Numerical Investigation on Mesoscale Evolution of Hydraulic Fractures in Hydrate-Bearing Sediments

Abstract

Hydraulic fracturing is widely recognized as a potential stimulation technology for the development of challenging natural gas hydrate. However, the fracturing behavior of non-diagenetic hydrate reservoirs has peculiar characteristics that are different from those of conventional oil and gas reservoirs. Herein, a fully coupled fluid-mechanical model for simulating hydraulic fracturing in hydrate-bearing sediments (HBS) was established based on the discrete element method, and the influence of hydrate saturation, in situ stress, and injection rate on the meso-fracture evolution was investigated. The results indicate that with the increase in hydrate saturation, the fracture morphology transitions from bi-wing to multi-branch, thereby enhancing fracture complexity. Both tensile and shear failure modes exist, and the tensile failure between the weakly cemented sediment particles is dominant. The tensile strength of HBS is an exponential function of hydrate saturation, with the breakdown pressure being governed by hydrate saturation and in situ stress, with the form being consistent with the classical Kirsch equation. Additionally, lower in situ stress and higher injection rates are conducive to the generation of microcracks, whereas an excessive injection rate reduces the fracture length. These findings contribute to understanding the meso-evolution mechanism of hydraulic fractures and guide the design of on-site hydraulic fracturing plans of natural gas hydrate reservoirs.

1. Introduction

Natural gas hydrate (NGH) is a crystalline compound consisting of natural gas molecules (mainly methane) and water molecules [1]. Global NGH deposits are estimated to contain 1.8–2.1 × 10¹⁶ stm³ of natural gas, at least twice the carbon content of all proven conventional fossil fuels [2]. Given this, NGH is considered to be an important part of the global carbon cycle and ideal alternative energy in the 21st century [3,4].

NGH accumulation in marine sediments is controlled by many factors, such as geological, formation temperature, and capillary pressure [5,6,7]. In most cases, the fracture system provides a channel for gas to migrate to the occurrence zone and is the key factor affecting NGH accumulation [8]. However, when the steady state of hydrates is disrupted, the pore overpressure generated by their dissociation can lead to uncontrolled hydraulic fracturing, which may harm marine ecosystems [9]. For example, gas hydrate destabilization may generate soft-sediment deformation and seafloor collapses or trigger large-scale continental slope instability [10]. Moreover, the released methane leaks into the ocean through the fracture systems, affecting the atmospheric methane budget and exacerbating the greenhouse effect [11,12,13,14,15], and also seep-carbonate masses associated with sedimentary instability in the host sediments are reported as evidence for the occurrence of paleo-gas hydrates [16]. Therefore, studying the fracturing behavior of hydrate-bearing sediments (HBS) is quite important to understand the NGH occurrence mechanism and its roles in the global carbon cycle.

From the perspective of resources, approximately 97% of hydrates occur in clayey–silty sediments, making gas recovery challenging because the pressure transmission, fluid injection, thermal diffusion, hydrate dissociation, and fluid seepage are inhibited by the low reservoir permeability [17,18,19]. Given this, hydraulic fracturing (HF) has been proposed for enhancing the productivity of NGH reservoirs [20,21,22,23,24,25]. Moreover, the stimulation effects of HF on hydrate reservoirs have been confirmed through recent simulations [20,21,22,23]. Our latest studies discovered that HF combined with multi-well inject–production systems increase the gas production by 1–2 orders of magnitude [25]. Consequently, studying the fracturing behaviors of HBS is of great importance for the commercial development of hydrates.

At present, a handful of researchers have studied the fracturing behavior of HBS on a lab scale. Konno et al. (2016) conducted fracturing tests in synthesized HBS using a triaxial pressure autoclave and observed fractures perpendicular to the minimum principal stress plane [26]. Too et al. (2018) carried out HF in a penny-shaped crack on synthesized hydrate-bearing sand and discovered vertical fractures [27]. Yamada et al. (2018) used porous media composed of glass beads, NaHCO₃, C₂H₅OH, and ice to simulate HBS and then implemented HF by injecting C₂H₅OH aqueous solution [28]. Their results indicate that the high permeable seepage channels formed are conducive to the fluid flow. Liu et al. (2020) evaluated the fracability of HBS based on the analytic hierarchy process-entropy method [29]. Their findings suggest that HBS with high hydrate saturation and great stress anisotropy have fracability. Ma et al. (2022) proved the fracability of clayey silt hydrates even with a low hydrate saturation of 25% [30]. Nie et al. (2022) analyzed the fracability of clayey–silty HBS based on the multi-layer perceptron and analytic hierarchy process [31]. The above studies confirm that various fractures can be formed in HBS. However, hydrate distribution in synthesized HBS is difficult to control, and the experimental conditions are too harsh for conducting HF tests. Additionally, capturing the fracture evolution is difficult owing to the inside of the sample being undetectable. Accordingly, the use of numerical simulation to study the fracturing behavior of HBS is another effective means.

There are four main numerical simulation methods for hydraulic fracturing.

• Finite-element method (FEM): Numerous researchers have used the FEM to simulate HF in various oil and gas reservoirs, among which the cohesive zone method (CZM) based on the viscoelastic–plastic damage constitutive criterion is widely used [32,33,34,35,36]. However, the mesh needs to be constantly reconstructed to simulate the fracture growth, which greatly increases the number of calculations. Additionally, fracture propagation is disturbed by grid boundaries, thus affecting the reliability of the results.
• Extended finite-element method (XFEM): XFEM is a variation of FEM in which discontinuous enrichment functions are introduced to describe the discontinuity of fractures, thereby solving the calculation burden caused by mesh reconstruction [37]. Although the method is effective, the XFEM introduces additional difficulties such as the need for special integration techniques to resolve the stiffness matrix, blending of enriched and non-enriched elements, and ill-conditioned stiffness matrices [38]. In addition, the most suitable enrichment functions for a particular type of stress concentration are usually not known in advance. This poses new challenges when applying the XFEM to arbitrary types of singularities [39,40].
• Boundary element method (BEM): Unlike the FEM, which divides elements in the continuum domain, the BEM only meshes elements on the domain boundary [41]. The displacement discontinuity method (DDM), an indirect DEM, is widely used to simulate HF [42]. However, the BEM is not good at solving nonlinear problems and cannot accurately describe the stress and displacement fields at the fracture tip [43]. For example, it is difficult to simulate fracture propagation in heterogeneous materials and handle the coupling of fluid–solid between the rock matrix and fractures [32].
• Discrete element method (DEM): The DEM was proposed by Cundall and Strack (1979) for simulating the interaction between granular materials [44]. Unlike methods based on continuum mechanics, the DEM idealizes the material as separate particles bonded together at their contacts, and the microcracks generated by the breakage of the bonds. Since there are no precise meshing and numerous continuum assumptions, the DEM has a unique model heterogeneity and thus can simulate the HF process of discrete materials. However, limited by the number of particles and computational resources, the DEM is primarily used for small-scale simulations and is good at revealing the meso-mechanical behavior of discrete materials, whereas methods such as the unconventional fracture model [45], the simplified 3D DDM [46], and the FEM-based method [47] are capable of mine-scale simulations.

Several HF simulations have been conducted based on the DEM. Shimizu et al. (2011) studied the fracturing behavior of hard rock and obtained AE data similar to those obtained via experiments [48]. Wang et al. (2014) found that the simulated fracture length and breakdown pressure of a coal seam were consistent with the field results [49]. Zhou et al. (2017) confirmed that the DEM was reliable for understanding the interaction between fractures in a laminated rock [50]. The above studies proved the applicability of the DEM in various fracturing conditions, considering that HBS is a discontinuous complex medium composed of loose sediment particles and cemented hydrate particles and thus specifically suitable for simulation with the DEM. Additionally, the DEM can reconstruct samples at the material level, allowing good control of hydrate distribution and clear capture of the evolution of fractures.

This study intended to investigate the meso-mechanical behavior of HBS during hydraulic fracturing, which, to the best of our knowledge, is still not well understood. To this end, a hydraulic fracturing model of HBS was established based on a fully coupled fluid-mechanical algorithm in the DEM, and the meso-parameters of the DEM model were calibrated by the drained shearing experiments of HBS in the Nankai Trough. Subsequently, the effects of hydrate saturation, in situ stress, and injection rate on fracturing behaviors, such as the initiation pressure, evolution of microcracks, bond failure mode, and fracturing effect, were thoroughly analyzed. The findings of this study reveal the meso-evolution mechanism of fractures, contributing to the understanding of the peculiar fracturing behavior of this heterogeneous complex geologic material and the implementation of hydraulic fracturing in hydrate reservoirs.

2. Methodology

2.1. HBS Simulation Methodology

A DEM code, that is, the particle flow code (PFC), was employed. In the PFC, materials are modeled as an aggregate of small rigid circular particles that interact through normal, shear, and rotary springs at contacts; therefore, the constitutive behavior of various materials can be modeled by assigning linear-based and/or bonded-contact models [44]. In addition, with the load changes, the force and position of particles are updated by the force-displacement law using a time-stepping algorithm.

The details of the DEM algorithm can be found in Ref. [51]. Only the simulation methodology of HBS is presented here. HBS can be regarded as a composite medium composed of sediment particles and hydrate particles (Figure 1a). Thus, there are three contact types in HBS, namely sediment–sediment (s_s), sediment–hydrate (s_h), and hydrate–hydrate (h_h). In s_s, the bond force is as weak as the capillary tension of meniscus water [52]; thus, the flat-joint model is adopted (Figure 1b) in which the circular particles are constructed into polygonal particles and a flat-joint contact simulates the behavior of an interface between two notional surfaces. In s_h and h_h, the bonds are stronger due to the cemented nature of hydrates [53]. Consequently, the parallel-bond model is used to characterize the mechanical behavior of hydrates (Figure 1c) [54,55]. Once the bonds fail, they degenerate into the rolling resistance model, in which a rolling resistance coefficient is used to characterize the shape effect of particles (Figure 1d).

2.2. Fluid–Mechanical Coupling Algorithm

The coupling between fluid and mechanical fields is described using Cundall’s fluid flow algorithm [56,57,58]. In Cundall’s fluid flow algorithm, a closed chain formed by particles in contact with each other is defined as a “reservoir” for storing pore fluid, and each contact is assumed to be a “pipe” for exchanging fluid from adjacent reservoirs, resulting in the formation of a flow network, as shown in Figure 2a.

The flow in pipes is assumed to be Poiseuille channel flow (Figure 2b), and the flow between reservoirs can be described by the Hagen–Poiseuille equation:

$$q=\frac{{\alpha }^3}{12\mu }\frac{\Delta p}{L_p},$$

(1)

where $q$ is the flow rate (m³/s), $\mu$ is the fluid viscosity (mPa·s), $\Delta p$ is the pressure difference (MPa), $L_p$ is the pipe length (m), and $\alpha$ is the pipe aperture (m).

$L_p$ is the harmonic mean of the two-particle radii $r_i$ and $r_j$ and can be described as

$$L_p=\frac{4r_i{r}_j}{r_i+r_j},$$

(2)

When $F^n$ is the compressive stress, $\alpha$ is calculated by

$$\alpha =\frac{{\alpha }_0{F}_0^n}{F^n+F_0^n},$$

(3)

where ${\alpha }_0$ is the initial aperture (m), and $F_0^n$ is the normal contact force (N).

When $F^n$ is the tensile stress or the bond at the contact fails, then $\alpha$ is described as

$$\alpha ={\alpha }_0+d-r_i-r_j,$$

(4)

where $d$ is the distance between particles (m).

The value of ${\alpha }_0$ determines the permeability of the medium [59] and is written as

$$k=\frac{{{\displaystyle \sum }}_{pipes}{L}_p{\alpha }_0^3}{12\pi {{\displaystyle \sum }}_{particles}{r}^2},$$

(5)

The permeability of HBS is governed by hydrate saturation $S_h$ and described by the Tokyo mode [60]:

$$k=k_0{\left(1-S_h\right)}^n,$$

(6)

where $k_0$ is the permeability of the hydrate-free sediment (m²), and $n$ is the exponential parameter and its value is 2.5 in this study [61].

Based on Equations (5) and (6), the value of ${\alpha }_0$ is calculated by

$${\alpha }_0=\sqrt[3]{\frac{12\pi k_0{\left(1-S_h\right)}^n{{\displaystyle \sum }}_{particles}{r}^2}{{{\displaystyle \sum }}_{pipes}{L}_p}},$$

(7)

During the time $\Delta t$, the variation in the fluid pressure $\Delta P$ is

$$\Delta P=\frac{K_f}{V}\left({{\displaystyle \sum }}_{i=1}^{N}q\Delta t-\Delta V\right),$$

(8)

where $K_f$ is the bulk modulus of the fluid (Pa), $V$ is the volume of the reservoir (m³), $\Delta V$ is the change in the reservoir volume (m³), and $N$ is the number of pipes.

Simultaneously, the fluid pressure $p_p$ acts on the surface of the surrounding particles (Figure 2c). The applied force $F_{\alpha }$ acting on the particle center is given by

$$F_{\alpha }={{\displaystyle \int }}_{-\beta }^{\gamma }{p}_{p}cos\theta rd\theta ,$$

(9)

When the bond fails, it is assumed that the flow instantaneously occurs, and the pore pressure in the new reservoir is their average value $p_p^{\prime }$ described as

$$p_p^{\prime }=\frac{p_p^i+p_p^j}{2},$$

(10)

where $p_p^i$ and $p_p^j$ are the pore pressure of reservoir “i” and “j”, respectively (MPa).

The fully coupled fluid–mechanical models are shown in Figure 3. The simulation of the mechanics of particles is presented on the left side, where the movement of particles and the contact force between particles are updated by the motion equation and force-displacement law; the simulation of fluid flow is presented on the right side, where Cundall’s fluid flow algorithm is employed; and in the middle is the fluid–mechanical coupling process, where the variation in contact force between particles causes the change in pipe aperture, and the variation in reservoir pressure updates its force on the surrounding particles. After the fluid calculation step, a kind of additional force is exerted on the particles calculated by fluid pressure in the reservoir. Consequently, the above two interactive processes realized the full coupling of the mechanical and fluid fields.

2.3. Microcrack Growth and Failure Criteria

When the contact force is greater than the bond strength, a microcrack is generated. The crack length is assumed to be the same as the length of the pipe $L_p$, and its direction is perpendicular to the line connecting the centers of the two particles [48,49,50]. Notably, this assumption is mainly used for crack visualization and not for representing the actual crack lengths.

The tensile stress $\sigma$ and shear stress $\tau$ acting on the bond can be calculated by

$$\sigma =\frac{T}{A}+\frac{⎡M⎤}{I}\overline{R},$$

(11)

$$\tau =\frac{\left|S\right|}{A},$$

(12)

where $T$ is the axial force (N), $S$ is the shear force (N), $M$ is the moment (N·m), $I$ is the moment of inertia (kg·m²), $A$ is the cross-sectional area (m²), and $\overline{R}$ is the bond radius (m).

For the flat-joint model, as the joint is regarded as a plane that cannot bear moments, the calculation of its tensile stress excluded the moment term in Equation (11).

Figure 4 shows the bond failure modes and the failure envelope. When the tensile stress $\sigma$ or shear stress $\tau$ exceeds the tensile strength ${\sigma }_c$ or shear strength ${\tau }_c$, a corresponding tensile or shear microcrack occurs. The shear strength ${\tau }_c$ is determined by the normal stress $\sigma$, cohesion $c$, and friction angle $\varphi$ and described as

$${\tau }_c=c-\sigma tan\varphi ,$$

(13)

3. Modeling

3.1. HF Model

Figure 5 shows the HF model of HBS. The radius of sediment particles is 0.1–0.2 mm, and the median size is 0.113 mm, which is consistent with that of sediments in the Nankai Trough [61]. Referring to the method used in previous studies, smaller particles (0.04 mm radius) are generated to represent hydrates [15,54,55]. In this manner, HBS with a certain hydrate saturation can be generated by controlling the number of hydrate particles. The densities of sediments and hydrates are the same as those in the Nankai Trough, at 2650 and 320 kg/m³, respectively [61]. Limited by computing resources, the model size is set to 10 mm × 10 mm. A circular hole with a radius of 0.3 mm is reserved in the center of the model for fluid injection. Additionally, the surrounding smooth servo walls are used for simulating the in situ stress through the servo mechanism [51].

3.2. Calibration of Model Parameters

Proper calibration of meso-parameters is crucial for DEM simulation. Here, drained shearing tests are mostly used to calibrate meso-parameters, and the calibration principle is that the mechanical response of HBS generated by the DEM is qualitatively consistent with the indoor test results. Based on the drained shearing test results of HBS in the Nankai Trough [62], the “trial–error” method was adopted to calibrate the meso-parameters. Through repeated attempts, the values of the meso-parameters were obtained (

Table 1. Values of meso-parameters.

Contact Model	Flat-Joint Model	Parallel-Bond Model	Rolling Resistance Model
Effective modulus Ec (N)	3.8 × 107	4.2 × 108	3.8 × 107
Normal-to-shear stiffness ratio kr	1.5	1.5	1.5
Friction coefficient μ	0.25	0.04	1.0
Rolling resistance coefficient μr	_	_	1.0
Bond effective modulus Ecb (N)	_	4.2 × 108	_
Bond normal-to-shear stiffness ratio krb	_	1.5	_
Gap interval gr (m)	1.0 × 10−5	1.0 × 10−5	1.0 × 10−5
Tensile strength of bond σcb (N)	2.1 × 106	2.4 × 107	_
Cohesion cb (N)	2.1 × 106	2.4 × 107	_
Friction angle ϕb (°)	10	10	_

). The drained shear model of HBS is shown in Figure 6. It can be seen that the simulated deviatoric stress–axial strain is in good agreement with the indoor test results, proving the reliability of the values of the meso-parameters in Table 1.

Additionally, the Brazilian splitting test was simulated to obtain the tensile strength ${\sigma }_t$ of HBS, as shown in Figure 7. When the hydrate saturation is 0%, 20%, 40%, and 60%, the simulated tensile strength of HBS is 0.59, 0.80, 1.17, and 1.71 MPa respectively, indicating that the tensile strength increases as hydrate saturation increases.

Figure 8 depicts the ${\sigma }_t$ obtained in our simulation and previous indoor tests. It can be seen that the tensile strength obtained from previous indoor tests is highly variable. This is due to the strong non-homogeneity of HBS, especially since sediment components, grain size, and hydrate distribution pattern all have significant influence on the mechanic properties of HBS. Nevertheless, a general trend of increasing tensile strength with increasing hydrate saturation can be observed as the existence of hydrates improves the consolidation of HBS. The simulated values are within the range of the test results of 0.28−12.5 MPa [26,27,63,64,65,66,67], which further supports the reliability of the DEM model. Additionally, the simulated tensile strength also increases as hydrate saturation increases and is found to be an exponential function of hydrate saturation with a fitting coefficient R² of 0.9981. This indicates that the DEM model well simulated the bonding characteristics of hydrate particles.

4. Results and Discussion

4.1. Effects of Hydrate Saturation

The fracturing process was simulated for HBS with hydrate saturations of 0, 25%, 35%, 45%, 55%, and 65%, respectively. Notably, the injection rate was 5 × 10⁻⁷ m³/s, and the in situ stresses in the X and Y directions were 1 and 2.5 MPa, respectively.

4.1.1. Injection Pressure

In our simulation, the number of microcracks developed within 5 × 10⁻⁴ s was recorded as acoustic emission (AE) events. The injection pressure and AE events at different hydrate saturations are shown in Figure 9. As can be seen, injection pressure increases with the injection of fracturing fluid, and a sudden drop occurs at the AE event. Furthermore, the fluctuations in injection pressure are synchronous with the occurrence of AE events. The maximum injection pressure is regarded as the breakdown pressure $P_b$, which represents the overall strength of the HBS. It can be seen that the breakdown pressure increases as hydrate saturation increases, which is 4.25 ($S_h$ = 0%), 5.74 ($S_h$= 25%), 7.92 ($S_h$ = 35%), 8.82 ($S_h$ = 45%), 9.46 ($S_h$ = 55%), and 12.06 MPa ($S_h$ = 65%). The mechanism of this phenomenon is twofold: first, the presence of hydrates reduces the effective permeability of the HBS, which increases the resistance to fluid flow and reduces the rate of fluid leaching; second, the bonding behavior of hydrates improves the consolidation strength of the HBS, and greater fluid energy is required for fracture initiation and propagation.

The relationship between breakdown pressure and hydrate saturation is fitted in Figure 10. It can be seen that $P_b$ is an exponential function of $S_h$, and the fitting coefficient R² reached 0.9711, proving the reliability of the equation.

4.1.2. Contact-Force Chains and Applied Force Induced by Fluid

The contact force between the particles and the applied force induced by the fluid was recorded, as shown in Figure 11. As can be seen, contact-force chains were mainly distributed along the vertical direction to bear the maximum in situ stress. However, in the near frac zone, contact-force chains were distorted as the particles were compressed by the fluid pressure. With the increase in hydrate saturation, the particles on the frac zone were subjected to greater fluid pressure, resulting in more severe distortion of contact-force chains. In the frac zone, particles separated under the fluid pressure, owing to the lack of contact force there. Additionally, as hydrate saturation increases, the strength of the HBS gradually increases and the permeability gradually decreases, which leads to a gradual reduction in the region of fluid action. Consequently, the presence of hydrates affects the interaction behavior between particles and fluid, including a slower fluid penetration rate, and increasing the applied force on particles induced by fluid pressure.

4.1.3. Evolution of Microcracks

The microcrack distribution at different times was recorded, as shown in Figure 12. When hydrate saturation was lower than 25% (Figure 12a,b), a macro fracture along the direction of maximum in situ stress was formed. Different from the bi-wing fractures in Figure 12a,b, multi-branch fractures at hydrate saturations of 45%–65% (Figure 12c–f) were generated. This indicates that with an increasing hydrate saturation, the strong cemented zone of hydrates forced hydraulic fractures to divert to weakly cemented sediments, causing the formation of multi-branch fractures. However, the CZM-based simulations in Refs. [35,36] all obtained regular flat fractures because of the assumption that HBS were homogeneous and isotropic. In contrast, the heterogeneity of HBS caused by the uneven distribution of hydrates was well simulated in the DEM model, which makes our simulation results more consistent with the existing experimental results [27,28,29,30,31].

To reveal the fracture evolution mechanism, the microcracks with different failure modes generated at different contacts were recorded, as shown in Figure 13. As can be seen, both tensile and shear microcracks exist at s_s for the hydrate-free sediment (Figure 13a), with more tensile microcracks. When the hydrate saturation is 25% (Figure 13b), a few tensile microcracks occur at s_h. When the hydrate saturation increases to 35% (Figure 13c–f), there is further emergence of a few tensile microcracks at h_h, whereas shear microcracks are difficult to occur at s_h and h_h. Overall, tensile and shear microcracks at s_s play a dominant role, followed by the tensile microcracks at s_h.

4.2. Effects of the In Situ Stress

The fracturing process was simulated for HBS with X-direction stress of 1, 2.5, and 5 MPa, respectively. Notably, the hydrate saturation was 55%, the injection rate was 5 × 10⁻⁷ m³/s, and the in situ stress in the Y-direction was 2.5 MPa.

4.2.1. Injection Pressure

The injection pressure under different in situ stress conditions is shown in Figure 14. As can be seen, when the in situ stresses in the X-direction were 1, 2.5, and 5 MPa, the breakdown pressures were 8.82, 11.02, and 11.82 MPa, respectively, indicating that the breakdown pressure increases as the in situ stress increases. This is because great contact force needs to be overcome for the generation of microcracks under high in situ stress.

In addition, according to the Kirsch equation of stress distribution around the elastic circular hole [68,69,70], the breakdown pressure is described as

$$p_b={\sigma }_t-{\sigma }_1+3{\sigma }_2,$$

(14)

where ${\sigma }_1$ and ${\sigma }_2$ are the principal stresses in the two directions (MPa), and ${\sigma }_1\ge {\sigma }_2$.

Here, the relationship between $p_b$ and ${\sigma }_t$, ${\sigma }_1$, and ${\sigma }_2$ are fitted as

$$p_b=5.7781{\sigma }_{\mathrm{t}}+0.1813{\sigma }_1+0.7005{\sigma }_2,$$

(15)

The regression coefficient of Equation (15) is 0.9981, indicating its reliability. Furthermore, this also reveals that the initiation mode belongs to the tensile failure mode, which is consistent with the experimental results of Ref. [31].

By comparing Equations (14) and (15), we discovered that their forms were consistent, but the coefficients were different. As the Kirsch equation is derived on the assumption that the formation is elastic, isotropic, and homogeneous, a certain degree of correction is required in the actual application. Additionally, Equation (14) is developed under a relatively ideal condition. Especially, the phase transition may occur for the hydrate in the change in environmental conditions, which may further influence the fluid flow and fracture propagation [71]. Consequently, further applications should take full account of the peculiar physico-mechanical properties of HBS.

Combined with the fitted equation in Figure 10 and Equation (15), the breakdown pressure can be described as the function of hydrate saturation and in situ stress:

$$p_b=3.3328e^{1.7863S_h}+0.1813{\sigma }_1+0.7005{\sigma }_2,$$

(16)

4.2.2. Distribution and Number of Microcracks

The microcrack distribution as well as the number of microcracks under different in situ stresses are shown in Figure 15 and Figure 16, respectively. As can be seen, when the X-direction in situ stress increases from 1 to 5 MPa, the microcrack growth direction gradually shifts from X-axil to Y-axil. This is consistent with the classic fracturing theory and the HF experiments of HBS conducted in Refs. [27,29]; that is, the magnitude of the minimum in situ stress difference controls the fracture propagation direction.

Additionally, the total number of microcracks decreases from 263 to 107 as the X-direction in situ stress increases from 1 to 5 MPa, indicating that the large in situ stress prevents the generation of microcracks. Thus, NGH reservoirs with shallow depth are the “sweet spots”, which is in line with the fracability evaluation results in Ref. [31].

4.3. Effects of Injection Rate

To analyze the effect of the injection rate on fracturing behavior, the fracturing process was simulated for injection rates of 1 × 10⁻⁷, 5 × 10⁻⁷, and 1 × 10⁻⁶ m³/s, respectively. Notably, the hydrate saturation was 55%, and the in situ stresses in the X- and Y-directions were 1 and 2.5 MPa, respectively.

4.3.1. Injection Pressure

The injection pressure at various injection rates is shown in Figure 17. With the increase in the injection rate, the fluctuation frequency of injection pressure increased, indicating that more microcracks were generated. Additionally, the breakdown pressures are 8.76, 9.45, and 10.86 MPa for injection rates of 1 × 10⁻⁷, 5 × 10⁻⁷, and 1 × 10⁻⁶ m³/s, respectively, indicating that the breakdown pressure increases as the injection rate increases. The HF experiments of HBS in Ref. [30] also observed this phenomenon, and the mechanism for this is the rapid accumulation and hysteresis dissipation of bottom-hole fluid energy at a high injection rate.

4.3.2. Distribution and Number of Microcracks

The microcrack distribution as well as the number of microcracks under different injection rates are shown in Figure 18 and Figure 19, respectively. As can be seen, when the injection rate is enhanced from 1 × 10⁻⁷ m³/s to 1 × 10⁻⁶ m³/s, the number of microcracks increases from 121 to 306, suggesting that microcrack generation is more favorable at large injection rates. However, when the injection rate is 1 × 10⁻⁶ m³/s, microcracks are mainly distributed in the near-well region, resulting in insufficient fracture penetration depth. Overall, when the displacement is 5 × 10⁻⁷ m³/s, the number of microcracks is sufficient and the fracture penetration depth is ideal. Therefore, appropriate injection rates should be selected when designing the fracturing plans for hydrate reservoirs.

5. Conclusions

In this study, a fully coupled fluid–mechanical model was established to simulate HF in HBS, and the effects of hydrate saturation, in situ stress, and injection rate on the fracturing behavior were investigated. The following conclusions can be drawn.

• Hydrate saturation has a significant effect on fracture morphology, and complex fractures tend to form in HBS with high hydrate saturation. When the hydrate saturation is lower than 25%, bi-wing fractures are formed. With the increase in hydrate saturation, the strength of HBS significantly increased due to the bond nature of hydrates, thus increasing the injection pressure and forming multiple fractures.
• The tensile and shear stresses induced by the injected fluid jointly promote the generation of hydraulic fractures, with tensile stress being dominant. Additionally, fractures tend to propagate in the weakly cemented sediments, followed by those at the interface between hydrates and sediments. Owing to the high bond strength between hydrate particles, it is difficult for fractures to pass through them.
• Without considering the fracturing conditions, the breakdown pressure of HBS is governed by hydrate saturation and in situ stress. Given that the tensile strength of HBS is an exponential function of hydrate saturation, the fitted equation of the breakdown pressure of HBS is consistent with the classical Kirsch equation.
• Small in situ stress and high injection rates are conducive to the generation of microcracks, whereas an excessive injection rate causes over-reconstruction in the vicinity of the wellbore and reduces fracture penetration depth. Therefore, the shallow hydrate reservoir with elevated hydrate saturation is the “sweet spot” for fracturing, and an appropriate injection rate should be chosen for this non-diagenetic geomaterial.

The findings of this study reveal the mesoscale evolution of hydraulic fractures in HBS, which can help to understand the particular fracturing behaviors of HBS and be an important guide for the design of on-site fracturing schemes. Nevertheless, the simulations were conducted under specific conditions, especially since the hydrate phase transition, sediment components, and geochemistry issues were not considered. Additionally, to better guide engineering, combining the DEM and FEM for mine-scale numerical modeling is also an important potential research direction.