocks. In general, the tests included an unconfined room temperature granite sample, a true-triaxially confined and heated granite sample simulating an enhanced geothermal system (EGS), and two unconfined room temperature acrylic glass samples, one without proppant and one with proppant. Proppant was not injected into either of the granite samples.

Unconfined Granite Reservoir Test

The unconfined granite test specimen, shown in Figure 3, was cut to 30x30x24 cm with one face left rough. A vertical borehole was drilled at the center of the specimen’s top face where the upper cased interval had a depth of 101 mm and the uncased fracturing interval had a length of 51 mm, giving a total borehole depth of 152 mm. No significant pre-existing fractures were visually observed on the specimen prior to stimulation so the material was considered to be intact. The hydraulic stimulation of the sample was performed with a constant injection flow rate of 0.05 mL/min and total test duration of 14 hours. As shown

in Figure 4, the first breakdown event was observed at 33.2 MPa and a peak injection pressure of 49.4 MPa followed shortly after. Injection was continued long after breakdown as no steady state pressure condition was achieved, a complex fracture network was desired, and fracture extension to the edge of the sample was needed to attain a stable geometry during water flow testing.

Figure 3: Image of the unconfined granite test

specimen after stimulation.

Figure 4: Unconfined granite test hydraulic fracture

stimulation pressure plot. Both of these breakdown pressures were significantly greater than a predicted breakdown pressure of 7.5 MPa as calculated using Equation 1 (ASTM D4645, 2008) and an estimated rock tensile strength of 7.5 ± 1.8 MPa, measured following ASTM D3967-08.

THhb uP 3 (1)

Following stimulation, the hydraulic system was purged and tap water was injected to characterize the pressure-flow response of the reservoir. The total number of injection tests performed, according to each different control methodology, included eleven constant flow, two constant pressure, seven stepped

constant flow, and five stepped constant pressure tests. An example stepped constant pressure injection data plot is provided in Figure 5 for reference. For this particular example, each pressure controlled step was 15 min duration and 500 or 1000 kPa in magnitude with smaller pressure increments used at the beginning of the test.

Figure 5: Example raw and refined stepped constant

pressure data plot for the unconfined granite injection testing.

A simple analysis was performed to summarize the raw data and allow for comparison of the results between flow tests. This analysis involved isolating the last 30% of the data for each step, performing a linear regression on the pumped fluid volume to obtain a best estimate of flow rate, and taking an average of the measured pump pressure. Only the final 30% of the data for each step was used for this analysis as it provided a consistent and reliable representation of the system behavior at a pseudo-steady flow condition. This approach also mitigated the effects of the PID induced flow spikes that occurred with the start of each step as the new control pressure was being established. When applied to the raw data presented, the plot shown in Figure 6 was produced where excellent hysteresis was observed.

Figure 6: Analyzed stepped constant pressure data

using flow rate regression and average pressures from the final 30% of each 15 minute step.

The nearly overlapping increasing and decreasing pressure step data provides an early indication of the repeatability of pressure-flow response. This observation is further reinforced when observing the compiled data plot from all flow testing as shown in Figure 7.

Figure 7: Compiled unconfined granite test pressure-

flow data separated into testing completed before and after a pressure-pulse induced change in the fracture network.

In this compiled data plot, the analyzed flow test data was separated into groups with respect to their collection before or after a high pressure pulse injection was performed on the hydraulic fracture stimulated specimen. For the pulse injection, water at 75 MPa was dynamically injected into the sample by opening a valve that separated the injection borehole at reservoir pressure and the upstream hydraulic pump system held at high pressure. This pulse was applied with the intent of changing the fracture geometry and flow network without the need to reintroduce the SAE 80W90 oil or perform high flow rate injection to overcome the fluid loss due to having the fractures intersect the external surfaces of the specimen. In this case, the hydraulic conductivity of the fracture network was significantly increased, giving relatively higher flow rates for a given injection pressure, which indicated that the physical fracture flow network had been successfully changed. During individual injection stages, a plot similar to that shown in Figure 8 was obtained. In this plot, the permeability was be observed to be decreasing with time indicating that the reservoir was not at steady state during any individual testing stage. Such behavior was expected as this phenomenon has been consistently observed during API RP61 proppant pack conductivity testing. This issue is so common that the standard even warns against using measured proppant pack permeability values for field treatment design. Comparing the pressure-flow data between injection tests, it was found that the observed permeability decline with time was very similar between tests. Focusing on data collected from constant injection rate tests performed at 1.0 mL/min and duration of 60 min, the similarity of the pressure-flow response between consecutive tests may be clearly seen. In this case, the data shows that the tests results are very repeatable while also suggesting that the expected permeability decline with continued injection is not permanent. Due to the small difference in results

between consecutive tests, error analysis was applied to better represent the confidence in the obtained numbers.

Figure 8: Typical non-steady state behavior showing

effective permeability decline with time as observed during an injection test stage.

Figure 9: Plot comparison of pre-pulse 1 mL/min

data showing repeatability between tests. To conclude the unconfined granite test, cross-section analysis was performed on the specimen to allow for physical measurement of the stimulated fracture network. Photogrammetric methods were applied to obtain a digital representation of this network with the corresponding results as shown in Figure 10. Here the complex multi-wing fracture may be visualized and all of the fracture wings were found to intersect the uncased injection borehole interval. Flow through the all of the surface-intersecting fractures was also observed during injection testing.

Figure 10: Digitized stimulated fracture geometry for

the unconfined granite test.

Granite EGS Reservoir Test

A second granite specimen was tested to investigate fracture flow phenomenon at elevated pressures and temperatures. For this test, a 30x30x30 cm

3 specimen

was subjected to elevated temperatures of 50 ºC and confining stresses of 4.3, 8.6, and 12.8 MPa for the minimum horizontal, maximum horizontal, and vertical principal stresses respectively. A centered vertical injection borehole was drilled while reservoir conditions were maintained using percussive rotary methods. The upper cased interval had an installed depth of 107 mm and the uncased interval had a length of 74 mm, giving a total borehole depth of 181 mm. A pre-test inspection of the specimen found no significant existing fractures but a prominent quartz rich band was identified as may be seen in Figure 11.

Figure 11: Pre-test images of the EGS granite

specimen showing a lighter-colored quartz-rich discontinuity band passing nearly horizontally through the sample.

Hydraulic fracture stimulation was performed using three separate treatments of SAE 80W90 oil injected at 0.05 mL/min. The first treatment was about 20 min in duration and the stimulation treatment was stopped prior to the fracture intercept with the edges of the specimen. This technique was applied with the intent of generating a fully contained fracture for EGS reservoir simulation purposes. After this stage, a second borehole was drilled to intercept the stimulated fracture and create a producible binary-borehole EGS system. Two additional reopening and extension injection treatments were performed when preliminary flow testing revealed that communication between the injection and production boreholes was less than desired. The breakdown pressures for the stimulation treatments were measured at 18.1 MPa, 15.4 MPa, and 17.4 MPa chronologically. The predicted breakdown pressure was 11.8 MPa. During the stimulation treatments, AE data was used as the primary tool for estimating the stimulated fracture geometry. As shown in Figure 12, the

stimulated fracture was observed to be perpendicular to the minimum horizontal stress and contained beneath the quartz-rich band which had been identified prior to testing. This data also showed that the stimulated fracture had intercepted the specimen surface once all three treatments were completed.

Figure 12: Plot of AE source location data for the

third stimulation treatment showing the activity occurring on the bottom half of the sample, the planar orientation of the fracture, both drilled boreholes, and the extension of the fracture to the specimen surface.

After stimulation, the hydraulic system was purged and tap water was injected to characterize the pressure-flow response of the EGS reservoir. The total number of injection tests performed included 6 constant flow, 3 constant pressure, 1 stepped constant flow, 16 stepped constant pressure tests, and 1 high pressure pulse injection using a 65 MPa upstream pressure. Compiling the results from this testing, the pressure-flow response plot shown in Figure 13 was produced. Here the data has again been divided into pre-pulse and post-pulse sets.

Figure 13: Compiled EGS granite test pressure-flow

data separated into testing completed before and after a pressure-pulse induced change in the fracture network.

σh min σHmax

σh min

Top View

Front

View

Side View

As evident in the plot, the data collected during stepped constant pressure testing provided the most consistent and repeatable behavior. This is best noted when comparing the circular points and triangular points which represent pressure controlled and flow rate controlled injection tests respectively. Here, the effective fracture conductivity was typically found to be higher but less consistent between tests as there is significantly more scatter in the flow controlled tests than the pressure controlled tests. Focusing on the pre-pulse fracture network data, shown in yellow, the curves from seven different stepped constant pressure tests was so similar that the curves are overlapping on the plot. This repeatable behavior was less prominent during the post-pulse fracture flow data, shown in green, where a permanent decrease in fracture conductivity with total flow time was observed between tests. Separating the pre- and post-fracture data into two plots further emphasizes the repeatable similarity between consecutive tests, as shown in Figures 14 and 15. In the plot of the pre-pulse data, the shaded upper region highlights data collected using stepped constant pressure tests with 30 minute 1000 kPa intervals and injection pressures stepping up from 2000 kPa, peaking at 6000 kPa, and returning back to 2000 kPa. The non-shaded region contains data collected using a higher step count procedure with added steps at 500, 1000, 7000, 8000, 9000 and 10000 kPa. Thus, the distinct difference between the two tests appears to be mostly caused by the extra 500 and 1000 kPa steps where flow is occurring at pressures well below the theoretical 4300 kPa critical fracture reopening limit. As an additional note, the overlap of the rising and decreasing pressure curves signifies that the fracture-flow network was not significantly changed even though injection pressures greatly exceeding the reopening limit were used. This observation assists in validating the expectation that fracturing to the edge of the sample had allowed for a stable fracture geometry to be produced.

Figure 14: Plot of EGS granite pre-pulse pressure-

flow data with explanation of differences caused changing the stepped constant pressure injection procedure.

A slightly different perspective is gained when inspecting the post-pulse data where a notable decline in hydraulic conductivity between consecutive stepped constant pressure tests was observed even though the same procedure is used for all injection tests. As emphasized in the figure, the permeability declined with total flow time both within and between consecutive tests. The reason for this behavior is not currently understood but it is expected that the phenomenon is likely to be the result of a less stable fracture geometry existing after the pulse treatment was performed.

Figure 15: Plot of EGS granite post-pulse pressure-

flow data with explanation of total flow time effect not observed during pre-pulse pressure-flow testing.

To investigate the repeatability of the pressure-flow response, data from all pre-pulse 2000 kPa constant pressure injection stages was collected and plotted as shown in Figure 16. The shaded region on the left of the plot provides a direct comparison between 60 minute duration single stage data while the region on the right of the plot shows data taken from 30 minute steps executed during stepped constant pressure tests. Generally, the effective permeability was found to fluctuate between tests but no permanent decline effect resulted. Error bars are not shown on this plot as the estimated standard error for these values was less than 0.3%.

Figure 16: Plot of pre-pulse 2000 kPa constant

pressure injection step data showing repeatability between consecutive tests.

A preliminary cross-section analysis was performed after the completion of reservoir flow testing. For this analysis, a core was taken from the specimen concentric with the borehole axis and disked. The resulting fracture geometry, as shown in Figure 17, was found to be vertical, bi-wing, semi-planar, and perpendicular to the minimum principal stress.

Figure 17: Image of post-test EGS granite specimen

core and corresponding disked sections numbered increasing with elevation.

Acrylic Reservoir Tests

Two unconfined acrylic glass (PMMA) samples were also hydraulic fracture stimulated and subjected to reservoir flow testing. The more ideal homogenous linear-elastic properties of this material, as well as its water-clear transparency, provided a very useful reference for better understanding the behavior of flow through a stimulated fracture network. The first specimen was cut at 3.8 cm diameter and 14.0 cm length while the second was slightly larger at 7.6 cm diameter by 12.7 cm length. The injection boreholes were axially drilled into the specimens to a nominal 60 mm cased length and 25 mm uncased length. Percussive rotary drilling was used to induce near-borehole damage with small factures and to replicate the same methods used for granite testing. For the second acrylic specimen, the uncased interval was completely filled with loose Ballotini

® to allow for

immediate contact between the proppant and fracture opening during stimulation. The stimulation procedure for both specimens was captured with pressure-flow data recording and video at 30 fps. The resulting hydraulic breakdown plots are provided in Figure 18 and the corresponding videos are freely available on YouTube

® (Gutierrez

et al., 2012). The first acrylic specimen was observed to breakdown at a pressure of 9.6 MPa with a single wing fracture resulting. The second specimen fractured at a higher pressure of 11.7 MPa with one dominant wing reaching the boundary of the sample and a second smaller wing contained within the sample. Images of the two fracture geometries are provided in Figures 19 and 20. The included angle between the primary and secondary wings was estimated at 150º.

Figure 18: Hydraulic fracture breakdown plots for

both acrylic specimens.

Figure 19: Image of the first acrylic specimen

fracture geometry without proppant.

Figure 20: Image of the second acrylic specimen

fracture geometry with proppant. Ballotini

® proppant was visually confirmed to have

entered the fracture in the second acrylic specimen. Some of the particles were found to have exited the specimen through the tip of the primary wing while the remaining particles were dispersed throughout both wings. No residual particles were observed to be in close proximity to the fracture front or the

borehole itself. It is possible that a closer inspection of the proppant distribution could reveal additional insight on the active research topics of near-wellbore issues and proppant flow but additional testing would be required to draw any useful conclusions on this topic. For the first acrylic specimen, both oil and water were injected for pressure-flow analysis with the total number of tests including 4 constant flow and 4 stepped constant flow injections. The resulting data is provided on the plot in Figure 21 where a consistent behavior between tests is observed once again. During this testing, visual inspection of the fracture confirmed that no significant additional fracture propagation had occurred during the flow testing procedures.

Figure 21: Plot of first acrylic pressure-flow data. An additional plot of the 0.1 mL/min flow rate stage data has been provided in Figure 22 where the effects of long term flow injection were observed. As with the granite specimens, no permanent permeability decline with time was observed.

Figure 22: Plot of first acrylic specimen pressure-

flow repeatability with 0.1 mL/min oil injection data.

For the second acrylic specimen, a total of 3 constant flow, 10 stepped constant flow, and 2 stepped constant pressure injections were performed. These tests included both oil and water injection but focus is given to the water injection results as shown in Figure 23. The pressure measurements for this data were near the edge of the sensitivity for the pressure transducer used so high flow rates were implemented.

No change in the fracture extents was observed during this high flow rate injection but the proppant within the fracture was found to have shifted closer towards the fracture tip while also being vertically concentrated towards the center of the fracture.

Figure 23: Plot of second acrylic pressure-flow data. A plot of the 10.0 mL/min injection rate, as shown in Figure 24, suggests that the effective permeability did permanently decrease with time for the second acrylic sample. It is currently expected that this behavior was primarily induced by the migration of the proppant and an incomplete flushing of the residual oil from the fracture.

Figure 24: Plot of second acrylic specimen pressure-

flow repeatability with 10.0 mL/min water injection data.

Combined Analysis

A plot of the combined pressure-flow data from all four specimens is provided in Figure 25. Comparing the step pressure and step flow curves between the different tests, specimens, and sample geometries provides a further verification of the repeatability for the pressure-flow responses during fluid injection into fractured reservoirs. While not attempted within this paper, these curves may also reveal additional insight into the behavior of flow through fractured media when compared with previous experimental data, the physical measurements of the fracture dimensions, or modeling results. Effectively, these curves show the dramatic changes occurring with the pressure pulse treatment where change in the fracture dimensions was expected to have occurred.

Figure 25: Compiled pressure-flow data from all four

specimen tests.

ANALYSIS AND INTERPRETATION

The current analysis of the pressure-flow response data suggests that the flow of fluid through a fracture network is complex and involves the interaction of many different processes. Through a comparison of the data collected from the different specimen tests, the following consistent pressure-flow characteristics were observed:

Effective permeability decreased with time during a given injection stage with constant pressure or constant flow rate control.

Repeatable pressure-flow behavior was observed across consecutive injection tests with pressure controlled injection often providing the greatest level of similarity between tests.

The stimulated fracture extents were stable during flow tests once the surface of the specimen was breached.

Permanent permeability decline effects with total flow time were not observed.

Pseudo-steady state flow conditions were reached slower in the heterogenous granite material than in the more homogenous acrylic material.

Hysteresis in the pressure-flow data consistently indicated higher permeabilities during decrementing pressure-flow steps than with incrementing steps.

In an attempt to provide a qualitative explanation for the observed fractured reservoir flow phenomenon, two potential mechanisms for this behavior are now proposed. First, the decrease in permeability with time observed during continuous injection at a constant pressure is proposed to be caused by a particle clogging phenomenon. As diagrammed in Figure 26, it is reasonable to assume that the movement of particles within a fracture could create blockages and flow restrictions whenever a mobile particle encounters a constriction on the flow path smaller than its dimensions. In this case, the

aggregation of particles within constrictions would be expected to increase with time as long as near steady-state flow conditions are maintained. As the particles aggregate, the fluid flow paths would become increasingly constricted with time giving a continuing decrease in effective permeability. Following the assumption of clogging phenomenon, it is reasonable to infer that a change in the flow state could also cause the particles to dislodge and unclog. The current data suggests that these clogs could be removed or the original state regained by stoppage of the flow as well as forced changes in the flow rate or pressure.

Figure 26: Proposed particle clogging phenomenon

with continuous constant pressure head injection into a fracture.

Second, the hysteresis between decrementing and incrementing pressure-flow injection could possibly be explained by dynamic pressure storage within the fracture-pore volume of the specimen. Recognizing the surface of a fracture as a low-permeability boundary, any fluid pressurization of the fracture network would be expected to induce a mechanical strain response within the specimen material matrix. Typically, this response is expected to induce opening of the fracture aperture as with linear-elastic fracture mechanics and general hydraulic fracture theory (Valko and Economides, 1995). These induced strains could also be described as energy storage as some measure of elastic rebound could be expected to occur with a release of the fluid pressure. However, the dynamic friction losses associated with fluid flow through the fracture would be expected to induce a time-lag in the decay of excess pressure stored in a fracture. Thus, if the injection pressure inside the borehole were decreased, as would occur with a decrement of flow rate or pressure, the stored fluid would temporarily flow at a rate higher-than the equilibrium state and hysteresis of the form observed during would be expected to result. From these observations, it is expected that the effective permeability of a fractured media is a function of the fracture dimensions, continuous active flow time, fracture pressure history, and distribution

of propping particles with all other variables assumed constant. Extending from these observations, these results suggest that it may be possible to achieve higher long-term effective permeability values within fractured fluid reservoirs through the application of non-steady injection or production pumping methods.

CONCLUSIONS AND RECOMMENDATIONS

The results from four recently completed hydraulic fracture stimulation and reservoir flow testing experiments has provided an opportunity to better understand the hydraulic fracturing process as well as the behavior of fluid flow through stimulated fracture networks. These recent tests, performed on granite and acrylic, consistently demonstrated that:

Fracture pressure-flow response is repeat-able between tests

Effective permeability decreases with time during constant flow rate or constant pressure injection

Pressure controlled injection provides more consistent and repeatable data than flow controlled injection

Some elements of permeability decline with flow are reversible simply by changing the injection flow rate or pressure.

Fracture fluid storage may have a role in the flow characteristics of an active stimulated fracture.

Injection of fluids at high flow rates without inducing stimulation is possible when high-leakoff regions are intersected by a stimulated fracture.

Extending from these recent observations, it is recommended that the following options be considered during field well completion and production operations:

Consider non-steady pumping methods as a potential means of increasing the long term effective permeability and fluid recovery rates from a stimulated reservoir.

Implement a stepped constant pressure injection test during field well testing operations to characterize the full pressure-flow response of the reservoir.

REFERENCES

API (1989). “Recommended Practices for Evaluating Short Term Proppant Pack Conductivity: API Recommended Practice 61 (RP 61).” American Petroleum Institute, First Edition, Oct. 1.

Ennis, B.K. (1989). “Case History of Restimulations in Western Oklahoma.” SPE Production Operations Symposium, Oklahoma City, Oklahoma, 13-14 Mar.

Fetkovich, M.J., Veinot, M.E., Bradley, M.D., Keisow, U.G. (1987). “Decline Curve Analysis Using Type Curves: Case Histories.” SPE Formation Evaluation, Vol. 2, No. 4, pp. 637-656.

Frash, L. (2012). Laboratory Simulation of an Enhanced Geothermal Reservoir. M.S. thesis submitted to Colorado School of Mines, Department of Civil and Environmental Engineering.

Frash, L. and Gutierrez, M. (2012). “Development of a New Temperature Controlled True-Triaxial Apparatus for Simulating Enhanced Geothermal Systems (EGS) at the Laboratory Scale.” PROCEEDINGS: Thirty-Seventh Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford, California, Jan. 30 – Feb. 1.

Frash, L., Gutierrez, M., and Hampton, J. (2012). “On the Experimental Simulation of EGS Reservoirs Using a Heated True-Triaxial Apparatus.” PROCEEDINGS: New Zealand Geothermal Workshop, Auckland, New Zealand, Nov. 19-21.

Gutierrez, M., Frash, L., and Hampton, J. (2012). “Hydraulic Fracturing in Acrylic with Proppant,” available through YouTube at http://youtu.be/ rbE4nisWlyA, 2012.

Gutierrez, M., Frash, L., and Hampton, J. (2012). “Water Clear Acrylic Laboratory Hydraulic Fracturing Test,” available through YouTube at http://youtu.be/PEXOE2FTDlI, 2012.

Ishibashi, T., Watanabe, N., Hirano, N., Okamoto, A., Tsuchiya, N. (2012). “Experimental and Numerical Evaluation of Channeling Flow in Fractured Type of Geothermal Reservoir.” PROCEEDINGS: Thirty-Seventh Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford, California, Jan. 30 – Feb. 1.

Ishida, T., Chen, Q., Mizuta, Y., and Roegiers, J. (2004). “Influence of Fluid Viscosity on the Hydraulic Fracturing Mechanism”. Transactions of the ASME, Volume 126, Sep.

Johnson, E. and Cleary, M.P. (1991). “Implications of Recent Laboratory Experimental Results for Hydraulic Fractures.” SPE Low-Permeability Reservoirs Symposium, Denver, Colorado, 15 - 17 Apr.

Valko, P. and Economides, M.J. (1995). Hydraulic Fracture Mechanics. John Wiley & Sons: New York.

ACKNOWLEDGEMENTS

Financial support provided by the U.S. Department of Energy under DOE Grant No. DE-FE0002760 is gratefully acknowledged. The opinions expressed in this paper are those of the authors and not the DOE.