The Role of Coal Mechanical Characteristics on Reservoir Permeability Evolution and Its Effects on CO2 Sequestration and Enhanced Coalbed Methane Recovery

Elastic modulus is an important parameter affecting the permeability change in the process of coalbed methane (CBM)/enhanced coalbed methane (ECBM) production, which will change with the variable gas content. Much research focuses on the constant value of elastic modulus; however, variable stiffness of coal during CO2 injection has been considered in this work. The coupled thermo-hydro-mechanical (THM) model is established and then validated by primary production data, as well as being applied in the prediction of CO2/N2-ECBM recovery. The results show that the harder coal seam is beneficial to primary production, while the softer coal seam results in greater CO2/N2-ECBM recovery and CO2 sequestration. N2 and CO2 mixture injection could be applied to balance early N2 breakthrough and pronounced matrix swelling induced by CO2 adsorption, and to prolong the process of effective CH4 recovery. Besides, reduction in stiffness of coal seam during CO2 injection would moderate the significant permeability loss induced by matrix swelling. With the increase of the weakening degree of coal seam stiffness, CO2 cumulative storage also shows an increasing trend. Neglecting the weakening effect of CO2 adsorption on coal seam stiffness could underestimate the injection capacity of CO2. Injection of hot CO2 could improve the permeability around injection well and then enhance CO2 cumulative storage and CBM recovery. Furthermore, compared with ECBM production, injection temperature is more favorable for CO2 storage, especially within hard coal seams. Care should be considered that significant permeability change is induced by mechanical characteristics alterations in deep burial coal seams in further study, especially for CO2-ECBM projects.


Introduction
Coal seams are typified as dual-porosity systems consisting of micropores in a matrix and two sets of cleats. Coal matrix contains more than 95% adsorption of methane (CH 4 ) with the cleat systems providing an effective flow path for both water and gas [1,2]. Primary coalbed methane (CBM) recovery begins with the dewatering to reduce reservoir pressure and increase gas effective permeability [3,4]. However, it is generally acknowledged that less than 50% of methane in situ could be extracted by traditional method due to higher adsorption capacity of coal, which is not efficient or economical [5,6]. Therefore, CBM production enhancement techniques have been conducted in some field trials to stimulate the methane recovery rate in recent twenty years.
The ECBM recovery technique is possible using two methods, involving injection of N 2 (N 2 -ECBM) and CO 2 (CO 2 -ECBM) [7][8][9]. Both the two injectants could reduce the CH 4 partial pressure in the cleats, then promote CH 4 desorption from the coal matrix to achieve the new partial pressure equilibrium [10,11]. Besides, CO 2 can also displace CH 4 from coal seams due to a greater affinity to coal [12]. N 2 is weakly adsorbed than CH 4 in coal, which results in a better sweeping efficiency, and thus is mainly maintained in the free gas phase [13]. This process is also referred to as CH 4 stripping. As an added benefit, injecting N 2 /CO 2 can sustain the positive effects of a higher total reservoir pressure on permeability and accelerate the gas flow rate by adding the additional driving force [14]. For the project of CO 2 -ECBM, besides the effect of ECBM recovery, accompanying geological storage of CO 2 can be viewed as a potential means to mitigate greenhouse gas emission [15]. Since the first field trial of CO 2 -ECBM in the Allison unit, San Jan Basin in 1995, a number of field pilots were conducted in Canada, Japan, Europe, and China, subsequently [16]. Meanwhile, N 2 , which is cheap and abundant, has been introduced to overcome pronounced permeability loss during CO 2 injection in several field applications, including Yubari pilot, Japan and Tiffany unit, San Jun Basin [17].
Coal permeability is an important parameter for primary CBM production or ECBM recovery. There are generally two competing effects on the absolute permeability, i.e., changed effective stress and coal matrix shrinkage/swelling. For primary production, depressurization increases the effective stress and causes a reduction in the permeability due to cleat compression, and then the permeability may tend to rebound due to coal matrix shrinkage [18][19][20]. However, CO 2 has a great affinity to coal than CH 4 and N 2 ; thus, significant matrix swelling caused by CO 2 adsorption may result in permeability and well injectivity loss during CO 2 injection [21,22], which has been one of the technical obstacles suffered in CO 2 -ECBM recovery or CO 2 storage. A reduction of over two orders of magnitude in injection well permeability of Allison CO 2 -ECBM pilots has been reported [23]. Unlike primary production, where permeability changes due to coal matrix shrinkage may show an important effect at the late production stage, net matrix swelling induces severe permeability loss which can be observed at early or whole stages of CO 2 injection [17]. Furthermore, the dramatic reduction in injectivity and permeability has not been observed in field trials and laboratory tests, where pure N 2 or flue gas was used [11,24,25] due to net matrix shrinkage caused by a much lower sorption capacity and strain of N 2 .
Coal mechanical properties are important in the design of primary CBM/ECBM recovery and CO 2 sequestration due to the influence of elastic modulus on controlling the stiffness of the cleat systems [26]. Compared with N 2 injection, CO 2 adsorption could not only cause matrix swelling but also pronouncedly reduce the stiffness and strength of coal (9.6%-82.1% for uniaxial compression; 12.16%-20% for triaxial compression), and consequently the permeability changes. To date, many field and laboratory experiments have been conducted on investigating the mechanical behavior of CO 2 interaction with coal [27][28][29][30][31]. The consequent mechanical properties alteration depends on its geoenvironment and coal seam characteristics, including confining stress, CO 2 adsorption pressure, CO 2 phase state, CO 2 saturation time, coal rank, cleat density and orientation, and moisture [26]. For instance, CO 2 -induced coal strength and stiffness reduction is comparatively less under higher in situ stress state [27,32,33], due to the decreased CO 2 adsorption and potentially hindered by matrix swelling. With the increase of adsorption pressure, the mechanical degradation caused by CO 2 adsorption is also gradually elevated [33]. Similarly, the CO 2 phase changes into supercritical state would create a significant strength and stiffness reduction since the higher adsorption capacity and polymerization capacity than that in the subcritical state. Note that the reduction of the mechanical strength and the stiffness (elastic modulus) with the CO 2 saturation pressure could be mathematically described by Langmuir-type curves [26,34]. Moreover, almost overall previous investigations imply that mechanical degradation largely occurs during the initial exposure to CO 2 , while the additional CO 2 exposure only reduces the strength parameters slightly. There are a number of factors that contribute to the comprehensive coal mechanical properties alteration, including changed surface energy, plasticization and swelling effects [27,35], microcracks induced by differential swelling/shrinkage strain [36], and dissolution of minerals due to chemical interactions [26,37,38]. Therefore, there might be the third effect that dominates the evolution of permeability besides effective stress and matrix strain-the stiffness of coal decreases, which would enhance the positive effect of the reduction in effective stress and moderate the significant matrix swelling (the influence may be more pronounced under supercritical CO 2 state) during CO 2 injection. The ultimate performance of reservoir permeability should be the result of coupling and competition of the above three aspects.
Understanding the mechanism of coal mechanical properties on coupled processes is key to evaluate the efficiency of CO 2 sequestration and ECBM recovery. Especially, the anomalous phenomenon, which shows enhanced injectivity in a few CO 2 -ECBM field trials [39] and CO 2 permeability rebound in laboratory experiments [40], might be potentially attributed to the CO 2 -coal interaction. In this study, we firstly address these key issues of a lack of investigations-involving geomechanical response on the effect of CO 2 /N2-ECBM recovery through developing coupled thermo-hydromechanical (THM) modeling. Then, the modeling is employed to inversing the reservoir permeability evolution, CH 4 production rate/cumulative production, CO 2 storage rate/cumulative storage, and CO 2 /N 2 breakthrough. Additionally, analysis and discussion are offered to link the modeling results with the field/laboratory observations.

Coupled Model for THM
2.1. Conceptual Model and Assumptions. CO 2 /N 2 -ECBM recovery involve the processes of competitive adsorption induced by a more reactive gas (CO 2 ), inert gas stripping (N 2 ), gas diffusion between matrix and cleats, gas and water two-phase flow in cleats, and heat transfer, together with coal deformation induced by the change in effective stress. The general processes of complex interactions are manifest in the response to THM coupling in the coal seams and schematically shown in Figure 1. A set of field-governing equations are defined that govern coal deformation and the where ρ p is the gas or water density, kg/m 3 ; v p indicates the Darcy velocity of the gas or water phase, m/s; t denotes the time, s; Q s is a source term, kg m -3 s -1 ; and m indicates the methane or water content, kg/m 3 . The binary gas content in the REV comprises both freephase gas in the fractures (m f g ) and adsorbed gas content (V sg ) in matrix and is expressed as where ρ c is the coal density, kg/m 3 ; ρ gsi represents the gas density under standard conditions (the subscript i represents the gas component, i = 1 for CH 4 and i = 2 for CO 2 /N 2 ); φ f denotes the fracture porosity; S g indicates the gas saturation; ρ f g is the free gas density within the fractures, m 3 /kg; ρ mg represents the free gas density in the matrix micropores, m 3 /kg; and p m is the gas pressure in the matrix, Pa. Then, according to the ideal gas law, the gas density can be described as ρ f gi = M gi p f gi /RT, where M gi is the molar mass of the gas component i, g/mol; R is the universal gas content, J/(mol·K); and T is the reservoir temperature , K. The absorbed gas content in per unit coal mass under variable temperature can be improved by the modified Langmuir equation [41][42][43]: where c 1 and c 2 are the thermal coefficients of gas adsorption; V Li is the Langmuir volume constant of gas component i,  Figure 1: Mass transport processes for binary gas (CO 2 /N 2 , CH 4 ) and water during CBM, CO 2 /N 2 -ECBM recovery, and injection heat.

Geofluids
Then, the water content in the REV can be expressed as follows: where ρ w is water density, kg/m 3 , and S w is water saturation, S w + S g = 1.
The reservoir pressure is defined as [42] where p f g is gas pressure in the fractures, MPa, and p f w indicates the water pressure in the fractures, Pa. The relationship between the gas pressure and water pressure can be expressed as [44] where p cgw is the capillary pressure, Pa. Buoyancy is not considered for both gas and water, defining Darcy's law for two-phase flow in the fractures as where the subscripts g and w refer to the gas and water, respectively; v gi and v w indicate the Darcy law velocity of the gas component i and water, respectively, m/s; μ gi and μ w denote the dynamic viscosity; k eg and k ew are the effective permeability of the gas and water, respectively, m 2 . The mass exchange between matrix and fractures are dominated by diffusion, which may be defined as [45,46] where q is the gas exchange rate between the matrix and the fractures, kg/(m 3 ·s); σ c indicates the coal matrix block shape factor, m -2 ; D is the gas diffusion coefficient, m 2 /s; ρ mg indicates the concentration of gas in the matrix blocks, kg/m 3 ; and we use adsorption time to estimate the effective gas diffusion coefficient in the coal matrix as [47] where τ is the sorption time of the coal matrix, which is numerically equivalent to the time for 63.2% of the coal gas to be recovered, s.

Governing Equation of the Mechanical Field.
For a homogeneous, isotropic, and elastic medium, the straindisplacement relationship and the equilibrium equation can be expressed as follows: where ε ij represents the strain tensor (i, j = 1, 2, 3); u i is the displacement within the element; σ ij is the total stress tensor; and σ e ij is the effective stress tensor, and effective stresses are defined as σ e ij = σ ij − αp f . The constitutive relationship of an isotropic linear poroelastic medium is expressed as [48] where G = E/2ð1 + υÞ is the shear modulus, Pa; K = E/3ð1 − 2υÞ represents the bulk modulus of the coal, and K s represents the bulk modulus of the coal grains, Pa; α is the Biot coefficient and can be expressed as α = 1 − K/K s ; E is Young's modulus of the coal seam, Pa; E s is Young's modulus of the coal grains, Pa; υ is Poisson's ratio; δ ij is the Kronecker delta tensor defined as 1 for i = j and 0 for i ≠ j; f i denotes the components of the body forces; σ kk = σ 11 + σ 22 + σ 33 . ε s = ε sg1 + ε sg2 is the total volumetric strain of matrix swelling/shrinkage induced by binary gas adsorption/desorption; α T is the 4 Geofluids thermal expansion coefficient, 1/K; and ε T = α T T is the thermal expansion volumetric strain. By analogy, the extended Langmuir-type equation is then used to define the sorption-induced volumetric strain, which can be expressed as where ε Li is the maximum volumetric strain of gas component i. The Langmuir-type curve is used to describe the elastic modulus reduction (ΔE), and elastic modulus of CO 2 saturated coal mass can be written as [26,34] where p CO2 is the gas component of CO 2 , Pa; ΔE max is the maximum reduction in the elastic modulus, Pa; and P E is the curve-fitting parameter, Pa. Combining Eqs. (14), (15), and (16) yields a modified Navier-type equation defining deformation.

Governing Equation of
the Thermal Field. The energy conservation equation of the skeleton and fluid can be obtained based on the energy conservation law. For the projects of CBM and CO 2 /N 2 -ECBM recovery, variation of internal energy within REV caused by temperature change is equal to the sum of heat convection of fluids (binary gas and water), heat conduction among the solid and fluid phases, strain energy induced by skeleton deformation, and isosteric heat induced by gas adsorption. The governing equation of the thermal field can be expressed as [49][50][51] where C T c is the effective specific heat capacity of the coal mass, J/(m 3 K); η ef f is the effective heat convection coefficient of the fluids, J/(m 2 ·s); λ T c is the effective thermal conductivity, W/(m K); and q sti is the isosteric heat of gas adsorption of component i, kJ/mol.
where C s , C gi , C w are the specific heat capacities of coal skeleton, binary gas, and water, respectively, kJ/(kg·K); λ s λ gi , λ w are the thermal conduction coefficients for the coal skeleton, binary gas, and water, respectively, W/(m·K).

2.2.4.
Cross-coupling. The permeability and porosity represent the key cross-coupling parameters linking the multiphysic fields. The cubic law is widely applied to describe the absolute permeability change relative to the porosities as where the subscript 0 refers to the initial state and k, φ are the absolute permeability and porosity, respectively. Based on the constitutive relationship of coal mass (Eq. (16)), the volumetric strain of the REV is expressed as where σ − = −σ kk /3 is the mean compressive stress. Considering a porous medium containing solid volume of V s and pore volume of V P , we assume the bulk volume V b = V P + V s , the porosity, and its differential form can be expressed as Eq. (27) and Eq. (28), respectively.
According to Eq. (4), the volumetric evolution of the porous medium can be described in terms of the volumetric strain of coal mass (dε b ) and pore space (dε P ), respectively. The relations are where β = 1 − K P /K s . By substituting Eqs. (29) and (30) into Eq. (28) and recognizing that the solid matrix modulus (K s ) is commonly several orders of magnitude larger than the pore volume modulus (K P ), we can obtain 5 Geofluids We assume that the volumetric variation of the porous medium satisfies the Betti-Maxwell reciprocal theorem ∂V b /∂p⋮ σ − = ∂V P /∂σ − ⋮ P [48], and we obtain Eq. (32).
The Biot's coefficient is considered one in this study, due to the soft coal seams; thus, we obtain K P = φK [21]. By assuming φ ≪ 1%, Eq. (31) can be integrated as and then dpÞ; thus, it can be expressed as By substituting Eq. (34) into Eq. (25), the permeability of cleats is obtained as where C f = 1/K P is cleat volume compressibility. For the conditions of uniaxial strain (ε xx = ε yy = 0 and σ xx = σ yy ), the horizontal stress σ xx or σ yy is given from Eq. (15) and Eq. (20) as We assumed that the reservoirs are under constant vertical stress (Δσ zz = 0), and the changed mean stress becomes Substituting Eq. (37) into Eq. (35) and combining the relationship between C f and φ 0 yield Water and gas are coexisting in many CBM reservoirs. Therefore, the effective permeability, as a function of the relative permeability with the absolute permeability, represents the most significant parameter for the two-phase flow. The relative permeability models of Eqs. (39) and (40) at saturation S w are widely used [52]. Gas slippage is not considered, and the gas/water endpoint relative permeability is calibrated in this study; a dynamic effective permeability model may be expressed as where k rg represents the relative permeability of the gas and is dimensionless; k rw is the relative permeability of water; k rg0 indicates the endpoint relative permeability of the gas; k rw0 denotes the endpoint relative permeability of water; S wr represents the irreducible water saturation fraction; and S gr is the residual gas saturation fraction.

Coupled
Relationship. The hydraulic, mechanical, and thermal fields are defined by Eqs. (12), (13), (20), and (21), and the cross-coupling term of Eqs. (38), (39), and (40) complete the THM coupled model, as shown in Figure 2. These equations are implemented into the software of COMSOL Multiphysics to solve for reservoir evolution of CBM/ECBM recovery. Figure 3 shows the solution process for the model by COMSOL Multiphysics.

Model Validation and Simulation Schemes
3.1. Model Description. Four CO 2 -ECBM field projects in China have been completed (three in the Qinshui Basin and one at the eastern margin of Ordos Basin), and the Qinshui Basin is one of the most representative commercial CBM basin [53]. In 2002, a pilot testing of CO 2 -ECBM in deep unminable coalbed in Qinshui Basin was undertaken by Chinese Commerce Department and the Canadian International Development Agency [15]. The target formation is the Permian Shanxi formation #3 coal seam of uniform thickness (~6 m), high CBM content (28.9-30.5 m 3 /t), permeability (0.002-12.6 mD), burial depth (~472-972 m), and reservoir pressure (2.4-6.1 MPa) [54]. The main parameters for the basic geological model are shown in Table 1-mainly obtained from related literature. Vertical well spacing with in situ primary production is usually arranged on a rectangle pattern of 300 × 300 m as shown in Figure 4(a). Moreover, the multiwell pilot testing-an injection well-is located at the center of a near-square array of four production wells in a traditional five-spot pattern [55] (Figure 4(b)). Quarter of the near-regular five-spot pattern is represented by a 150 × 150 m block to simulate the CO 2 /N 2 -ECBM pilot test (Figure 4(c)). The reference section (Line A-B) and three points (P1, P2, and P3) are used to investigate the reservoir parameters evolution. This study is performed in two parts by the proposed CBM simulation model. The established THM model is validated by using history matching of pressure depletion production of a typical production well and then implemented for the performance prediction of CO 2 /N 2 -ECBM production and CO 2 sequestration. Table 2 lists the parameters used in the study of model validation. Table 3 lists the related parameters for the simulation of CO 2 /N 2 -ECBM recovery.

Model Validation.
History matching is used to conduct the model validation, and the field data is obtained from some scholars, who have reported the CBM production rate from an unstimulated production well subject to pressure deletion recovery in situ of Qinshui Basin. Unsurprisingly, the simulated gas production in Figure 5 is not perfectly consistent with the actual production shown, especially for the time of actual peak gas production (slightly lagging behind). This might be mainly attributed to permeability anisotropy, heterogeneity, and the single-phase flow of water during the dewatering stage (no gas production) in the field. Coincidentally, the average relative error of CH 4 production between this simulation and field data is~16% ( Figure 5), with the corresponding of values for the other study of 16.3% [43]. Note that, compared with the values of gas production rate, the shape of simulated gas production profile should be more concerned. Although the average relative error, both in this study and the literature, are consistent, the gas production profile of this research is more in agreement with the field profile. It indicates that the mathematical model of THM coupling can be used to simulate the primary CBM production, as well as extend the CO 2 /N 2 -ECBM production prediction.

Effect of Different Injection Gas on ECBM
4.1.1. CH 4 Recovery. During primary production and pure CO 2 /N 2 -ECBM recovery, CH 4 production rates all show a trend of rising first as the coalbed water continuously discharged, and then reduction ( Figure 6(a)). The peak production rates for primary injection of CO 2 and N 2 are 304.6, 406.1, and 1615.1 m 3 /day, with the corresponding time of 308, 906, and 423 days, respectively. Compared with primary production, injection of CO 2 and N 2 not only elevate CH 4 recovery rate (even up to 33.3% and 432%) separately but also delay the peak production rate. The cumulative CH 4 production for the projects of primary production and CO 2 /N 2 -ECBM recovery at 4000th day reach 0:68 × 10 6 , 1:21 × 10 6 , and 1:73 × 10 6 m 3 , respectively ( Figure 6(b)).
Making the case of primary as a reference, the cumulative productions for CO 2 /N 2 -ECBM recovery are increased by 78.3% and 155.3%, separately. For primary production, the recovery ratio at 4000th day is 25.1%, with the corresponding values for CO 2 /N 2 -ECBM recovery of 44.8% and 62.7%, separately. Meanwhile, the enhancement factors, which are defined as the proportion of enhanced recovery ratio to that of primary production, are 1.78 and 2.5. All those state that the CH 4 recovery might be effectively enhanced by reactive gas (CO 2 ) or inert gas (N 2 ) injection, and the enhancement effect of N 2 injection is better than that of CO 2 . However, an important issue cannot be ignored during enhanced recovery-CO 2 /N 2 breakthrough, which may cause deterioration of produced gas purity and decrease the calorific value-thereby early well shutdown. Especially, N 2 breakthrough occurs shortly after the start of injection (~265 days), and it is attributed to lower dynamic viscosity and weak adsorption capacity of N 2 (Figure 7(b)). In contrast, the time of CO 2 breakthrough will be dramatically delayed (~1500 days) due to the larger  7 Geofluids adsorption capacity decreasing the passing ability (Figure 7(a)). Therefore, there is a tradeoff between incremental CBM production and the earlier N 2 breakthrough. For instance, N 2 and CO 2 mixture injection could be applied to balance early N 2 breakthrough and pronounced matrix swelling induced by CO 2 adsorption and prolong the process of effective CH 4 recovery.

Permeability
Ratio. Similar to the findings of previous studies [42][43][44], with the primary production continued, reservoir permeability at different reference points first decreases due to the increase of effective stress and then rebounds due to CH 4 desorption-induced matrix shrinkage (Figure 8(a)). With the approaching to the production well, permeability rebound appears earlier, and the ultimate permeability recovery is also the largest. The minimum permeability ratios at points P1, P2, and P3 are~0.947 (672 days),~0 .949 (1117 days), and~0.949 (1253 days), and these points reach the maximum values of~1.045,~1.012, and 1.001 at 4000 days.
However, the permeability evolution during CO 2 /N 2 -ECBM becomes more complex, compared to primary production. In the case of CO 2 -ECBM, near the production well (P1), the permeability ratio evolution is similar to primary production in early time, while decreases dramatically with the arrival of CO 2 due to continued injection (Figure 8(b)). The minimum permeability ratio at point P1 is 0.54 at 4000 days. In the middle of the reservoir (P2), the permeability remains stable over the first 420 days due to the dual opposing effects-decreasing effective stress and matrix swelling-and then continuously decreases. Near the injection well, it is noted that the permeability first slightly increases due to the dominant factors of the reduction for effective stress and sharply declines due to CO 2  18) and (19).

The updated material parameters
Increase time step i = i+1 Reach the maximum weakening stiffness ?
Near the production well, the increase of effective stress, matrix shrinkage, and swelling are the main controlling factors of permeability evolution, successively. Near the injec-tion well, the decrease of effective stress and matrix swelling are the dominant factors, and then the final reservoir permeability will decrease to less than 60% of the initial value, which will further reduce the CO 2 injection rate. The net matrix shrinkage caused by N 2 injection and sustaining the total cleats pressure as the double positive effect factors make Table 2: Key parameters for model validation.

Effect of Coal Stiffness on ECBM.
The simulation scheme for the effect of coal mechanical properties on ECBM is shown in Table 4.

CH 4 Production.
During the primary CBM recovery under different mechanical properties of coal seam, CH 4 production rates all show a trend of sharply decreasing first due to the rapid release of free gas in the coal seam near the production well. Subsequently, CH 4 recovery rates first increase in early time and then reduce at later time. The peak production rates for different scenarios of mechanical properties (from lower stiffness to higher stiffness) are 204, 226, and 250 m 3 /day, respectively-illustrating that the higher stiffness of coal seams would result in a larger CH 4 production rate (Figure 9(a)). Making the case of lower stiffness of coal seam of CH 4 cumulative production at 4000 days (0:57 × 10 6 m 3 ) as a reference, the corresponding CH 4 cumulative values under the medium and higher stiffness are increased by 14% (0:65 × 10 6 m 3 ) and 22.8% (0:7 × 10 6 m 3 ), separately (Figure 10(a)). For CO 2 -ECBM, the CH 4 recovery rate at early production is greater for the case of higher stiffness; however, the peak production rate for lower stiffness of coal seam would be elevated and delayed, subsequently (Figure 9(b)). The   11 Geofluids peak production rates for different scenarios of mechanical properties (from lower stiffness to higher stiffness) are 494.6 (1544 days), 348 (1580 days), and 325.1 m 3 /day (1150 days), respectively. The CH 4 cumulative production for soft (lower stiffness) coal seam within 1000 days is lower than that scenarios of medium and higher stiffness, while the value would rebound dramatically after 1000 days (Figure 9(b)). Making the case of harder coal seam CH 4 cumulative production as a reference (1:06 × 10 6 m 3 ), by 4000th day, the corresponding CH 4 cumulative productions under the medium and lower stiffness are increased by 4.7% (1:11 × 10 6 m 3 ) and 28% (1:36 × 10 6 m 3 ), separately ( Figure 10(b)).
As analyzed in "Effect of Different Injection Gas on ECBM," injection of N 2 can significantly enhance CBM production rate and cause early N 2 breakthrough; thus, we reduce the initial permeability to 0.1 mD. Different with CO 2 -ECBM, with the decrease of coal seam stiffness, the peak production rate for N 2 -ECBM would be elevated and advanced. The peak production rates for different scenarios    (Figure 9(c)). However, it is noted that the production rate for harder coal seam is larger than those cases of medium and lower ones during the decline stage, and this phenomenon is similar to the primary CH 4 production. With the CBM recovery continued, CH 4 cumulative production for coal seams with different mechanical properties gradually increases. Making the case of harder coal seam CH 4 cumulative production at 4000th day as a reference (0:65 × 10 6 m 3 ), the corresponding CH 4 cumulative productions under the medium and lower stiffness are increased by~27.7% (0:83 × 10 6 m 3 ) and~55.4% (1:01 × 10 6 m 3 ), separately (Figure 10(c))-illustrating that N 2 -ECBM in softer coal seam is more favorable. Compared with primary recovery and CO 2 -ECBM, N 2 -ECBM is more sensitive to the stiffness of coal seam.

CO 2 Storage.
The CO 2 -ECBM project not only recovers additional CBM in unminable coal seams-utilizing CO 2 displacement and sweeping-but also effectively sequesters greenhouse gas. The CO 2 storage rates first increase and then decline as the coalbed water continuously discharged. With the decrease of coal seam stiffness, the peak storage rates are elevated and delayed (Figure 11(a)). The peak CO 2 storage rates of lower, medium, and higher stiffness CBM reservoirs are~1132,~687, and~580.5 m 3 /day, respectively, appearing at~1610, 1580, and 1397 days. 14 Geofluids Although, the CO 2 storage rate of lower stiffness CBM reservoirs is larger than the other scenarios during dewatering and stable production stage, it would decrease sharply at decline stage and even much lower than the other scenarios after 2750 days. The CO 2 cumulative storage decrease from the lower stiffness to the medium stiffness, and then higher stiffness, with a maximum cumulative storage of the three scenarios reaching approximately 2:8 × 10 6 , 2:25 × 10 6 , and 1:96 × 10 6 m 3 at the 4000th day ( Figure 11(b)), respectively. Making the case of harder coal seam of CO 2 cumulative storage as a reference, the corresponding cumulative storage under the medium and lower stiffness is increased bỹ 14.8% and~42.9%, separately, indicating a significant and practicability CO 2 storage capacity within the soft coal seam.

Permeability Evolution.
For the project of primary production, due to the small decline in temperature and princi-pal matrix shrinkage over the negative effect of increased effective stress, matrix shrinkage dominates the evolution of permeability near the production well, thus leading to an increase in permeability. On one hand, the dominant factor (matrix shrinkage) is gradually weakened as it is far away from the production well. On the other hand, with the decrease of coal seam stiffness, the enhanced stress sensibility of coal permeability results in lower permeability ratio distribution within CBM reservoir (Figure 12(a)). For instance, the permeability ratio of the entire soft coal seam is almost below the initial permeability during the whole production process and even decreases by 20%. In contrast, the reservoir permeability for medium and higher stiffness coal seams would exceed their corresponding initial values at 4000th day. Making the point P2 as a reference, the higher stiffness of coal seam and the earlier permeability rebound and recovery appear (Figure 12(b)). The minimum permeability ratios (permeability rebound) of higher and medium stiffness CBM reservoirs are~0.99 (at 370 days) and~0.95 (at 1100 days), and these reach a maximum ratio of 1.16 and 1.01, separately, at 4000 days. However, the permeability ratio of soft coal seam decreases dramatically over time, from 1 to 0.81 (at 4000 days), and there is no rebound or recovery of permeability due to the dramatically negative effect of increased effective stress.
For the project of CO 2 -ECBM recovery, permeability evolution around the production well is similar to primary CH 4 recovery. Conversely, near the injection well, CO 2 injections cause the increase of reservoir pressure, reduce the effective stress, and then weaken the significant matrix swelling induced by CO 2 adsorption. Therefore, the softer the coal seam is, the higher permeability ratio would reach (Figure 13(a)). In addition, matrix swelling still dominates the evolution of permeability, especially for medium and higher stiffness coal seams; thus, reservoir permeability drops sharply during the whole CO 2 -ECBM production project. Making the point P2 as a reference, the permeability ratio of medium and higher stiffness of coal seam first decreases slightly due to the increase of effective stress and   16 Geofluids then reduces sharply due to CO 2 arrival (Figure 13(b)). However, the permeability ratio curves of soft coal seam exhibit undulated shape due to the more sensitive effect of effective stress (Figure 13(b)). Compared to hard coal seams, the permeability of soft one is increased bỹ 35.1% at 4000th day. For the project of N 2 -ECBM recovery, the permeability evolution is simple due to net matrix shrinkage. Near the production well, the permeability distributions under different scenarios are also similar to CO 2 -ECBM and primary production, before N 2 breakthrough. With approaching the injection well and N 2 injection continued, double positive effects-net matrix shrinkage and reduction of effective stress-dominate the permeability evolution; thus, reservoir permeability would increase dramatically, exhibiting the softer the coal seam, the higher the permeability (Figure 14(a)). Compared with hard coal seams, the permeability of soft one is increased by~63.1% at 4000th day (Figure 14(b)).
Congruent with those previous studies [57,58], the elastic modulus of the coal seam has a significant effect on the  17 Geofluids permeability within the reservoir. It can be illustrated that fluid injection into a more deformable reservoir (lower elastic modulus) opens up the fractures more easily and then results in a higher value of permeability, compared to a hard reservoir (with larger modulus).

CO 2 /N 2 Breakthrough.
With the rising stiffness of coal seam, the migration rates of CO 2 and N 2 gradually decrease in CBM reservoirs (Figures 15 and 16). CO 2 breakthrough for lower, medium, and higher stiffness coal seams appears at 1415, 2046 and 2400 days, respectively, when the CO 2 partial pressure begins to increase (Figure 17(a)). The corresponding values for N 2 breakthrough are 426, 600, and 1126 days, respectively (Figure 17(b)). It is noted that CO 2 breakthrough is delayed slightly from the condition of medium stiffness to higher stiffness for CO 2 -ECBM. In contrast, N 2 breakthrough is delayed sharply under the same condition ( Figure 18). The major reason for this phenomenon could be attributed to the double positive effects of permeability evolution on N 2 injection and the single positive effect on CO 2 injection caused by permeability change. The    18 Geofluids increase of stiffness restrains the double or single positive effect on permeability evolution during N 2 or CO 2 injection. Note that the restraint for CO 2 injection are weakened due to the dominant factor of matrix swelling induced by CO 2 adsorption. However, the restraints for N 2 injection are significant due to the potential for transforming from double positive effect to singe positive effect-the net matrix shrinkage.

Effect of CO 2 Interaction Induced Mechanical Property
Alteration on CO 2 -ECBM. Compared with N 2 /CH 4 adsorption, CO 2 adsorption not only induces matrix swelling but also accompanies the reductions in stiffness and strength and then shows a significant effect on permeability changes and CO 2 storage. CO 2 storage and reservoir permeability evolution are investigated due to the reduction in stiffness in this paper. The simulation scheme is shown in Table 5.

CO 2 Storage.
With the increase of the weakening degree of coal seam stiffness, both CO 2 storage rate and CO 2 cumulative storage all show a trend of increase. The peak CO 2 storage rates of no stiffness weakening and weakening degree from 10% to 30% are~684.8,~714.9,~753.5, and 804.2 m 3 /day, respectively (Figure 19(a)). The CO 2 cumulative storage of the conditions gradually increases, with a maximum cumulative storage reaching approximately 2:25 × 10 6 , 2:32 × 10 6 , 2:40 × 10 6 , and 2:50 × 10 6 m 3 at the 4000th day, respectively (Figure 19(b)). Making the case of no weakening as a reference, the corresponding cumulative storage under the weakening degree from 10% to 30% is increased by~3 .1%,~6.5%, and 11.1% separately, indicating the higher the weakening degree, the better the CO 2 storage.

Permeability
Ratio. CO 2 interaction with coal seam would induce mechanical characteristics alterations, such as elastic modulus-which controls the stiffness of the coal cleat systems. Reduction in elastic modulus can alleviate the sharp decline of permeability caused by CO 2 adsorption due to the elevated positive effect of effective stress decrease. The process is positive feedback-CO 2 injection decreases the elastic modulus and then causes a large injection volume, and a large injection volume would also cause a large scale and degree of weakening on coal seam stiffness. The lower elastic modulus is firstly distributed near the injection well and then propagates to the production well subsequently. With the increase of the maximum weakening degree of coal seam stiffness, sharper reduction of coal seam stiffness appeared ( Figure 20). Variation of stiffness will affect the dynamic permeability. The evolution of permeability ratio along the reference section A-B is divided into three zones before CO 2 breakthrough (Figure 21(a)). Making the case of 400th day of production, Zone 1 represents the dominant factor caused by the decline of effective stress under different weakening degrees. The weakening degree of 30% shows a larger permeability ratio than 20% and 10%, due to the more significant effect of effective stress. However, the larger reservoir permeability will cause more CO 2 injection, and then matrix swelling becomes the dominant factor of permeability changes within Zone 2-indicating that the permeability ratios decreases slightly from the stiffness weakening degree of 10% to 20%, and then 30%. The same phenomenon also occurs at the reference point P2. By the 4000th day, the minimum permeability ratios of stiffness weakening degree from 10% to 30% at this point is 0.61, 0.63, and 0.66, respectively ( Figure 21(b)). For Zone 3, reservoir permeability is controlled by two opposing effects-increased effective stress caused by CH 4 pressure deletion and matrix shrinkage induced by CH 4 desorption-due to no CO 2 arrival at this moment. The ultimate permeability distributions within the entire reservoir show a trend alike with Zone 1, indicating the higher stiffness weakening degree caused by CO 2 adsorption, the larger permeability ratio reached ( Figure 22). Therefore, neglecting the weakening effect of CO 2 adsorption on coal seam stiffness will underestimate the injection capacity of CO 2 .

Effect of Injection
Temperature on CO 2 -ECBM. Gas desorption consumes energy, causing a dropping in reservoir temperature; conversely, gas adsorption would release energy and then elevate the reservoir temperature. For the project of injection temperature in CO 2 -ECBM, the evolution of reservoir temperature is the competitive result of binary gases adsorption/desorption and injected thermal field. However, due to the large volume of coal seam, the migration rate of the apparent temperature rise frond is restricted, resulting in a limited extend of this elevated temperature zone. Taking a case of injection temperature at 320 K, near injection well, reservoir temperature sharply decreases due to the large amount of CH 4 desorption caused by hot CO 2 injection. With the injection continued, reservoir temperature gradually propagates within coal reservoirs, and then it is elevated, while the temperature around production well is lowest due to the net CH 4 desorption caused by reservoir pressure depletion ( Figure 23). With the increasing of injection temperature, both CH 4 production rate and CO 2 storage rate are elevated. The peak production rates for injection at 320, 340, 360, and 380 K are 358.0, 367.8, 375.1, and 380.5 m 3 /day, respectively (Figure 24(a)). The corresponding peak CO 2 storage rates are 716.7, 740.7, 760.5, and 775.8 m 3 /day, respectively (Figure 24(b))-indicating that higher injection temperature is favorable for CH 4 production and CO 2 storage. At 4000 days, the cumulative CH 4 production and CO 2 storage for no temperature injection are 1.11 and 2.25 million m 3 . The cumulative CH 4 production for injection at 320, 340, 360, and 380 K is increased by 1.8% (1:13 × 10 6 m 3 ), 3.6% (1:15 × 10 6 m 3 ), 4.5% (1:16 × 10 6 m 3 ), and 5.4% (1:17 × 10 6 m 3 ), respectively (Figure 25(a)). The corresponding CO 2 cumulative storage is increased by 3.1% (2:32 × 10 6 m 3 ), 5.3% (2:37 × 10 6 m 3 ), 7.1% (2:41 × 10 6 m 3 ), and 8.4% (2:44 × 10 6 m 3 ), respectively-illustrating that the effect of injection thermal on ECBM production and CO 2 sequestration is gradually moderate with the increase of temperature ( Figure 25(b)). Therefore, there is a tradeoff between incremental production/storage and the cost of injection heat. Furthermore, compared with ECBM production, injection temperature is more favorable for CO 2 storage (8.4% versus 5.4% at 4000th day shown in Figure 26).
The mechanism of enhanced CBM recovery and CO 2 sequestration using temperature injection method can be attributed to injection thermal effects on gas adsorption, and thus the permeability evolution. Near the injection well, there is a region (Zone 1 in Figure 27), we defined the thermal dominant area. In this region, matrix swelling induced by CO 2 adsorption is eliminated in some degree due to the raising of reservoir temperature. In addition, higher temperature results in a large amount of CH 4 desorption and then significant matrix shrinkage. The absolute permeability around the injection well is improved by these factors, compared with the case of no injection. However, remote from the injection well (Zone 2 in Figure 27), the effects of thermal on permeability disappear due to the rapid reduction in thermal gradient-showing that lower temperature injection results in a greater permeability. Note that, although the improvement of permeability is slight, it is important for CO 2 injection and CO 2 -CH 4 displacement ( Figure 28). The effect of coal mass thermal expansion on permeability is not obvious due to the lower magnitude of thermal expansion coefficient (2:4 × 10 −5 1/K).  Figure 26: The cumulative CH 4 production and CO 2 injection.

Geofluids
With the analysis of "Effect of Coal Stiffness on ECBM," within the coal seam with weak stress sensitivity, the injection of hot CO 2 can moderate a high-stress zone created by matrix swelling near to the injection well and then improve the permeability to achieve more significant production and storage effect.

Discussion
In the process of CBM recovery, the reservoir pressure continues to decrease and the permeability will increase significantly, which is mainly due to the influence of matrix shrinkage. However, CO 2 sequestration and CO 2 -ECBM in deep coal seams usually face the problem of matrix swelling caused by CO 2 adsorption. It is generally recognized that CO 2 injection would be hindered by permeability loss due to the higher adsorption-induced coal matrix swelling in most field-scale pilots, which has also become one of the major obstacles to the implementation of the technology. But it is found in Alberta field trials, opposite to the case of reduction in CO 2 injection rate-CO 2 injectivity was even greater than for weakly adsorbing N 2 . This is attributed to the result of coal weakening [39], while noting that the impact has not been quantified. The increase of CO 2 injection rate has also been encountered in other field trials; for instance, the Allison CO 2 -ECBM pilot shows the reduction in injection rate during early times, and then the rebound in injectivity during later time. Nevertheless, the rebound in injectivity is believed to be due to the overall reservoir pressure reduction and resulting matrix shrinkage around injection wells instead of coal weakening. Three schemes are designed to investigate the effect of coal weakening on CO 2 injection as shown in Table 6. With the decrease of elastic modulus from 3 to 2.1 GPa (scenario 1), CO 2 injection rate is gradually reduced due to significant matrix swelling during the whole CO 2 -ECBM. With the minimum elastic modulus reaching to 1.4 GPa (scenario 2), CO 2 injection rate first decreases and then rebounds slightly during later times. In the case of elastic modulus decreased by 67% from 3 to 1 GPa (scenario 3), CO 2 injection rate shows a trend of slightly decreasing first and then rising sharply due to the   23 Geofluids more pronounced effect of effective stress than matrix swelling ( Figure 29). All these discussed above illustrate that the rebound in CO 2 injection rate can be partially attributed to coal weakening, when the elastic modulus decreases by more than 50%.
Interestingly, it is confirmed by many laboratory experiments that CO 2 injection will reduce permeability dramatically. However, the opposite phenomenon is yet shown in some laboratory experiments, with similar to the result of the Alberta field trial. For instance, Robertson and Christiansen [59] describe a new permeability equation derived for sorption-elastic media such as coal specifically for confining conditions found commonly in the laboratory, but not in the field. This model can especially useful when dealing with laboratory experiments where many of the other factors that cloud field measurements are eliminated. The model can be expressed as follows: where c 0 is the initial cleat compressibility, α is the cleat compressibility change rate, Δp is the change of cleat gas pressure, φ 0 is the initial porosity of the cleat, υ is Poisson's ratio, E is the elastic modulus, s L and p SL are sorption-induced Langmuir strain, and p 0 and p are cleat pressure at initial and current state, respectively. The Robertson-Christiansen permeability model (R-C model) shown in Eq. (43) was applied to coal permeability data measured in the laboratory under hydrostatic confinement pressure. The permeability data was taken from Robertson and Christiansen [40], and the main parameters are listed in Table 7. The proposed model shows a good match with the measured data from N 2 (Figure 30(a)). CH 4 had a larger sorption strain value than did N 2 . Permeability ratios first decrease due to matrix swelling caused by CH 4 adsorption and then rebounds due to the decline of effective stress. However, the prediction value of the R-C model underestimated the actual permeability (Figure 30(b)). Robertson and Christiansen [40] attribute the poor fit of permeability data to the difference between stress-free sorption-induced strain and constrained sorption-induced strain-the amount of sorption-induced strain measured under confining stress condition will be significantly less than that measured under unconfined state (stress-free). Therefore, freestanding sorption-induced strain could be used into the permeability model due to easily accessible, but it should be modified   24 Geofluids before being inputted into the model to account for the depression of sorption-induced strain caused by partially confined matrix blocks. Here, we can try to correct the effect of this discrepancy by decreasing the input parameter of sorption-induced strain from 0.00931 to 0.00325, and then the result shows that the modified model matched the measured data for CH 4 very well (Figure 30(b)). Though, the adsorption of CO 2 has a much larger sorption-induced strain (0.03527) than CH 4 (0.00931) and N 2 (0.00305). When CO 2 is a flowing fluid, the actual permeability ratio is still vastly underestimated by the prediction model (Figure 30(c)). The model is modified by sharply reducing the CO 2 sorptioninduced strain from 0.03527 to 0.00437, while there is still a larger difference between the theoretical value and the actual. Thus, considering the effect of CO 2 adsorption on coal mechanics-reduction in elastic modulus-we update the R-C model again, and then the predictions of this modified model are in perfect agreement with the permeability data from CO 2 . With the increase of CO 2 pressure from 0.7 to 5.32 MPa, the elastic modulus of this perfect model decreases by 63% from 2.7 to 1 GPa, illustrating that the effect of matrix swelling on permeability variation may be not pronounced compared to the decreasing effective stress under high confining pressure. CO 2 sequestration and CO 2 -ECBM projects are preferred to be carried out in deep unminable coal seams where supercritical CO 2 is most likely to be encountered due to a high enough in situ stress and temperature. In this case, care should be considered that CO 2 interaction induced significant mechanical characteristics alterations and then changed  [40]). 25 Geofluids permeability in coal. Therefore, it is necessary to further investigate the influence of CO 2 adsorption on coal mechanics under constrained adsorption state.

Conclusion
(1) Injection of CO 2 /N 2 all can reach the purpose of ECBM, while the significant coal matrix swelling induced by CO 2 adsorption can reduce the reservoir permeability by one order of magnitude at least, and thus, it is not conducive to continuous CO 2 injection and ECBM. The effect of N 2 -ECBM overmatches CO 2 injection. However, the net matrix shrinkage could cause a sharply increase of permeability and then early N 2 breakthrough (2) The elastic modulus of coal seam affects reservoir permeability by controlling the stiffness of the coal cleat system. Harder coal seam is beneficial to the primary production due to the larger permeability recovery. On the contrary, for the CO 2 /N 2 injection project, softer coal seam results in greater ECBM production and CO 2 sequestration. Compared to primary CBM recovery and CO 2 -ECBM, N 2 -ECBM is more sensitive to the stiffness of coal seam (3) CO 2 adsorption not only induces matrix swelling but also accompanies the reductions in stiffness of coal seam and then shows a significant effect on permeability changes and CO 2 storage. With approaching the injection well, reservoir permeability for different CO 2 weakening degree presents the distribution of "S" type, showing that permeability evolution is dominated by decreasing of effective stress around injection well and then controlled by matrix swelling remote from injection well. With the increase of the weakening degree of coal seam stiffness, CO 2 cumulative storage shows an increasing trend. Neglecting the weakening effect of CO 2 adsorption on coal seam stiffness will underestimate the injection capacity of CO 2 (4) Injection of hot CO 2 could improve the permeability around injection well and then enhance CO 2 cumulative storage and CBM recovery. The effect of injection thermal on ECBM production and CO 2 sequestration is gradually moderate with the increase of temperature. Furthermore, compared with ECBM production, injection temperature is more favorable for CO 2 storage, especially within hard coal seams.

Data Availability
The data used to support the findings of this study are available from the first author upon request.

Conflicts of Interest
The authors declare no conflict of interest.