Active Pressure Management for Seasonal Aquifer Gas Storage: From Artificial Gas-Cap Establishment to Cyclic Injection–Production

Abstract

Seasonal balancing of energy supply and demand in China is constrained by the limited capacity of existing salt-cavern and depleted-reservoir storage facilities. To evaluate aquifer storage as a supplementary option, this work investigates the Baiju coastal aquifer using a CMG-GEM compositional model and examines seasonal multi-cycle operation under active pressure management. Sensitivity analyses were carried out for the artificial gas-cap construction period, brine-well bottomhole pressure, daily gas discharge rate, vertical-to-horizontal permeability ratio, and cushion-gas composition (N₂, CH₄, and CO₂), with emphasis on injection–discharge performance, pressure response, and discharged-gas purity. The simulations show that extending cushion-gas injection moderates pressure oscillation during subsequent cycling, and the working-gas mole fraction in discharged gas rises above 97% by the fifth cycle, while values varied across the tested operating and cushion-gas scenarios. Within the fixed eight-well configuration, lowering the bottomhole pressure of brine-discharge wells was associated with higher injectivity, although the incremental benefit became limited once the pressure dropped below the initial reservoir pressure. Within the tested range, higher daily discharge rates were associated with lower capacity-expansion efficiency; the magnitude depends on the metric normalization. A lower vertical permeability mainly constrains gas injection, while its influence on gas discharge is comparatively minor. Among the candidate cushion gases, CH₄ provides the largest gas-cap volume and CO₂ provides the strongest damping of pressure disturbance; considering both economics and storage performance, N₂ provides the best overall compromise because its working-gas injection and discharge volumes are close to those obtained with CH₄.

1. Introduction

Natural gas [1,2] and hydrogen are often generated and consumed in different regions and seasons. Underground Gas Storage (UGS) and underground hydrogen storage (UHS) therefore play a key role in interseasonal regulation and supply–demand balancing. In the United States, about 386 UGS facilities distributed across 31 states are currently in service, of which roughly 12% are unused groundwater or saline aquifers [3]. In China, both salt-cavern storage [4,5] and depleted-reservoir storage [6,7] have progressed rapidly, yet the available formations are still insufficient to satisfy the growing demand for storage capacity.

Because of their wide distribution, aquifers represent an important geological option for storing natural gas and hydrogen [8,9], as well as for CO₂ sequestration [10], compressed-air energy storage [11,12], and thermal energy storage [13,14]. China is actively evaluating aquifer gas storage as a supplementary option, and pilot projects such as the Baiju aquifer storage project are currently under feasibility study.

Although aquifer gas storage has already been implemented in several countries, many scientific and engineering questions remain open. Bi et al. [15] showed that lower porosity and permeability can be favorable for working-gas recovery, whereas hysteresis reduces performance. Jadhawar [16] reported that hysteresis may suppress the hydrogen discharge rate while improving the purity of the withdrawn gas, and a subsequent study [17] identified heterogeneity and hysteresis as the dominant controls on underground hydrogen storage in a North Sea aquifer. Giacomi et al. [18] further demonstrated that stronger heterogeneity is detrimental to the recovery of injected gas. At the pore scale, Nazari et al. [19] found that mixed wettability mainly affects injection, whereas lower wettability more strongly influences recovery. Pereira et al. [20] compared seasonal and hourly cycles and suggested that heterogeneity, although unfavorable in some respects, can help restrain plume migration.

Beyond geological properties, operational choices also exert a strong influence on storage efficiency. Zamehrian et al. [21] indicated that thinner aquifers are more advantageous for recovering working gas. Mahmoodpour et al. [22] evaluated 22 controlling parameters and found that placing wells near the structural crest improves operational performance. Moradi et al. [23] showed experimentally that fluorinated surfactants can enhance injection and recovery in saline aquifers. Other studies have emphasized the importance of well spacing [24], injection–discharge strategy [25], heterogeneity [26,27], and the design of cushion-gas injection [28,29,30].

Operational safety is equally critical when evaluating aquifer storage schemes. Zoback et al. [31] warned that large-scale subsurface fluid injection and discharge may induce earthquakes. Bai et al. [32] showed numerically that aquifer storage can cause surface uplift closely linked to injection and discharge volumes. Previous studies on aquifer gas storage have mainly focused on geological controls (porosity, permeability, heterogeneity, hysteresis, wettability) or on individual operational parameters (well placement, injection–discharge strategy, cushion-gas design). The concept of active pressure management via targeted brine extraction was introduced by Birkholzer et al. [33], but only in the context of CO₂ sequestration, where the objective is permanent storage rather than cyclic operation. To the best of our knowledge, no previous study has coupled active pressure management (through brine-discharge wells with controlled bottomhole pressure) with the entire life cycle of an aquifer gas-storage facility, i.e., from artificial gas-cap establishment to seasonal multi-cycle injection–production.

To fill this gap, the present study investigates the Baiju aquifer on the eastern coast of China. A compositional CMG-GEM model was established, and active pressure management was implemented through constraints on brine-well bottomhole pressure and injection–discharge pressure limits. The influences of artificial gas-cap construction time, brine-well pressure, daily gas discharge rate, vertical-to-horizontal permeability ratio, and cushion-gas type (N₂, CH₄, and CO₂) were quantified in terms of injection–discharge efficiency, pressure stability, and gas purity. The present work fills this gap and further provides quantitative operational limits that can directly support the engineering design of the planned Baiju project.

2. Methodology

2.1. Numerical Model

All simulations were performed with the CMG 2021.10, which provides both black-oil and compositional (GEM) models. Although the black-oil model is computationally faster, the present problem involves multiphase and multicomponent gas behavior during injection and withdrawal, so the compositional approach was adopted. For the relatively simple gas systems considered here, GEM still offers acceptable computational efficiency while providing a more rigorous description of phase behavior and compositional effects. The compositional model contained 3 gas components (N₂, CH₄, and CO₂) and 2 phases (gas and water). In the GEM model, the mass conservation equation for the water phase is

$$-\nabla \cdot \left({\rho }_w{u}_w\right)\begin{array}{cc}={\displaystyle \frac{\partial }{\partial t}}\left(\varphi {\rho }_w{S}_w\right)+{\rho }_w{q}_w,& i=1\cdots n_c\end{array}$$

(1)

where ϕ is porosity, ρ_w is the density of the water phase, u_w is the velocity of the water phase, and q_w is the source term of the water phase. The mass conservation equation of component j in different phases i is expressed as follows:

$$\begin{array}{l}\begin{array}{cc}-\nabla \cdot {\displaystyle {\displaystyle \sum _{j=1}^{n_p}}\left(x_{\mathit{ij}}{\rho }_j{u}_j\right)}={\displaystyle \frac{\partial }{\partial t}}\left[\varphi {\displaystyle \sum _{j=1}^{n_p}\left(x_{\mathit{ij}}{\rho }_j{S}_j\right)}\right]+{\displaystyle {\displaystyle \sum _{j=1}^{n_p}}{x}_{\mathit{ij}}}{\rho }_j{q}_j,& i=1\cdots n_c\end{array}\end{array}$$

(2)

where x_ij is the phase ratios, ρ is the component density, u is the seepage velocity, and q is the injection or discharge rate. The momentum equation for fluid flow in porous media follows Darcy’s law. The flow velocity of each phase can be expressed as follows:

$$u_j=-{\displaystyle \frac{k{k}_{\mathit{rj}}}{{\mu }_j}}\left[\nabla p_j-{\rho }_{j}g\nabla Z_j\right]$$

(3)

where k is reservoir permeability, k is the relative permeability, p is the fluid pressure, μ is the dynamic viscosity, g is the gravitational acceleration constant, and Z is the depth. The relative permeability is calculated using the Corey relative permeability model, which can be expressed as follows:

$$k_{\mathit{rw}}=k_{\mathit{rw},\max }{\left({\displaystyle \frac{S_w-S_{\mathit{wc}}}{1-S_{\mathit{or}}-S_{\mathit{wc}}-S_{\mathit{gc}}}}\right)}^{n_w}$$

(4)

$$k_{\mathit{ro}}=k_{\mathit{ro},\max }{\left({\displaystyle \frac{S_o-S_{\mathit{or}}}{1-S_{\mathit{or}}-S_{\mathit{wc}}-S_{\mathit{gc}}}}\right)}^{n_o}$$

(5)

$$k_{\mathit{rg}}=k_{\mathit{rg},\max }{\left({\displaystyle \frac{S_g-S_{\mathit{gc}}}{1-S_{\mathit{or}}-S_{\mathit{wc}}-S_{\mathit{gc}}}}\right)}^{n_g}$$

(6)

Because real-gas and real-liquid behavior deviates from ideality, phase equilibrium was formulated in terms of fugacity rather than pressure. Under thermodynamic equilibrium, the fugacity of component i in the gas phase must be equal to its fugacity in the liquid phase [34], which gives

f_j,l(p, T, x_j) = f_j,g(p, T, x_j)

(7)

Component properties at different temperatures and pressures [35,36] were evaluated with the PR equation of state (PR-EOS), a formulation widely used because it provides a practical balance between accuracy and computational cost. The PR-EOS is written as

$$p={\displaystyle \frac{\mathit{RT}}{V_m-b}}-{\displaystyle \frac{\mathit{a\alpha }}{V_m^2-2b{V}_m-b^2}}$$

(8)

where V_m is the molar volume, R is the gas constant, p is pressure, and T is temperature. Other parameters in the equation can be determined by Equations (9)–(12):

$$a={\displaystyle \frac{0.45724R^2{T}_c^2}{P_c}}$$

(9)

$$b={\displaystyle \frac{0.07780\mathit{RT}}{P_c}}$$

(10)

$$\alpha ={\left[1+(0.37464+1.54226\omega -0.26992{\omega }^2)(1-T_r^{0.5})\right]}^2$$

(11)

$$T_r={\displaystyle \frac{T}{T_c}}$$

(12)

where T_r is the reduced temperature, T_c is the absolute critical temperature, P_c is the absolute critical pressure, and ω is the Pitzer’s acentric factor.

2.2. Research Program

The study area is the Baiju aquifer in Xinghua City, Jiangsu Province, China (Figure 1a). Gas storage is mainly targeted in the E1t0 and K2c sandstone formations, whereas the overlying E1t1 mudstone serves as an effective seal. According to preliminary results from China National Petroleum Corporation, the maximum tolerable formation pressure is 15.0 MPa and the effective storage space is 3864 × 10⁴ m³. The E1t0 layer is 15 m thick with a permeability of 38 mD and a porosity of 14.9%, while the K2c layer is 300 m thick with a permeability of 53 mD and a porosity of 15.3%. The reservoir temperature is 40 °C, and the rock compressibility is 0.58 MPa⁻¹. The mesh size for the E1t0 layer is 40 m × 10 m × 5 m, and the mesh size for the K2c layer is 40 m × 10 m × 15 m. The relative permeability hysteresis was modeled using the Land model implemented in GEM, with a maximum trapped residual gas saturation of 0.3. A geological model reflecting these conditions was constructed (Figure 1b), and the corresponding well configuration is shown in Figure 1c. All eight wells are vertical; four of them are brine-discharge wells introduced to actively regulate formation pressure and accelerate storage construction. The relative permeability curves used in the model are presented in Figure 1d. To protect the vertical sealing capacity of the reservoir, the maximum bottomhole pressure of the injection wells was limited to 15 MPa [37]. The original formation pressure is calculated based on the reservoir depth to determine the hydrostatic pressure, with a hydrostatic pressure gradient of 0.102 MPa/m. The original formation water saturation is 1. Both the vertical and horizontal boundary conditions in the model are constant-pressure boundary conditions.

Figure 1. Geological information of the Baiju aquifer. (a) Stratigraphic profile of the Baiju aquifer. (b) 3D view of reservoir depth. (c) Top view of reservoir depth. (d) Relative permeability curves.

A series of numerical cases was designed to examine gas-storage behavior under coordinated brine extraction and pressure control. The sensitivity factors considered were cushion-gas injection period, bottomhole pressure of brine-discharge wells, daily working-gas discharge rate, vertical-to-horizontal permeability ratio, and cushion-gas type. In the base case (Case 2), when the cushion gas was N₂, the artificial gas cap was established over 3 years, and the maximum injection and discharge rates of working gas were both 50 × 10⁴ m³/day. Five annual operating cycles were then simulated, each consisting of 8 months of injection and 4 months of withdrawal, to represent progressive capacity expansion. The minimum bottomhole pressure and the maximum brine-discharge rate of each brine-discharge well was 10 MPa and 3000 m³/day, respectively. Cases 1 and 3 shortened or extended the cushion-gas injection period by 1 year to evaluate the role of cushion-gas volume. Cases 4 and 5 decreased or increased the minimum brine-well bottomhole pressure by 2 MPa. Cases 6 and 7 reduced or increased the working-gas discharge rate by 10 × 10⁴ m³/day. Cases 8 and 9 set the vertical-to-horizontal permeability ratio to 0.15 and 0.95, respectively. Cases 10 and 11 replaced N₂ with CH₄ and CO₂ to examine the effect of cushion-gas composition. Full case settings are listed in Table 1, and the stage-by-stage simulation results are provided in the Supplementary Materials, alongside all gas volumes reported at standard conditions, the working-gas mole fraction in discharged gas, the cumulative brine-discharge volume, and the cumulative gas injection/discharge volumes.

Table 1. Input parameters for the case study.

Case	Artificial Gas-Cap Construction Duration	Bottomhole Pressure of Brine-Discharge Wells	Daily Gas Discharge Rate	Vertical-Horizontal Permeability Ratio	Cushion-Gas Type
1	2	10	50	0.45	N2
2	3	10	50	0.45	N2
3	4	10	50	0.45	N2
4	3	8	50	0.45	N2
5	3	12	50	0.45	N2
6	3	10	40	0.45	N2
7	3	10	60	0.45	N2
8	3	10	50	0.15	N2
9	3	10	50	0.95	N2
10	3	10	50	0.45	CH4
11	3	10	50	0.45	CO2

3. Results and Discussion

3.1. Effect of Artificial Gas-Cap Construction Duration

The influence of cushion-gas volume was evaluated by comparing Cases 1–3, in which the artificial gas cap was established with 2, 3, and 4 years of cushion-gas injection, respectively. Figure 2 summarizes gas injection–discharge rates, brine-discharge behavior, discharged-gas composition, and cumulative gas volumes. During gas-cap construction (Figure 2a), the cushion-gas injection rate in all three cases first increased and then gradually declined. The maximum rate reached 131.9 × 10⁴ m³/day, whereas the rates at the end of Years 1–4 were 87.179 × 10⁴, 71.879 × 10⁴, 68.879 × 10⁴, and 68.29 × 10⁴ m³/day, respectively; by the second year, the rate had fallen by nearly one-half from its peak value. After the gas-cap construction stage, cushion gas was retained in the reservoir and working gas was injected directly to enlarge storage capacity. As a result, the injection rate during the first working-gas injection period was relatively low, but it increased markedly after one injection–discharge cycle and continued to rise with cycle number. Figure 2b shows that reservoir pressure rose steadily during cushion-gas injection, accompanied by increasing brine-discharge rates of 5110, 5930, 6053, and 6075 m³/day at the ends of Years 1–4, after which the brine rate became nearly stable. During cyclic operation, larger cushion-gas volumes reduced the fluctuation in brine-discharge rate, indicating that a larger artificial gas cap helps buffer pressure variation. According to Figure 2c, when the cushion-gas volume remained sufficiently large, its impact on the working-gas fraction in discharged gas was limited. At the end of discharge in the first cycle, the working-gas mole fraction was about 57%, but it increased progressively with repeated cycling and exceeded 97% by the end of the fifth cycle. This implies that early-stage operation may require either lower discharge rates or additional purification capacity. From Figure 2d, cumulative cushion-gas injections for the three cases were 6.678 × 10⁸, 9.229 × 10⁸, and 11.74 × 10⁸ m³, while cumulative working-gas injections were 11.28 × 10⁸, 11.14 × 10⁸, and 11.03 × 10⁸ m³. After five cycles, the difference in cumulative working-gas injection among the cases was only 2.3%, and the discharged volumes of both cushion gas and working gas also differed little. Deng et al. [29] also pointed out that as the number of injection–production cycles increases, the purity of the discharged working gas gradually improves. Across the five cycles, the average cushion-gas fraction in discharged gas was only about 3.2%.

Figure 2. Numerical simulation results of reservoir development and cyclic injection–discharge processes under different gas cushion injection durations. (a) Gas injection and discharge rate. (b) Brine-discharge rate and cumulative brine-discharge volume. (c) Molar fraction of the discharged gas. (d) Cumulative gas injection volume and cumulative discharge volume.

3.2. Effect of Bottomhole Pressure of Brine-Discharge Wells

Reservoir pressure can be actively regulated by modifying the bottomhole pressure of the brine-discharge wells. Cases 4, 2, and 5 were therefore simulated with brine-well bottomhole pressures of 8, 10, and 12 MPa, respectively, and the results are compiled in Figure 3. During artificial gas-cap construction (Figure 3a), the peak gas injection rates were 134.9 × 10⁴, 131.9 × 10⁴, and 131.6 × 10⁴ m³/day for Cases 4, 2, and 5, whereas the corresponding rates at the end of the construction stage were 87.09 × 10⁴, 68.87 × 10⁴, and 49.59 × 10⁴ m³/day. These values correspond to decreases of 35.44%, 48.08%, and 62.31% after three years. In the subsequent capacity-expansion stage, Cases 4 and 2 were able to maintain the maximum injection–discharge rates in the second and fifth cycles, respectively, whereas Case 5 was restricted by the safety-pressure limit and could not sustain the maximum injection rate for long. Thus, lowering the bottomhole pressure of the brine-production wells not only strengthens active pressure management but also enhances the injectivity of both cushion gas and working gas. Figure 3b shows peak brine-discharge rates of 7980, 6053, and 4187 m³/day during gas-cap construction, while the rates at the end of the fifth cycle decreased to 4085, 3003, and 1660 m³/day. Although a lower bottomhole pressure increases brine output and creates more storage space, excessive brine discharge raises treatment costs and may leave pumping capacity underutilized later in operation. Figure 3c further shows that the working-gas fractions at the end of the first discharge cycle were 67.81%, 56.71%, and 41.28% for Cases 4, 2, and 5, increasing to 98.03%, 98.02%, and 95.62% by the end of the fifth cycle. The large early-cycle differences under identical discharge volumes are attributed to the relatively small amount of injected working gas in the initial stage. As shown in Figure 3d, cushion-gas injections were 10.79 × 10⁸, 9.229 × 10⁸, and 7.689 × 10⁸ m³ for Cases 4, 2, and 5, respectively, and the first-cycle working-gas injections were 1.941 × 10⁸, 1.562 × 10⁸, and 1.159 × 10⁸ m³. Case 5 was therefore 67.47% lower than Case 4 and 34.77% lower than Case 2 in the first-cycle working-gas injection volume. Moreover, the fifth-cycle discharged-gas composition and cumulative working-gas injection in Case 5 were close to the fourth-cycle results of Cases 4 and 2, suggesting that, at a fixed withdrawal rate, the cumulative volume of injected working gas is an important factor influencing produced-gas purity under the conditions considered here. Lowering the brine-well bottomhole pressure is therefore effective, but the benefit becomes progressively smaller once the pressure is reduced below the initial reservoir pressure.

Figure 3. Numerical simulation results of reservoir development and cyclic injection–discharge processes under different bottomhole pressures of brine drainage wells. (a) Gas injection and discharge rate. (b) Brine-discharge rate and cumulative brine-discharge volume. (c) Molar fraction of the discharged gas. (d) Cumulative gas injection volume and cumulative discharge volume.

3.3. Effect of Daily Gas Discharge Rate

Because the storage system also serves a peak-shaving function during capacity expansion, Cases 6, 2, and 7 were analyzed with working-gas discharge rates of 40 × 10⁴, 50 × 10⁴, and 60 × 10⁴ m³/day, respectively. Figure 4 presents the corresponding simulation results. As shown in Figure 4a, Cases 2 and 7 could maintain the target discharge rate, whereas in Case 6 the rate declined near the end of the first cycle but reached the target value in subsequent cycles. A higher discharge rate promoted a higher reinjection rate in later cycles, partly compensating for the negative effect on expansion speed. Nevertheless, Figure 4b indicates that increasing discharge rate reduced the brine-discharge rate during cycling, which implies weaker capacity-expansion efficiency (the volume of working gas stored for the same construction time). After five cycles, cumulative brine-discharge volumes in Cases 6, 2, and 7 were 1370 × 10⁴, 1332 × 10⁴, and 1290 × 10⁴ m³, respectively. Thus, every 1 × 10⁴ m³/day increment in discharge rate lowered cumulative brine discharge by about 4 × 10⁴ m³. Figure 4c shows pronounced differences in working-gas fraction during the first discharge cycle: the end-of-cycle values were 69.46%, 56.71%, and 47.0% for Cases 6, 2, and 7, respectively. These differences diminished gradually with continued cycling. From Figure 4d, the stored natural-gas volumes after five cycles were 6.416 × 10⁸, 5.481 × 10⁸, and 4.552 × 10⁸ m³, indicating that the capacity-expansion efficiency decreased by about 2% for every 1 × 10⁴ m³/day increase in discharge rate.

Figure 4. Numerical simulation results of reservoir construction and cyclic injection–discharge processes at different maximum daily gas discharge rates. (a) Gas injection and discharge rate. (b) Brine-discharge rate and cumulative brine-discharge volume. (c) Molar fraction of the discharged gas. (d) Cumulative gas injection volume and cumulative discharge volume.

3.4. Effect of Vertical-Horizontal Permeability Ratio

To investigate the role of permeability anisotropy, Cases 8, 2, and 9 were assigned vertical-to-horizontal permeability ratios of 0.15, 0.45, and 0.95, respectively. The resulting injection–discharge behavior is shown in Figure 5. During artificial gas-cap construction (Figure 5a), the peak cushion-gas injection rates were 116.6 × 10⁴, 131.9 × 10⁴, and 141.3 × 10⁴ m³/day, and the rates at the end of the construction stage were 63.48 × 10⁴, 68.87 × 10⁴, and 72.14 × 10⁴ m³/day for Cases 8, 2, and 9, respectively. These results demonstrate that lower vertical permeability hampers gas injection. The same effect persisted during cyclic operation, where lower vertical permeability caused a stronger reduction in working-gas injection rate under pressure constraints, suggesting that pressure accumulation is more likely in highly anisotropic reservoirs. Figure 5b shows peak brine-discharge rates of 5499, 6053, and 6394 m³/day during gas-cap construction, and values of 2922, 3003, and 3061 m³/day at the end of the fifth cycle. Because the gas injection–discharge wells were placed near the structural crest and the brine-discharge wells near the structural flank, a low vertical permeability made pressure transmission from the injection wells to the brine wells more difficult, thereby lowering brine output. By contrast, the discharged-gas composition was only weakly affected by permeability anisotropy (Figure 5c). The cumulative cushion-gas injections in Cases 8, 2, and 9 were 8.504 × 10⁸, 9.229 × 10⁸, and 9.649 × 10⁸ m³, and the corresponding cumulative working-gas injections were 10.58 × 10⁸, 11.14 × 10⁸, and 11.36 × 10⁸ m³ (Figure 5d). Overall, vertical permeability exerts a strong control on gas injection but only a limited influence on gas withdrawal.

Figure 5. Numerical simulation results of reservoir construction and cyclic injection–discharge processes under different vertical-horizontal permeability ratios. (a) Gas injection and discharge rate. (b) Brine-discharge rate and cumulative brine-discharge volume. (c) Molar fraction of the discharged gas. (d) Cumulative gas injection volume and cumulative discharge volume.

3.5. Effect of Cushion-Gas Type

The effect of cushion-gas composition was assessed by comparing Cases 10, 2, and 11, which used CH₄, N₂, and CO₂, respectively. Figure 6 summarizes the results. During artificial gas-cap construction (Figure 6a), CO₂ exhibited the highest injection rate, followed by CH₄. During cyclic capacity expansion, however, the increase in injection rate was slower when CO₂ served as the cushion gas than when CH₄ or N₂ was used. Figure 6b shows that CH₄ discharged the largest cumulative brine-discharge volume during gas-cap construction, implying the largest artificial gas-cap volume; N₂ ranked second. In contrast, CO₂ led to much smaller fluctuations in brine-discharge rate, indicating a stronger ability to suppress pressure disturbance during cycling. The discharged-gas composition also depended strongly on cushion-gas type. At the end of the first discharge cycle, the working-gas fractions in Cases 2 and 11 were 56.71% and 50.61%, respectively, and by the end of the fifth cycle they reached 98.03% and 90.29%. Thus, N₂ preserved product-gas purity far better than CO₂. A comparison of Figures S2 and S11 suggests that the large density contrast between CO₂ and CH₄ promotes obvious gravitational layering, whereas CH₄ injected into an N₂ cushion gas shows a funnel-shaped distribution. Saeed and Jadhawar [28] also reported this phenomenon. Under rapid working-gas withdrawal, cushion gas in a layered distribution is more readily discharged. Figure 6d also indicates that CO₂ improves working-gas injection efficiency but compromises discharge efficiency. Considering that N₂ yields working-gas injection and discharge volumes similar to those of CH₄ while also being less expensive, N₂ appears to be the most practical cushion-gas choice.

Figure 6. Numerical simulation results of reservoir construction and cyclic injection–discharge processes under different gas cushion types. (a) Gas injection and discharge rate. (b) Brine-discharge rate and cumulative brine-discharge volume. (c) Molar fraction of the discharged gas. (d) Cumulative gas injection volume and cumulative discharge volume.

4. Discussion and Comparison

The effectiveness of active pressure management observed in this study can be compared with analogous strategies developed for geologic CO₂ storage. Buscheck et al. [38] evaluated pre-injection brine production using a model calibrated against field data from the Snøhvit CO₂ storage project and reported a volume-per-volume effectiveness of approximately 94%; i.e., each cubic meter of produced brine permitted an additional 0.94 m³ of CO₂ to be injected under the same peak-pressure constraint. Our finding that reducing the brine-well bottomhole pressure below the initial reservoir pressure yields diminishing returns is also consistent with the observation of Birkholzer et al. [31] that the marginal benefit of pressure relief decreases once the extraction-induced drawdown extends beyond the pressurized region around the injectors.

The trade-offs among CH₄, N₂, and CO₂ cushion gases identified in Section 3.5 are consistent with recent findings in the closely related field of underground hydrogen storage (UHS) in aquifers. Using a compositional CMG-GEM model of a North Sea aquifer, Saeed and Jadhawar [28] found that CO₂ provided the highest storage capacity owing to its high density and compressibility, whereas N₂ and CH₄ yielded higher recovery efficiency and produced-gas purity. Zhao et al. [39] further showed with an experimentally benchmarked equation-of-state simulator that the low aqueous solubility of CH₄ and N₂ keeps the working gas confined near the wellbore, whereas dissolved CO₂ increases brine density and triggers density-driven convection, enhancing mixing losses. Our results for natural-gas storage in the Baiju aquifer exhibit the same qualitative hierarchy: the CO₂ cushion reduced the peak injection pressure relative to the N₂ cushion but lowered the produced-gas purity in the fifth cycle from 50.61% to 90.29%. A quantitative distinction of the present system is that the density and viscosity contrasts between CH₄ working gas and N₂/CO₂ cushion gases are considerably smaller than those between H₂ and the same cushion gases, so gravity segregation is weaker and the mixing zone develops more slowly than in UHS scenarios. This explains why the purity penalty of the CO₂ cushion observed here is milder than the values reported for hydrogen storage [30,39].

The adverse effect of the gas withdrawal rate on both capacity-expansion efficiency and produced-gas purity found in Section 3.3 and Section 3.4 agrees with sensitivity studies of operational parameters in porous-media storage. Cui et al. [40] reported that increasing the hydrogen production rate from 8.18 × 10⁴ to 12.86 × 10⁴ m³/day significantly degraded the purity of the recovered gas because higher flow velocities intensify convective mixing at the working-gas/cushion-gas interface, whereas the injection rate had only a minor effect. Similarly, our simulations show a loss of approximately 2% in capacity-expansion efficiency per 1 × 10⁴ m³/day increase in withdrawal rate, together with a purity reduction. The consistency of this trend across natural-gas and hydrogen storage systems suggests that rate-dependent interfacial mixing is a general constraint on peak-shaving flexibility in aquifer storage, and that the withdrawal rate should be treated as a design variable jointly optimized with the brine-production schedule rather than fixed solely by surface demand.

5. Conclusions

Using the Baiju aquifer as a case study, this work numerically investigated artificial gas-cap establishment and seasonal multi-cycle operation of aquifer gas storage under active pressure management. The analysis focused on cushion-gas construction time, brine-well bottomhole pressure, gas-discharge rate, vertical-to-horizontal permeability ratio, and cushion-gas type. The main findings are summarized as follows.

(1) Active pressure management improves gas injectivity during both gas-cap construction and subsequent cycling. Increasing the cushion-gas volume dampens pressure oscillations during injection–discharge operation. When a sufficiently large cushion-gas inventory is maintained, its adverse effect on the working-gas fraction in discharged gas remains limited. Under the geological and operational conditions of the Baiju aquifer considered in this study. The working-gas mole fraction in discharged gas rises progressively with cycling and exceeds 97% by the end of the fifth cycle. In the early stage, however, reduced discharge or auxiliary purification may still be required to secure product-gas quality.

(2) Reservoir pressure can be effectively regulated by controlling the bottomhole pressure of the brine-discharge wells. Lowering this pressure slows the decline in gas injection rate, but the incremental gain becomes small once the target pressure is below the initial reservoir pressure. In practical applications, the treatment of discharged brine, pump selection, and overall economics must also be considered.

(3) The storage system can provide peak-shaving supply while its capacity is still being expanded, but higher daily discharge rates reduce the expansion rate. Across the tested cases, higher daily discharge rates were associated with lower capacity-expansion efficiency. But stronger withdrawal promoted higher reinjection rates in later cycles, partly offsetting this adverse effect.

(4) Lower vertical permeability makes gas injection more difficult and causes a stronger reduction in working-gas injection rate during cyclic operation under pressure constraints. The associated brine-discharge rate also decreases, indicating that pressure buildup is more likely in more anisotropic reservoirs. By contrast, the influence of vertical permeability on discharged-gas composition is limited, so its impact is pronounced for injection but comparatively weak for production.

(5) During artificial gas-cap construction, CH₄ generates the largest gas-cap volume, followed by N₂. During cyclic operation, CO₂ suppresses pressure disturbance most effectively, but it also leads to slower growth in injection rate and lower purity of the discharged working gas. N₂ provides working-gas injection and discharge volumes close to those obtained with CH₄, while offering better economy and accessibility. It is therefore the most suitable cushion gas among the three candidates considered here.

(6) In the numerical model used in this study, the dissolution of CH₄, N₂, and CO₂ in the aqueous phase is neglected. When CO₂ is used as the cushion gas, the estimation of the artificial gas-cap volume may result in slight deviations. Furthermore, the model neglects the effect of formation water salinity, which may also potentially influence the solubility of CO₂.