Journal of Energy Storage 73 (2023) 108886

Contents lists available at ScienceDirect

Journal of Energy Storage

journal homepage: www.elsevier.com/locate/est

Review Article

Challenges in the simulation of underground hydrogen storage: A review of

relative permeability and hysteresis in hydrogen-water system
Mehdi Bahrami a, Esmaeil Izadi Amiri a, Davood Zivar b, Shahab Ayatollahi a, *, Hassan Mahani a, *
a
Department of Chemical and Petroleum Engineering, Sharif University of Technology, Tehran, Iran
b
PanTerra Geoconsultants B.V., Leiderdorp, the Netherlands

A R T I C L E I N F O A B S T R A C T

Keywords: Underground hydrogen storage (UHS) and production in depleted gas reservoirs, aquifers, and salt caverns is a
Underground hydrogen storage (UHS) promising solution to balance supply and demand on a large scale. Because of the cyclical nature of the hydrogen
Hysteresis storage and production process and the presence of cushion gas, compared to other underground gas storage
Hydrogen-water relative permeability
operations like carbon capture and storage (CCS), accurate application of hydrogen-water dynamic properties in
Gas mixing
the UHS simulation process is critical. In this paper, we present a state-of-the-art literature review on relative
Wettability
Special core analysis permeability of the hydrogen-water system, and the determination methods. To further highlight the existing
data deficiency and simulation challenges, we performed reservoir-scale flow simulations using a conceptual
geological model to study the impact of hysteresis and reservoir mixing between hydrogen and cushion gas on
flow functions, and the efficiency of the operation. The results reveal that hysteresis and gas mixing significantly
affect the total volume of hydrogen produced, the water-gas ratio, the purity of the produced gas, and the
reservoir pressure. On average, the ultimate recovery factor in the presence of hysteresis decreased by 16 % to
25 %, while increasing water production from the field. The simulation results for the effect of in-situ gas mixing
show a decrease in the hydrogen recovery factor by about 5 %. This review emphasizes the importance of
considering hysteresis and gas mixing in UHS simulation studies. To improve the accuracy and reliability of the
simulation results of UHS process for field deployment, it is necessary to conduct further research and acquire
field-specific saturation functions for pure and mixed gases.

1. Introduction storage, existing methods are constrained in their capacity and

discharge duration. This underscores the potential for hydrogen-based
Energy generation from renewable sources, rather than fossil fuel- energy storage to be executed on a far greater scale than most conven­
based sources, is gaining attraction around the world to meet interna­ tional storage approaches. Consequently, the distribution and utilization
tional climate protection agreements and net zero‑carbon initiatives. of significant quantities of energy (ranging from GWh to TWh) at the
This is because it generates only negligible amounts of greenhouse gases community level necessitates storage facilities that surpass the limita­
(GHG) in comparison with fossil fuel-based energy sources and hence tions of the surface-level storage.
supports the environmental protection initiatives to tackle the issues Hydrogen is a suitable storage source for this purpose because it has a
caused by the use of fossil fuels [1,2]. specific energy capacity of 132 kWh/m3, and that its combustion re­
One of the aspects affecting the worldwide development of renew­ leases no greenhouse gases, and its burning product is water, and has a
able energies is the seasonal dependence on energy sources such as wind transformational nature. However, hydrogen has a low density of 0.084
and solar energy, such that, depending on the month of the year, the kg/m3 (at 20 ◦ C and 0.1 MPa), which means that to supply energy on a
amount of energy accessible varies from less than the requirement to terawatt hour (TWh) scale, large-scale storage is required [4–6]. Un­
more than the demand. Energy storage plays as a buffer to handle these derground hydrogen storage (UHS) and production is one of the most
changes and is one method for dealing with this matter. Fig. 1 shows suitable options for large-scale and long-term energy storage, and its
storage ranges and discharge durations provided by various energy usage as a buffer among the various energy storage techniques can be
storage options for renewable energies [3]. In the realm of electricity effectively operated in both offshore and onshore reservoirs [7–9].

* Corresponding author.
E-mail addresses: [email protected] (S. Ayatollahi), [email protected] (H. Mahani).

https://doi.org/10.1016/j.est.2023.108886
Received 13 June 2023; Received in revised form 28 August 2023; Accepted 30 August 2023
Available online 14 September 2023
2352-152X/© 2023 Elsevier Ltd. All rights reserved.
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

In addition to UHS, the natural hydrogen (or white hydrogen) re­ through the implementation of two hysteresis models and subsequent
sources are the potential resources that can produce a large amount of examination of simulation results. Moreover, we explore the potential
hydrogen depending on the size of the reservoir. Digging deep into the impact of hydrogen mixing with the initial reservoir gas or cushion gas
history of explorations shows several potential sites were identified in on the simulation outcomes. Additionally, we discuss the challenges
the past with considerable amounts of hydrogen concentration but were associated with accurately measuring the relative permeability of the
left behind due to the popularity of hydrocarbon resources. For example, hydrogen-water system in laboratory settings and provide tips for
it was reported that the gas sample taken from Robe 1 contained 25 % enhancing measurement accuracy. These laboratory measurements
hydrogen. This well was drilled in 1915 in the Otway Basin, Australia. serve as critical inputs for field-scale simulations of the UHS process.
Later, 60 % and 75 % were reported for the concentration of hydrogen This paper is organized as follows: Section 2 provides an overview of
for the American Beach Bore 1 and Ramsay Oil Bore 1 samples in the studies conducted to determine the relative permeability of the
Australia, respectively. In 1987, the drilling of a water borehole ended hydrogen-water system. Additionally, a comprehensive review of
up with an explosion in Mali. A high concentration of hydrogen (up to simulation studies on UHS and their corresponding relative permeabil­
98 %) was detected for this well. Although natural hydrogen could not ities has been conducted. Section 3 details the assumptions and model
draw attention in the past, this type of energy is gaining attention employed to simulate a UHS process, as well as the relative permeability
rapidly. It is claimed that the drilled well in Salina Basin in Fillmore and hysteresis models utilized. Furthermore, it outlines the methodol­
County, Nebraska in 2019 contains a considerable amount of hydrogen ogy for investigating the impact of gas mixing on relative permeability,
(aka NH2E). In addition, several companies awarded licenses to explore with the aim of determining how hysteresis and gas mixing influence the
natural hydrogen, like H2EX, SCIRO, Buru Energy, and Byrock simulation results. In Section 4, the discussion and results are presented.
Resources. We first analyze the impact of hysteresis and gas mixing on the perfor­
The world's top economies regard hydrogen as the fuel of the future mance of the UHS process. Then, the challenges associated with the
due to its high energy content and zero emissions. However, it has not experimental measurements that require attention and further research
yet met the threshold for safe and efficient production, storage, and are investigated.
consumption [10]. Several studies on various aspects of UHS have
confirmed the feasibility of this process. However, this process does not 2. Hydrogen-water relative permeability
appear to be very complex conceptually, though there are various
characteristics specific to hydrogen that needs to be considered. Firstly, There are a few experimental analyses reported in the literature to
compared to other fluids stored in geological structures, such as CO2 and accurately characterize the dynamic properties of the hydrogen-water
CH4, the physico-chemical properties of hydrogen are different. Sec­ system, such as relative permeability and capillary pressure. Overall,
ondly, the (abiotic) reaction of hydrogen with minerals and subsurface there exist very few studies on the relative permeability of the hydrogen-
fluids could have an impact on storage operations. Thirdly, the presence water system, and most of the research works have either used the
of hydrogen in the subterranean formation can favor the growth of available relative permeability data from the CH4-water system or the
microbes that consume hydrogen. Fourthly, the integrity of the reservoir empirical models like the Brooks-Corey model. In Table 1, we present
and caprock may be put at risk as a result of the stress field at hydrogen the usual models used in these studies.
storage sites changing during injection-production cycles [11]. In one of the very first studies by Yekta et al. [16], the drainage
Because of the absence of reliable data as necessary input parameters relative permeability curves under a capillary-dominated regime for
for reservoir simulation and process design of hydrogen storage and sandstone samples were obtained by using the steady-state method for
production, UHS in geological structures remains a challenge. One of the hydrogen-water system for two sets of temperatures and pressures.
these input parameters is the dynamic properties of hydrogen flow in During the flooding, 40 % water saturation was achieved. By utilizing
porous media, especially relative permeability. In this work, we high­ capillary pressure measurement and Darcy's law together, the hydrogen
light the crucial role of hysteresis in the UHS process, which is addressed relative permeability was computed at water saturations <40 %. These

Fig. 1. Storage capacity of different energy sources (adopted from [3]).

2
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Table 1 measurements [18].

Summary of the drainage relative permeability models used in UHS simulation. It is noteworthy that most numerical studies on the subject of UHS
Relative permeability Equations Reference have used empirical models of relative permeability curves without
model including hysteresis and the gas mixing effect. The mixing of the primary
Brooks-Corey 2 + 3λ [12] reservoir fluid with hydrogen gas, either when a depleted gas reservoir is
( )
krw = S*w λ used for UHS or when a cushion gas is used, has a key role in UHS
( 2+λ) feasibility studies [19]. Such mixing can affect the relative permeability
2 ( )
krg = S*g 1− S*w λ of the gas mixture and subsequently reduce the ultimate recovery of the
van Genuchten-Mualem √̅̅̅̅̅̅{ ( 1 }
[ ] )λ 2
[13,14] storage process and hydrogen purity as we will demonstrate later in this
krw = S*w 1 − 1 − S*w λ if Sw < paper. Hysteresis relates to the irreversibility of a process and its
Sws dependence on history and path. In two-phase systems, the wetting
krw = 1 if Sw ≥ Sws phase remains continuous through smaller pores and wetting film. As
krg = 1 − krw if Sgr = 0
imbibition progresses and the saturation of the wetting phase increases,
1
√̅̅̅̅̅[(
krg = S*g 1 − S*w λ
[ ] ) ]2λ
if Sgr > 0
some of the non-wetting phase becomes trapped in the pores. When the
( S ) ]m wetting phase reaches a certain saturation level, smaller pores that used
Corey type [
g [15]
krw = 1 −
1 − Swr
to have more of it flowing through them now have some non-wetting
[ S ]n
g
fluid trapped inside. As a result, the wetting fluid needs to flow
krg = korg
1 − Swr through larger pores, leading to an increase in the wetting phase relative
All models Sw − Swr –
permeability during imbibition compared to drainage (gas injection).
S*w =
1 − Swr
Sg
Furthermore, the non-wetting phase relative permeability becomes
S*g = lower in imbibition than in drainage [20]. In contrast to some other
1 − Swr
underground gas injection processes, like carbon capture and storage
(CCS), in which gas is only stored, hydrogen storage is a cyclic process of
findings suggest that the relative permeability of the hydrogen-water injection and production. The cyclic nature of underground hydrogen
system is not significantly affected by changing temperature and pres­ storage processes makes the concept of hysteresis relevant. For example,
sure in the range of potential hydrogen storage conditions because the hydrogen-water relative permeability in Berea sandstone as
hydrogen viscosity does not vary significantly [16]. Also, the unsteady- measured by Lysyy et al. [21] shows the significant effect of hysteresis
state method was used in [17] for the first time to obtain the relative for both water and hydrogen relative permeability curves. The PNM
permeability curve for the hydrogen-water system over a variety of results by Hashemi et al. [5] show that the secondary drainage relative
temperatures, pressures, water salinities, and rock types. The study permeability values were lower than those in the primary drainage due
compared hydrogen's relative permeability to CH4 and N2. The results to the trapped water in the porous medium and hydrogen phase
showed that pressure increases hydrogen viscosity, which reduces gas disconnection after secondary imbibition. However, the hysteresis effect
mobility and hydrogen relative permeability at higher gas saturations. remained the same as more cyclic processes occurred. In Table 2, a more
However, at lower gas saturations, higher pressures show higher comprehensive overview of the studies conducted on hysteresis in the
hydrogen relative permeability due to capillary forces overcoming hydrogen-water system is presented.
viscous forces. Water salinity increases the interfacial tension of the H2- It should be noted that relative permeability and hysteresis in the
water system, making it more difficult for H2 to transport through hydrogen-water system are different from the systems of other gases and
higher-salinity pore fluid systems. However, in general, salinity had oil-water-gas systems. Based on experimental measurements, one
almost no effect on the hydrogen relative permeability curve. The study notable difference is the level of hysteresis observed in the wetting
also evaluated relative permeability in different rock types and found phase. Typically, wetting phases in other systems exhibit minor hys­
that porosity is the most effective parameter on relative permeability teresis, but this is not the case in the hydrogen system. The liquid phase
curves, however, these curves were unrelated to pore size distribution relative permeability hysteresis for the hydrogen-brine system is (much)
and mineralogy. Under the same conditions, relative permeability stronger. Additionally, experimental data reveals that the relative
curves for CH4, N2, and H2 were obtained. The study found that CH4 has permeability of hydrogen during imbibition is extremely low compared
higher relative permeability at each saturation, but H2 and N2 have to the carbon dioxide-brine and oil-water-gas systems. This further
higher relative permeability at the endpoint as a result of hydrogen's highlights the unique behavior of the hydrogen-water system and em­
lower viscosity. Pore network modeling (PNM) has been used to phasizes the need for a thorough understanding of it [24]. For instance,
numerically generate the relative permeability of hydrogen-water sys­ Fig. 2 compares the experimentally measured relative permeability
tems [5]. The PNM results were modelled using Corey functions and the values for both drainage and imbibition processes in hydrogen-water
Corey's exponents for hydrogen and water were obtained 1.31 and 4.36, and CO2-water systems.
respectively. These values are reasonable as Corey's exponents are
related to pore geometry and wettability, and are greater for the wetting 3. Technical review
phase, i.e., water. In a subsequent study, an experimental coreflood test
was conducted on a Berea sandstone core to measure the relative In Table 3, a summary of the UHS simulation studies reported in the
permeability and capillary pressure of hydrogen and water during literature is presented. Field-scale simulations complement laboratory
drainage and imbibition processes. The saturation distribution was experiments, however the diverse input parameters across studies in­
visualized using a medical X-ray CT scanner. The findings of the study creases uncertainty in the results. Standardization of relative perme­
revealed that the interaction between viscous, capillary, and gravita­ ability inputs is essential for more reliable storage facility design and
tional forces can result in complicated displacement patterns during operation.
UHS. Moreover, the study identified that the formation of high-water While the reported studies provide valuable insights, there is still a
saturation fingers can establish preferential flow pathways for the significant knowledge gap in the understanding of the impact of hys­
water phase, potentially causing significant entrapment of hydrogen teresis and gas mixing on UHS performance. As such, we have chosen to
during reproduction. These findings emphasize the necessity of visual­ further explore this area in our technical review, to shed light on the
izing the saturation distribution map to ensure the accuracy of the ob­ effects of these two crucial phenomena.
tained multiphase flow parameters for the hydrogen-water system, as
gravity override (segregation) and channel flow can influence the

3
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Table 2 with hydrogen gas within the PVT model. Additionally, it can accurately
Summary of the conducted studies on hysteresis relevant to hydrogen-water simulate multiphase flow, hysteresis, and fluid behavior. Moreover, the
system. software has the ability to incorporate features such as aquifer, rock
Reference Method Main findings compaction, geomechanical effects, and geochemistry. These advanced
Hashemi Pore network modeling The findings suggest that the
functionalities align perfectly with our future research objectives. Fig. 3
et al. [5] relative permeability values for represents the geometry of the simulation model, which contains 43, 43,
secondary drainage are lower and 5 grid blocks in i, j and k directions, respectively. The details of the
than those for primary drainage simulation inputs are presented in Table 4.
when saturation levels are
The distribution of permeability data is illustrated in Fig. 4a. The
equivalent. Furthermore, the
hysteresis effect remains the same distribution of permeability data is obtained from Eq. (1) [40] and
throughout drainage and illustrated in Fig. 4b.
imbibition cycles after the initial
(1)
logK+7.9
injection/withdrawal process. φ = 10 7.1

The authors employed the

modified Morrow's model [22,23] where φ is porosity and K is the permeability.
to determine advancing and
The PVT model was established based on the Peng-Robinson equa­
receding contact angles in order to
obtain secondary imbibition and tion of state with a focus on tuning of acentric factor and volume shift for
secondary drainage relative hydrogen. The temperature was considered constant at 105◦ F. The sol­
permeability curves. ubility of hydrogen in water, and the diffusion of hydrogen (either in
Boon and Core scale experiment The observations indicate that
water or in cushion gas) were not considered in the model. The solubility
Hajibeygi hysteresis participates in drainage
[18] and imbibition stages for both of hydrogen in water/brine is very low, to the extent that hydrogen is
phases (hydrogen and water), often considered insoluble. The solubility of hydrogen in water.
where it is less pronounced for the expressed as mole fraction, is only 0.00002 at 25 ◦ C and 10 bar and is
water phase. 0.00018 at 100 bar [41]. In contrast, cushion gases such as CO2 may
Lysyy et al. Core scale experiment The relative permeabilities for
exhibit much higher solubility in water, and hydrogen recovery can also
[21] both hydrogen and water
exhibited a significant hysteresis be affected by increasing their solubility.
effect. The primary drainage Krg UHS in an aquifer was considered and the drainage relative perme­
values were higher than those ability and capillary pressure curves were taken from [32] and are
observed during imbibition, while
presented in Fig. 5.
the Krg values during secondary
drainage were intermediate
between the two. 3.2. Injection/withdrawal scenarios
Bo et al. [24] Numerical simulation with In the study, it was found that Kr
the help of experimental hysteresis of both phases
relative permeability data (hydrogen and water) plays a A single well was used for both the injection/withdrawal stages. Due
crucial role in UHS processes. to the low density of hydrogen gas and its high-density difference with
Neglecting gas hysteresis can the formation water, hydrogen in the reservoir tends to accumulate in
introduce very large uncertainties the upper parts of the reservoir, therefore the well was placed at the top
in cumulative hydrogen
production. This is primarily
of the reservoir, and the first two layers from the top were perforated,
attributed to its adverse effect on resulting in higher hydrogen production and lower water production.
injectivity and productivity, There are five storage cycles with injection, idle and withdrawal dura­
caused by limited variations in tion of 4, 1, and 7 months, respectively, which remain the same for all
hydrogen saturation and pressure
scenarios. The total simulation time is 6 years. In the first year, nitrogen
in different cycles. However,
hysteresis in the water phase (as a cushion gas) is injected at a surface rate of 500 ft3 /day until the
relative permeability can partially bottom hole pressure (BHP) reaches the initial pressure of the reservoir
enhance the displacement of the (650 psi) to 850 psi. N2 was chosen as cushion gas because of its
water phase and contribute to advantage over hydrogen to displace the resident water due to a more
resolving this issue.
Pan et al. [25] Numerical simulation with The phenomenon of Kr hysteresis
favorable mobility ratio and subsequently higher water relative
the help of experimental leads to a reduction in the H2 permeability values [21]. Hydrogen injection and withdrawal start in
relative permeability data withdrawal factor, but an increase the second year and continue for the next five years. The hydrogen is
in the H2 withdrawal purity.
injected at a surface rate of 2000 ft3 /day for 4 months per year with the
Ershadnia Numerical simulation Neglecting the impact of Kr
et al. [26] hysteresis can lead to significant
maximum allowable injection pressure of 1150 psi. There is one-month
underestimation of residual idle time between the injection and withdrawal stages. The reservoir
trapping, and overestimation of then produces at a rate of 1300 ft3 /day for seven months, in which the
the amounts of produced H2,
constraint for abandonment is 100 psi.
mobile phase, and extracted
water, while the impact on
solubility trapping is relatively 3.3. Hysteresis in relative permeability
minimal.

The linear [42] and Carlson [43] models were used to apply the
3.1. UHS simulation model hysteresis phenomenon to the simulation model. The primary rationale
behind this choice was the alignment of these models with the CMG
We considered a synthetic, dome-like, three-dimensional heteroge­ software package, which is specifically tailored for gas behavior anal­
neous model to study the hypothesis. A hypothetical dome model was ysis. In the linear model, for a specific value of the gas saturation, the gas
used because gas would spread less in the reservoir and accumulate in relative permeability on the drainage to imbibition scanning curve is
one area, minimizing the effect of other factors and phenomena affecting calculated by:
the results. Among different existing simulators (see Table 3), CMG-GEM ( ) ( )
Krg Sg = Krg drn
Sgf (2)
(2018.10) was used for several reasons. Firstly, its capabilities in dealing

4
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

0.4
Kr(H2) Drainage Hydrogen-Water [21]
Kr(H2) Imbibion Hydrogen-Water [21]
Kr(W) Drainage Hydrogen-Water [21]
Kr(W) Imbibion Hydrogen-Water [21]
0.3 Kr(CO2) Drainage CO2-Water [27]
KrCO2) Imbibion CO2-Water [27]
Kr(W) Drainage CO2-Water [27]
Kr Kr(W) Imbibion CO2-Water [27]

0.2

0.1

0
0.3 0.4 0.5 0.6 0.7 0.8
Sw
Fig. 2. Variations in gas and water relative permeabilities for hydrogen-water [21] and carbon dioxide-water [27] systems during drainage and imbibition processes
in Berea sandstone samples.

where Krg stands for the gas relative permeability, Sg is the saturation of on the core-scale studies, the residual gas saturation (Sgr ) is often be­
tween 0.24 and 0.45. Therefore, we chose a value in this range to adhere
gas, Kdrn
rg is the drainage gas relative permeability and Sgf is the free gas
to the published experimental results. In this regard, the specific value of
saturation. 0.35 was chosen based on Ref. [5]. Moreover, the critical gas saturation
We applied the Land's trapping function [44] as the trapping func­ (Sgcrit) was assumed to be zero due to the high mobility of hydrogen.
tion. The Land's parameter and the trapped gas saturation Sgrh are given However, it may not be zero in all cases, as it depends strongly on
by: permeability, pore structure of the porous medium and saturation,
1 1 which control the percolation threshold and the connectivity of the gas
C= − (3) phase.
Sgrmax − Sgcrit Sgmax − Sgcrit
( )
Sgh − Sgcrit
Sgrh = Sgcrit + ( ) (4) 3.4. Mixed gas relative permeability
1 + C Sgh − Sgcrit

where C is the Land's parameter, Sgrmax is the maximum residual gas Cushion gas is typically used in an underground gas storage (UGS)
saturation and Sgmax is the maximum gas saturation. operation to provide the necessary pressure for withdrawing working
Using the following relationship, the free gas saturation (Sgf ) is gas (hydrogen) and removing initial fluid(s) from the pores to make
calculated by: some space for hydrogen placement. The thermodynamic properties of
( ) ( ) hydrogen gas can vary depending on its composition ratio with cushion
Sg − Sgrh − Sgh − Sgcrit gas as well as the temperature and pressure of the reservoir (for further
Sgf = Sgcrit + ( ) (5)
Sgh − Sgrh details, see ref. [45]). In our work, the properties of the gas mixture,
such as viscosity and density, are handled by the simulator. In-situ
where Sgcrit is the critical gas saturation, Sgrh is the trap gas saturation mixing of cushion gas with hydrogen reates a mixing zone with rock
and Sgh is the maximum gas saturation. & fluid dynamical behaviour different from that of pure hydrogen or
In the next case, the gas phase (as the non-wetting phase) hysteresis cushion gas in the porous medium. In the presence of cushion gas, the
is governed by the Land and Carlson models. relative permeability for mixed gas varies depending on the saturation
The drainage to imbibition scanning curve's gas relative permeability distribution of hydrogen and cushion gas at different distances from the
can be calculated using Eq. (2). Land's equation is used to calculate the production or injection well. There are variable ratios of injection gas
free gas saturation: and cushion gas at different distances from the well, which results in a
{ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ } distribution of relative permeability (as a function of time) in the porous
Sgf = Sgcrit +
1 ( )
Sg − Sgrh +
(
Sg − Sgrh +
)2 4 (
Sg − Sgrh
)
(6) medium. This requires further research to obtain and apply the mixed
2 C gas flow functions in simulations, as there is currently no reported data
for the mixture of H2 and cushion gas. To highlight the effect of
This model focuses on the non-wetting phase and assumes that
considering mixed gas relative permeability, we considered a case with a
hysteresis does not affect the relative permeability of water as a wetting
similar scenario and properties to the base case and used the arithmetic
phase. It is important to note that, based on the simulations in Ref. [24],
average of hydrogen and nitrogen relative permeability instead of pure
the inclusion of relative permeability hysteresis for hydrogen over­
hydrogen. We further assumed that the residual water saturation of the
estimates hydrogen production by 338 %. However, when hysteresis is
mixed gas-water system is similar to that of the pure hydrogen-water
considered for both hydrogen and water fluids, there is a 17 % increase
system. Fig. 6 shows the relative permeability and capillary pressure
in the cumulative hydrogen production. This suggests that considering
curves that were used for pure hydrogen-water and nitrogen-water
hysteresis for the water phase will not greatly change the results. Based
systems [32,46] and the mixed gas-water system.

5
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Table 3
Summary of UHS simulation studies and their main findings.
Description and key points Storage medium Storage Cushion Software Relative permeability models Reference
duration gas

A numerical model is developed for the hydrodynamic effects Aquifer Short term H2 and DuMux Brooks–Corey [28]
of UHS and it explores the injection of hydrogen for storage CH4
and compares it with methane injection. At lower injection
rates, gravitational forces are the primary driving factor,
resulting in a uniform water displacement. However, the
dominance shifts to viscous forces at higher injection rates
lead to an unstable displacement pattern. Also, hydrogen
has been shown to spread laterally faster than methane.
The impacts of different injection and production cycles, gas Depleted gas Long term N2 and DuMux and Brooks–Corey [19]
mixing, as well as the hydrodynamics involved in UHS are reservoir CH4 COMSOL
to be examined. Gravity override and viscous fingering
complicated aquifer fluid displacement, although gas-
saturated reservoirs are unaffected
Present a mathematical model to simulate the combined Depleted gas – CO2 and DuMuX Brooks–Corey [29]
hydrodynamic and bio-chemical processes in UHS. Lateral reservoir CH4
fingers spread faster in UHS than in natural gas storage.
Simulate the subsurface hydrogen storage process and Saline aquifer Long term H2 COMSOL Brooks–Corey [30]
analyzed three different scenarios based on extraction well
configurations. The highest hydrogen recovery was
observed when withdrawal wells were completed slightly
below the reservoir top (caprock) and when wells were not
completed throughout the entire thickness of the reservoir.
Examining hydrogen storage in a hypothetical model based Saline aquifer Short term N2 Eclipse 300 Corey [31]
on a real geological structure to investigate the feasibility
of UHS. The storage facility has the potential to meet
approximately 20 % of the average electrical energy
demand in the state of Schleswig-Holstein, Germany, for a
week.
An existing anticlinal structure was examined to explore a Saline aquifer Short term N2 Eclipse 300 Corey [32]
potential UHS scenario within the North German Basin.
The storage performance exhibited a low initial efficiency,
which gradually improved with an increasing number of
cycles as more hydrogen became accessible.
The potential of hydrogen storage in a sandstone formation Saline aquifer Short term N2 Eclipse 300 Corey [33]
located within an anticline structure in a North German
Basin was explored through numerical simulation.
The feasibility of hydrogen storage in a deep aquifer was Aquifer Long term H2 petraSim- van Genuchten-Mualem [34]
assessed. The maximum gas saturation was observed near TOUGH2
the well and in the cap rock of the aquifer.
The effect of caprock presence and injection rate of hydrogen Aquifer Long term H2 TOUGH2 van Genuchten-Mualem [35]
on both hydrogen withdrawal efficiency and hydrogen
leakage rate was investigated. Without a caprock, higher
injection rates result in increased hydrogen leakage.
Conversely, lower injection rates combined with the
presence of caprock lead to a higher amount of recovered
hydrogen.
To explore the potential of storing hydrogen in the Norne Depleted oil and Long term CH4 ECLIPSE Norne relative permeability [36]
hydrocarbon field located offshore Norway Pure hydrogen gas field 100 functions for gas, oil, and water,
was injected into the different zones, such as oil, water, and as originally published
gas intervals, to investigate three different storage
schemes. The thin gas zone, with an 87 % ultimate
hydrogen recovery factor, is a preferable candidate. Also,
completion of the injector in the lower intervals causes a
decrease in storage efficiency.
The influence of different cushion gas types, including CH4, Depleted gas Long term CH4, N2, CMG- GEM Original relative permeability [37]
N2, and CO2, on the purity and production recovery of the condensate and CO2 functions for gas, oil, and water
UHS process was examined. The utilization of CH4 and CO2 reservoir
as cushion gases yielded the highest and lowest recovery
factors, respectively.
A sensitivity analysis was conducted on operational, Depleted gas Short term CH4 ECLIPSE Experimental data from [38]
geological, and reservoir factors that impact the efficiency reservoir 300 Ref. [16]
of the withdrawal stage. A set of criteria for selecting
suitable sites for storing hydrogen in depleted reservoirs
was established.
CO2 storage, hydrogen storage, and natural gas storage were Depleted oil Long term H2 CMG-GEM Experimental data from [39]
compared in real reservoir models. H2 has a wider lateral reservoir and Ref. [16]
spread than CO2 and natural gas. Also, knowledge of CO2 saline aquifer
and natural gas storage cannot simply be applied to
underground hydrogen storage.
The effects of injection/withdrawal plans, relative Aquifer Long term H2 CMG-IMEX Drainage data from Ref. [16] [25]
permeability hysteresis, and rock were investigated in a Imbibition data from Ref. [5]
PUNQ-S3 geological model. The hydrogen recovery factor
is reduced by hysteresis while the purity of the produced
hydrogen is increased, and lower H2 withdrawal
efficiencies are found in more water-wet rocks

6
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Fig. 3. Geometry of the simulation model (colour bar shows depth) and the location of injection/production well (red arrow). (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

restricted due to the limitation of gas production. This is because the gas
Table 4
that is trapped in the larger pores of the storage reservoir can block the
Properties of the simulation model.
fluid path, preventing fluctuations in the WGR trend.
Parameter Value Fig. 7b shows that the average reservoir pressure responds strongly
Dimensions 430 ft. × 430 ft. × 25 ft. to the WGR changes because the differential pressure in each cycle is
Average absolute permeability (i, j and k) 250, 250 and 25 mD related to the area below the WGR curve. For example, the average
Average porosity 0.25
reservoir pressure reduces more in the scenarios with hysteresis
Reservoir top 1500 ft.
Initial pressure 650 psi
compared to the no-hysteresis scenario for the early cycles due to a
Initial water saturation 1.0 larger area below WGR curves. Therefore, the observed results highlight
Injection/production rates 2000/1300 ft3 /day the importance of reservoir management after several cycles of injec­
Rock compressibility 10− 6 1/psi tion/withdrawal in which re-injection of cushion gas may help, espe­
Well's blocks 16,16,1:2 cially when there is no active aquifer. In addition to the re-injection of
cushion gas, the injection of water from the deeper layers is an alter­
4. Results and discussion native scenario that can maintain the reservoir pressure.
Fig. 8 shows the amount of hydrogen and nitrogen gases in the
4.1. Effect of hysteresis on hydrogen storage and production reservoir over six years, demonstrating that when the hysteresis is
considered, a higher volume of hydrogen is trapped in the reservoir after
Fig. 7 shows the water-gas ratio (WGR) and the average reservoir six years. In the absence of hysteresis, the amount of remaining
pressure over 6 years for three different relative permeability scenarios: hydrogen saturation at the end of each cycle stays constant as the
the no hysteresis case, the Carlson hysteresis model case, and the linear number of cycles increases. This means that we could theoretically
hysteresis model case. As seen in this figure, there is a decreasing trend withdraw all of the hydrogen that we injected. However, this is not
for WGR with cycles in the cases with hysteresis. This is because the gas realistic because, during the imbibition process, a certain amount of gas
tends to remain in the near wellbore area due to its reduced mobility. will be trapped in the porous medium in the form of bubbles and clusters
Therefore, the water replaces with hydrogen gas during the injection/ and is not produced. In the cases with hysteresis applied, the amount of
withdrawal stages and there is less water available to produce over time. residual gas saturation in the reservoir at the end of each cycle increases
In the case of no hysteresis, the WGR in the production stream increases with increasing cycles. In the Carlson model, the residual gas in the
for the first 3 cycles and sharply decreases in cycles 5 and 6. It is believed reservoir is roughly 4.54 times greater, and in the linear case, it is
that hydrogen can spread laterally into the reservoir and allows water to around 3.07 times greater than in the case without hysteresis. Also,
produce because its mobility is not limited by hysteresis [28,29,39]. The when hysteresis is not used, the majority of the injected nitrogen gas as
WGR decreases in the last two cycles due to the simulation assumption cushion gas is produced throughout the production and injection cycles,
(i.e., the no flow reservoir boundary condition) and high-water pro­ until at the end of the fifth cycle, its amount in the reservoir is nearly
duction in the previous cycles. The WGR trend without hysteresis is zero. But, when hysteresis is applied, a significant amount of cushion gas
characterized by significant fluctuations. This is because the gas is able remains trapped in the reservoir. Ignoring the hysteresis effect can lead
to move freely through the reservoir, and the amount of gas that is to an overestimation of hydrogen production, as the remaining
produced depends on the amount of gas that is in the reservoir. How­ hydrogen in the reservoirs after the withdrawal stage becomes relatively
ever, when hysteresis is considered, the movement of water is further minor. Additionally, it may underestimate the purity of the production
stream due to the low residual nitrogen in place present at the end of the

7
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Fig. 4. Distribution of a) permeability and b) porosity in the simulation model.

Fig. 5. Rock-fluid properties of hydrogen-water system, a) drainage relative permeability curves, b) drainage capillary pressure curve (data taken from [32]).

8
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Fig. 6. Rock-fluid properties of hydrogen-water system, a) relative permeability curves, b) capillary pressure curves (data taken from [32,46]).

simulation time. absence of hysteresis, the residual gas saturation in the reservoir in­
In order to examine the effect of hysteresis on gas saturation changes, creases with the number of cycles in the presence of hysteresis, which is
Fig. 9 shows the amount of gas saturation and residual saturation consistent with results of Ref. [47,48].
changes in the well block (16,16,1). According to Fig. 9a, the range of The amount of hydrogen gas recovered in each year is presented in
saturation changes is smaller in the presence of hysteresis since gas Fig. 10. The dome-like shape of the reservoir, as well as the placement of
saturation is a function of the history of prior drainage and imbibition the injection/production wells at the reservoir top, along with high
cycles. As seen in Fig. 9a, gas saturation in block (16,16,1), during the vertical permeability (25 mD), results in gas accumulation at the
first months of the first cycle of injection-production (365 to 730 days), reservoir top, and an overall high recovery factor for the case without
in the case with hysteresis is lower than in the case without hysteresis. hysteresis. The recovery factor in the linear and Carlson models is, on
While during the final months of the year, the gas saturation experiences average, 16 % and 25 % lower than that in the model without hysteresis
a significant decrease in the absence of hysteresis. This is because, in the over the years of hydrogen injection and production. From the fourth
absence of hysteresis, gas production can still occur even at low satu­ cycle onward, due to decreasing reservoir pressure and reaching the
rations. Toward the end of the production period, as water production minimum pressure constraint, production is halted in the final days. As a
usually reaches its maximum due to coning, there is a sudden increase in result, the production duration in each cycle decreases. Since production
the WGR that lasts only a few days. Hysteresis keeps gas saturation from is maintained at the maximum flow rate in the later cycles, we will
falling below a certain value. As a result, gas production declines and encounter a reduction in the volume of produced gas. This phenomenon
eventually ceases sooner, causing the WGR trend to rise earlier. is the reason behind the decline in hydrogen recovery in the later cycles.
Although the application of hysteresis may result in a shorter gas pro­
duction period, it can also result in a longer period of water production 4.2. Effect of gas mixing on hydrogen storage and production
before reaching its peak. This may prevent observing a sharp peak, but
generally, when hysteresis is applied, the cumulative water production Fig. 11 shows the changes in the average reservoir pressure under
is greater than in the absence of hysteresis. Thus, the reservoir pressure two different scenarios. When gas mixing is considered and the corre­
decreases less in the first year in the case without hysteresis because the sponding change in the relative permeability is factored in, the average
average WGR is lower. In the third year of production, the case without reservoir pressure becomes higher than that in the other case. This is due
hysteresis has a large WGR peak in the last few days, resulting in a to the fact that the gas relative permeability in the gas mixing scenario is
notable pressure drop in cycles 5 and 6. less than that of pure hydrogen, which hinders the gas from spreading
Also, according to Fig. 9b, while the residual saturation is zero in the throughout the reservoir. Consequently, the gas saturation in the

9
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

0.05
a no Hys
0.045 Linear
Carlson
0.04

0.035

0.03
WGR
0.025

0.02

0.015

0.01

0.005

0
0 365 730 1095 1460 1825 2190
Time (day)
1400
b Carlson
Linear
1200 no hys
Average pressure (psi)

1000

800

600

400

200

0
0 365 730 1095 1460 1825 2190
Time (day)
Fig. 7. Produced water-gas ratio (a) and average reservoir pressure (b) for different hysteresis models or scenarios.

vicinity of the well block increases, leading to reduced water production permeability and capillary pressure values is ignored, the final hydrogen
and higher reservoir pressure. The pressure difference caused by the recovery will differ significantly between the simulation and the actual
presence of the cushion gas can affect the selection of the appropriate condition.
injection and production pattern within the available capacity. This The difference in the recovery factor is due to the lower relative
enables the utilization of varying injection and production rates for permeability of the mixed gas compared to pure hydrogen, which causes
better economic efficiency, highlighting the importance of considering the injected gas to have a lower tendency to spread laterally and be
the effect of mixing gas on the dynamic behavior of reservoir fluids. retained around the injection well than in the case of using the relative
In Fig. 12, the gas moles in place and the recovery factor of hydrogen permeability of pure hydrogen. As a result, production increases. Fig. 13
for both scenarios during five cycles are compared. As the number of shows the saturation of gas in the idle time in the fifth cycle in both
cycles increases, due to water production and the reservoir's closed (no cases, confirming that the mixed gas spreads less than the pure
flow) boundaries, gas saturation around the well will increase and hydrogen.
become more sensitive to relative permeability curves. When we apply
the relative permeability of the mixed gas, the amount of difference in
4.3. Further discussion on the knowledge gaps and the future direction
recovery between the two scenarios indicates a relative increase
beginning in the third cycle of injection and production of hydrogen,
As shown via numerical simulations, relative permeability and
and increasing with time. Therefore, in the last cycle, we see a difference
capillary pressure curves are essential in simulating hydrogen storage
in recovery factor of roughly 5.7 %. The difference between the relative
and production in underground structures. These simulations are then
permeability and capillary pressure curves of the two cases grows as the
used to design and implement a proper reservoir management workflow.
amount of gas saturation in the two cases goes up. The maximum gas
However, due to the limited number of experimental studies on the
saturation in this simulation, as shown in the figure, is 0.4 around the
dynamic properties of the hydrogen-water system, researchers often rely
well and since most blocks have a gas saturation <0.4, the relative
on available models in their simulations. It is important to exercise
permeability difference is small, and the gas mixing effect is less
caution when using both models and experimental data to avoid
tangible. In more realistic models, the presence of more wells, layers
obtaining unrealistic results. While existing models are well-known to
with high permeability, or fractures increases the number of blocks with
researchers, there are still challenges associated with experimental
high gas saturation; therefore, if the role of gas mixing on relative
measurements that need to be addressed.

10
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Fig. 8. Hydrogen mole in place (a) and nitrogen mole in place (b) for different hysteresis models or scenarios.

• Experimental measurement of saturation functions underground storage structures. The reason behind this is strong spon­
taneous imbibition capillary forces that are present in the outlet/inlet
There are different methods for the measurement of relative faces of the sample due to the strong water-wet state of the hydrogen-
permeability curves, each of which has its advantages and disadvantages water system. This phenomenon was reported by Boon and Hajibeygi
that we need to consider before using. We focus on steady-state and [18] where they observed higher water saturation at the inlet face
unsteady-state methods for our favorite system (i.e., hydrogen-water) during the first step of imbibition. Therefore, the fractions of water and
due to their availability in the literature and the huge amount of expe­ hydrogen streams will be more important when we aim to interpret the
rience that exists in the special core analysis (or SCAL) area. The relative permeability data and then utilize them in the simulation.
following considerations should be taken into account when applying In the steady-state experiments, the fluid flow rates are kept con­
hydrogen-water relative permeability. stant, and the fluid properties are assumed to be uniform throughout the
In the steady-state method, both hydrogen and water phases are co- porous medium. However, this assumption may not be valid in the case
injected at a steady rate into the core sample, and the pressure drop and of a hydrogen-water system as it was observed by scholars that hydrogen
the production rate of each fluid are measured until their values become can create a capillary barrier [18]. It is important to apply bump rates to
constant in time. The same measurements are performed for various suppress the capillary-end-effect in the hydrogen flooding experiments
combinations of flow rates (or fractional flows) to obtain the complete and eliminate its effect on the inferred relative permeability curves due
range of relative permeability curves. This technique provides a full to the ultra-low density of hydrogen especially in horizontal systems.
curve for relative permeability data from initial water saturation to re­ Since the viscosity of hydrogen is very low (<0.01 cp), it is also chal­
sidual saturation (see Fig. 14) with a focus on intermediate saturations, lenging to create sufficient viscous pressure drop, even via bump rates,
especially for drainage. However, a large change in the saturation of the to suppress the capillary-end-effect in the drainage phase. Mid-point
wetting phase (water) can be seen for the imbibition, which can make pressure transducers, away from core ends can be considered to elimi­
the interpretation of the data more challenging. It is important to nate the capillary-end-effect region. Also, it is important to make sure
consider this effect when analyzing the relative permeability data and to that gravity segregation has a minimal effect on fluid displacement and
use appropriate techniques to correct them for the hysteresis observed in distribution.
the imbibition phase. Failure to account for hysteresis can result in An unsteady state method of measuring the relative permeability of a
inaccurate predictions of the behavior of the hydrogen-water system in hydrogen-water system requires first saturating the core with connate

11
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

0.5
a
0.45

0.4

0.35

Sg (block 16,16,1)
0.3

0.25

0.2

0.15
Linear
0.1 no hys
0.05 Carlson

0
0 365 730 1095 1460 1825 2190
Time(day)
b
0.3

0.295
Resitual Gas saturation

0.29

0.285

0.28

0.275

0.27
Linear
0.265
Carlson
0.26

0.255
0.37 0.38 0.39 0.4 0.41 0.42 0.43 0.44 0.45 0.46
Inial Gas Saturaon
Fig. 9. Gas saturation of the well block (a) and the residual saturation as a function of initial gas saturation (b) for different hysteresis models or scenarios.

1
no hys Linear Carlson
0.9
Hydrogen Recovery Factor

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
1 2 3 4 5
Number of Cycles
Fig. 10. Hydrogen recovery factor versus cycle number for different hysteresis models or scenarios.

water (often 100 % if the target reservoir is an aquifer), then injecting volume, and the gas flow rate on both sides of the core over time until
hydrogen gas into the core at a constant rate and measuring the pressure the cumulative volume of produced water becomes constant. Due to the
differential between the two sides of the core, the production water unfavorable mobility ratio, the derived relative permeability curves

12
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

1400
H2 Relave Permeability
1200

Average pressure (psi)

H2+N2 Relave Permeability
1000

800

600

400

200

0
0 365 730 1095 1460 1825 2190
Time (day)
Fig. 11. Average reservoir pressure during five cycles when using i) pure hydrogen and ii) mixed gas (mixture of hydrogen and nitrogen) relative permeability.

a
400000
H2 Relave Permeability H2 + N2 Relave Permeability
350000
Gas mole IPF(gmole)

300000

250000

200000

150000

100000

50000

0
0 365 730 1095 1460 1825 2190
Time(day)
b
1
H2+N2 Relave Permeability H2 Relave Permeability
0.98
Hydrogen Reecovery Factor

0.96
0.94
0.92
0.9
0.88
0.86
0.84
0.82
0.8
1 2 3 4 5

Fig. 12. Hydrogen mole in place with respect to time (a) and hydrogen recovery factor in different cycles (b).

from the unsteady-state experiments cover a limited saturation range methods (such as JBN [50], Jones and Roszelle (JR) [51]), or semi-
out of the entire mobile saturation range. Therefore, at low gas satura­ analytical methods (such as Toth et al. [52] and Civan and Donaldson
tions, there is no true data because relative permeability can not be [53]) or numerically by history matching using a numerical simulator.
obtained before gas breakthrough in drainage experiments. A similar Besides, understanding the quality of experimental data, the limitation
issue exists in imbibition experiments (see Fig. 14). of the measurements and the techniques used, and the inherent uncer­
The experimental data can be interpreted using various analytical tainty of the data and error bars, the use of numerical simulators comes

13
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

Fig. 13. Distribution of hydrogen in the structure using the average relative permeability for the hydrogen‑nitrogen mixture (left) and the relative permeability of
pure hydrogen (right).

0.1
Kr

0.01

0.001

0.0001
0 0.2 0.4 0.6 0.8 1
Sw
Fig. 14. Relative permeability data based on the methodology, blue lines are representative of the steady-state and red lines are representative of the unsteady state
(adopted from [49]). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

with its own challenges. For example, finding the best match between can then be used to match the experiment's history with that estimated
the experimental data and the simulation data requires tuning several by a fit-for-purpose numerical simulator. Borazjani et al. [55] and
unknown parameters of relative permeability and capillary pressure Hemmati et al. [56] developed a technique which make it possible to
models or correlation. In this case, multiple, equally valid matches may extract both relative permeability and capillary pressure from the
be found as this is an inverse problem with more unknowns than knowns transition periods in between two subsequent fractional flow steps in
(i.e., an ill-posed problem). This then requires further scrutiny of the steady-state experiments. Although, suitability of the technique for CO2-
findings to locate the global minima by invoking an optimization water systems has been already demonstrated, its use for hydrogen-brine
scheme and data sampling algorithms (see Berg et al. [54]). system can be further investigated.
Classical methods of interpreting relative permeability often fail to Nevertheless, direct measurement of capillary pressure (by means of
accurately represent the actual flow in the reservoir, as they do not mercury intrusion or MICP method, centrifuge or porous plate tech­
account for capillary pressure effects. Similarly, capillary pressure nique) and using it as an input to the numerical simulator for relative
measurements often overlook the influence of relative permeability. To permeability determination can reduce the uncertainty associated with
address this issue, capillary pressure and relative permeability can be non-unique solutions in the inverse modeling. Otherwise, in-situ satu­
estimated simultaneously by history matching the experimentally ob­ ration data (obtained via X-ray scanning or NMR along the core length
tained production and pressure drop profiles during the displacement and over injection time) can be used (together with production data) to
experiment over time. This requires the use of an (assisted) optimization aid the history-matching process and infer both capillary pressure and
technique to determine the optimal set of saturation functions, which relative permeability from one experiment.

14
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

There are several potential solutions to improve the classical sample's end-face. This end-effect acts as a back pressure, particularly
methods and address the capillary pressure effect. These include using noticeable at low injection pressures and high-water saturations. It is
bump-rates to suppress capillary-end-effect, use of long cores (a few feet important to consider this effect when analyzing the behavior of the
long) to change the balance of viscous and capillary forces, use of hydrogen-water system in underground storage structures to avoid
multiple pressure taps along the core to avoid capillary-end-effect re­ inaccuracies in simulations.
gion, and use of end pieces at the inlet and outlet of the core to minimize
the capillary-end-effect. • Initial wettability of hydrogen-rservoir fluid-rock

• Hydrogen-water relative permeability endpoints The most important factor affecting the immiscible displacement
process in porous media, and the relative permeability, is rock wetta­
The endpoints of the relative permeability curves represent the bility [58]. The non-wetting phase is trapped by the wetting phase when
highest relative permeability of each phase that can be reached at the the porous media shows a strong wettability preference to the wetting
irreducible saturation of the other phases. The consideration of relative phase [59]. Therefore, the chance of hydrogen trapping and conse­
permeability at the endpoints is one of the topics not covered explicitly quently hydrogen loss is high in such a system, especially when it comes
in the articles related to the relative permeability of hydrogen curves. In to a cyclic process [21,60]. Some studies have been conducted in recent
the presence of a trapped phase, the other phase's relative permeability years on evaluating the wettability of the water‑hydrogen-rock system
cannot reflect the rock's absolute permeability. Thus one cannot achieve in various lithologies and investigating the parameters impacting the
a relative permeability of one for hydrogen in a drainage experiment. wettability using dynamic and static methods [61–64]. Most techniques
This is particularly important if a rock contains active minerals such as used flat, smooth rock chips to measure ex-situ contact angle (using e.g.
clays which bind with water, and affect pore fluid flow and connectivity sessile drop method). Most of these studies found that the non-wetting
of the pore space. In this case, a high initial water saturation (Swi) can be fluid in all of the investigations is hydrogen. The rock wettability can
expected. Thus, hydrogen relative permeability at Swi cannot be equal be altered by the presence of residual hydrocarbon on the rock surface,
to one. However, in several studies reported in the literature [5], this fatty acids and polar molecules over the geological time, and become
problem was not noted and the relative permeability at the endpoint less water-wet toward even hydrocarbon-wet. This situation can be
reached the core absolute permeability. relevant for hydrogen storage in depleted oil fields or polluted aquifers
Here we have summarized the available experimental data for the with humic acid or other precipitates. Currently, there is a lack of
hydrogen-water system collected from the published literature in comprehensive understanding of wettability in these systems and how it
Table 5. A careful review of the available data for the hydrogen-water affects the performance of UHS.
system shows low relative permeability values for hydrogen compared Moreover, the link between ex-situ and in-situ wettability data needs
to water. Endpoint values for the steady-state experiments are within the more investigation, as the in-situ data may be more realistic. In only a
range of 0.019–0.044 (considering water saturation of 0.40 to 0.64), few cases the contact angle was measured in-situ using microfluidics and
while the unsteady-state method shows a value with one order of micro-CT scanning. Wettability can be affected by mineralogy, pore
magnitude higher (0.34) at 0.29 water saturation. The analysis shows fluids, pore structure, surface roughness, heterogeneity, etc. Thus core-
hydrogen acts as a strong non-wetting phase in the hydrogen-water scale experiments should be conducted at realistic initial wettability. In
system. Based on this observation, a higher end-point value was ex­ the case of hydrocarbon recovery, this can be achieved by proper
pected during relative permeability tests. However, there are other initialization of the core with fluids followed by aging at reservoir
points that we need to take into account. For instance, gas slippage and conditions. In the case of hydrogen-water, and depending on the pre-
its interplay with the capillary-end-effect and very low viscosity of existing pore fluids in the target reservoir, this should be done care­
hydrogen. The gas slippage effect is a phenomenon that describes the fully; whether the hydrogen is injected into an aquifer or a depleted
slippage of gas molecules on the surface of the porous media. We believe hydrocarbon reservoir the initial wettability would be different. This can
this effect should be more pronounced in the hydrogen-water system become even more complex when a cushion gas other than hydrogen is
due to the small molecule size of hydrogen [57]. Gas slippage can injected. The wetting properties of the hydrogen-cushion gas mixture
significantly cause an increase in Krg value at the endpoint. This is due should be determined and there is currently a lack of data in this regard.
to the expected increase in pore radius at lower water saturations (i.e. a Despite the experimental evidence suggesting minimal risk of hydrogen
thinner wetting phase layer), which leads to the underestimation of the loss and well & reservoir integrity issues due to abiotic geochemical
gas slippage factor. This reduction amplifies the impact of gas slippage reactions between hydrogen and reservoir rock or wellbore cement
on the measured Krg at higher saturations. However, the behavior of the [65,66], the in-situ geochemical and biochemical reactions may still
hydrogen-water system differs from this theory. It appears that, in a alter rock wettability and permeability/pore structure. Therefore a
hydrogen-water system, the slippage factor varies primarily in response deeper understanding of this topic is really necessary. Consequently, the
to the additional pressure drop caused by the capillary-end-effect on the data obtained from the core-scale experiments should be used with

Table 5
Summary of the relative permeability experiments related to hydrogen-water system.
End point Sw @ end Method Core permeability Flow rate m& Sample Capillary pressure Insitu saturation Reference
of H2 point (mD) (cc/min) n orientation determination method measurement

0.037 0.59 SS 107 1 3&2 Vertical Porous plate Yes [21]

0.044 0.41 SS 44 1 3&9 Vertical Calculation No [16]
0.034 0.4 Vertical
0.34 0.295 USS 34 N/A N/A Vertical MICP No [17]
0.35 0.227 N/A N/A Vertical
0.019 0.64 SS 203 5 2& Horizontal MICP+calculation Yes [18]
1.5

SS: steady-state.
USS: unsteady-state.
m: Corey exponent for water.
n: Corey exponent for hydrogen.

15
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

caution in reservoir-scale studies. In this regard, the use of computa­ similarities with conventional water-gas systems, but there are also
tional fluid dynamics (CFD) methods coupled with digital core analysis specific complexities. These complexities include more gas fingering due
techniques on the pore-scale may be fruitful for further understanding of to the high mobility of hydrogen, gravity segregation and the gas
the fluid displacement process and trapping mechanisms under the in­ override phenomenon. Besides, the hydrogen-rock geochemical in­
fluence of rock-fluid & fluid-fluid interactions and to better design the teractions, which are not-well-understood, were identified as the main
required experiments. challenging factors affecting the wettability and formation damage. The
primary source of uncertainty in UHS studies is the lack of reliable
• Hysteresis experimental relative permeability data for different conditions and
systems, as well as the failure to consider hysteresis and gas mixing
As shown in Section 4, hysteresis is an effective factor in various phenomena.
parameters like recovery efficiency and reservoir pressure along sea­ Our simulations demonstrated that hysteresis has a significant
sonal underground hydrogen storage. Therefore, considering hysteresis impact on the total volume of gas produced, the volume of water pro­
plays a key role in terms of predicting the possibility of hydrogen storage duced, the purity of the produced gas, and the reservoir pressure. Ac­
operations and economic feasibility in an underground reservoir. We counting for hysteresis in the simulation resulted in a 16 % to 25 %
used two models (Carlson and Linear) to evaluate the presence of hys­ reduction in the recovery factor of hydrogen gas. We also found that the
teresis on UHS in an aquifer model. However, hysteresis is a complex presence of cushion gas in the reservoir would enhance hydrogen pro­
phenomenon which is influenced by several parameters including the duction. This resulted in a higher recovery factor of approximately 5 %
number of phases involved. While there are other hysteresis models due to the lower relative permeability of the mixed gas (nitrogen and
available in the literature, including Killough [67], the Scanning Hys­ hydrogen) compared to pure hydrogen.
teresis Model (SHM) described in [68] on the basis of experimental data
provided by Gladfelter and Gupta [69], and Braun and Holland [70],
Declaration of competing interest
and Larsen & Skauge [71] model. However, further investigation is
needed to determine the suitability of different models for hydrogen-
The authors declare that they have no known competing financial
water flow, and new comprehensive experimental data should become
interests or personal relationships that could have appeared to influence
available.
the work reported in this paper.
There are some critical items that should be considered during
relative permeability hysteresis analysis. Initial and final saturation
Data availability
conditions matter the most during the quantification of the data. The
relative permeability hysteresis behavior is strongly dependent on the
No data was used for the research described in the article.
initial and final saturation conditions. Therefore, it is essential to
establish the initial and final saturation conditions carefully and accu­
rately. The pore structure, pore size distribution, local mineralogy, and Acknowledgments
wettability of the porous medium strongly influence the relative
permeability hysteresis behavior. It is important to understand the The authors express their gratitude for the financial support received
wettability of the medium and pore structure to interpret the results from Sharif University of Technology (G4000103).
correctly. The flow rate and injection method used during the experi­
ment can significantly affect the relative permeability hysteresis References
behavior. Therefore, it is important to control the flow rate and injection
[1] N.C.E.A. Plan, Communication from the Commission to the European Parliament,
method carefully and consistently. Our simulation results show that due
the Council, the European Economic and Social Committee and the Committee of
to hysteresis the hydrogen recovery is reduced by 16–25 %. Apart from the Regions A New Circular Economy Action Plan for a Cleaner and More
the amount of actual value which is strongly dependent on several fac­ Competitive Europe, 2009.
[2] Paris Agreement, United nations, United Nations Treaty Collect, 2015, p. 1e27.
tors, our observed results are in line with Pan et al., Bo et al., and
[3] A. Kabuth, et al., Energy storage in the geological subsurface: dimensioning, risk
Ershadnia et al., where they have reported a reduction in hydrogen re­ analysis and spatial planning: the ANGUS+ project, Environ. Earth Sci. 76 (2017)
covery [24–26]. 1–17.
In addition to the difference in the behavior during imbibition and [4] Hydrogen Storage, Available from: https://www.energy.gov/eere/fuelcells
/hydrogen-storage.
drainage, other factors, such as reservoir geometry, geological hetero­ [5] L. Hashemi, M. Blunt, H. Hajibeygi, Pore-scale modelling and sensitivity analyses
geneity, injection rate, etc. can also affect hysteresis in the hydrogen- of hydrogen-brine multiphase flow in geological porous media, Sci. Rep. 11 (1)
water system. Therefore, a comprehensive understanding of these pa­ (2021) 1–13.
[6] A. Aftab, et al., Toward a fundamental understanding of geological hydrogen
rameters is crucial for accurate interpretation of the hysteresis data and storage, Ind. Eng. Chem. Res. 61 (9) (2022) 3233–3253.
reliable simulation of the behavior of the system in underground storage [7] D. Zivar, S. Kumar, J. Foroozesh, Underground hydrogen storage: a comprehensive
structures. review, Int. J. Hydrog. Energy 46 (45) (2021) 23436–23462.
[8] A. Hassanpouryouzband, et al., Offshore geological storage of hydrogen: is this our
best option to achieve net-zero? ACS Energy Lett. 6 (6) (2021) 2181–2186.
5. Summary and concluding remarks [9] N.S. Muhammed, et al., A review on underground hydrogen storage: insight into
geological sites, influencing factors and future outlook, Energy Rep. 8 (2022)
461–499.
This work started with critical review of the literature on the dy­ [10] M.K. Singla, P. Nijhawan, A.S. Oberoi, Hydrogen fuel and fuel cell technology for
namic properties of hydrogen-water systems and numerical simulations cleaner future: a review, Environ. Sci. Pollut. Res. 28 (2021) 15607–15626.
of underground hydrogen storage and production. We utilized the CMG- [11] N. Heinemann, et al., Enabling large-scale hydrogen storage in porous media–the
scientific challenges, Energy Environ. Sci. 14 (2) (2021) 853–864.
GEM to simulate UHS scenarios for a conceptual geological model, with [12] R.H. Brooks, Hydraulic Properties of Porous Media, Colorado State University,
a focus on two critical phenomena, namely hysteresis and gas mixing. 1965.
Through our analysis, we identified the knowledge gaps in the UHS [13] Y. Mualem, A new model for predicting the hydraulic conductivity of unsaturated
porous media, Water Resour. Res. 12 (3) (1976) 513–522.
laboratory studies. We also highlighted the importance of incorporating
[14] M.T. Van Genuchten, A closed-form equation for predicting the hydraulic
the existing petroleum engineering core analysis expertise into conductivity of unsaturated soils, Soil Sci. Soc. Am. J. 44 (5) (1980) 892–898.
enhancing UHS laboratory research works, design and interpretation of [15] A.T. Corey, The interrelation between gas and oil relative permeabilities, Producers
core-scale data to find reliable and representative flow functions. This monthly (1954) 38–41.
[16] A. Yekta, et al., Determination of hydrogen–water relative permeability and
would critically improve the feasibility studies to become more reliable. capillary pressure in sandstone: application to underground hydrogen injection in
Our analysis revealed that the hydrogen-water system shares sedimentary formations, Transp. Porous Media 122 (2) (2018) 333–356.

16
M. Bahrami et al. Journal of Energy Storage 73 (2023) 108886

[17] A. Rezaei, et al., Relative permeability of hydrogen and aqueous brines in [44] C.S. Land, Calculation of imbibition relative permeability for two-and three-phase
sandstones and carbonates at reservoir conditions, Geophys. Res. Lett. 49 (12) flow from rock properties, Soc. Pet. Eng. J. 8 (02) (1968) 149–156.
(2022) p. e2022GL099433. [45] A. Hassanpouryouzband, et al., Thermodynamic and transport properties of
[18] M. Boon, H. Hajibeygi, Experimental characterization of H 2/water multiphase hydrogen containing streams, Sci. Data 7 (1) (2020) 222.
flow in heterogeneous sandstone rock at the core scale relevant for underground [46] K. Li, R.N. Horne, Experimental study of gas slippage in two-phase flow, SPE
hydrogen storage (UHS), Sci. Rep. 12 (1) (2022) 14604. Reserv. Eval. Eng. 7 (06) (2004) 409–415.
[19] F. Feldmann, et al., Numerical simulation of hydrodynamic and gas mixing [47] C. Ruprecht, et al., Hysteretic trapping and relative permeability of CO2 in
processes in underground hydrogen storages, Environ. Earth Sci. 75 (2016) 1–15. sandstone at reservoir conditions, Int. J. Greenh. Gas Control 27 (2014) 15–27.
[20] S.M. Fatemi, M. Sohrabi, Relative permeabilities hysteresis for oil/water, gas/ [48] R. Sedaghatinasab, et al., Relative permeability hysteresis and capillary trapping
water and gas/oil systems in mixed-wet rocks, J. Pet. Sci. Eng. 161 (2018) during CO2 EOR and sequestration, Int. J. Greenh. Gas Control 106 (2021),
559–581. 103262.
[21] M. Lysyy, et al., Hydrogen relative permeability hysteresis in underground storage, [49] C. McPhee, J. Reed, I. Zubizarreta, Core Analysis: A Best Practice Guide, Elsevier,
Geophys. Res. Lett. 49 (17) (2022) p. e2022GL100364. 2015.
[22] N.R. Morrow, The effects of surface roughness on contact: angle with special [50] E. Johnson, D. Bossler, V.N. Bossler, Calculation of relative permeability from
reference to petroleum recovery, J. Can. Pet. Technol. 14 (04) (1975). displacement experiments, Trans. AIME 216 (01) (1959) 370–372.
[23] P.H. Valvatne, M.J. Blunt, Predictive pore-scale modeling of two-phase flow in [51] S. Jones, W. Roszelle, Graphical techniques for determining relative permeability
mixed wet media, Water Resour. Res. 40 (7) (2004). from displacement experiments, J. Pet. Technol. 30 (05) (1978) 807–817.
[24] Z. Bo, et al., Impact of experimentally measured relative permeability hysteresis on [52] J. Toth, et al., Convenient formulae for determination of relative permeability from
reservoir-scale performance of underground hydrogen storage (UHS), Int. J. unsteady-state fluid displacements in core plugs, J. Pet. Sci. Eng. 36 (1–2) (2002)
Hydrog. Energy 48 (2023) 13527–13542. 33–44.
[25] B. Pan, et al., Impacts of relative permeability hysteresis, wettability, and [53] F. Civan, E. Donaldson, Relative permeability from unsteady-state displacements
injection/withdrawal schemes on underground hydrogen storage in saline aquifers, with capillary pressure included, SPE Form. Eval. 4 (02) (1989) 189–193.
Fuel 333 (2023) 126516. [54] S. Berg, E. Unsal, H. Dijk, Non-uniqueness and uncertainty quantification of
[26] R. Ershadnia, et al., Impact of geological and operational conditions on relative permeability measurements by inverse modelling, Comput. Geotech. 132
underground hydrogen storage, Int. J. Hydrog. Energy 48 (4) (2023) 1450–1471. (2021), 103964.
[27] M. Akbarabadi, M. Piri, Relative permeability hysteresis and capillary trapping [55] S. Borazjani, et al., Simultaneous determination of gas–water relative permeability
characteristics of supercritical CO2/brine systems: an experimental study at and capillary pressure from steady-state corefloods, J. Hydrol. 598 (2021), 126355.
reservoir conditions, Adv. Water Resour. 52 (2013) 190–206. [56] N. Hemmati, et al., A novel method to determine relative permeability and
[28] B. Hagemann, et al., Mathematical modeling of unstable transport in underground capillary pressure from corefloods application to CO2 storage, in: SPE Asia Pacific
hydrogen storage, Environ. Earth Sci. 73 (2015) 6891–6898. Oil and Gas Conference and Exhibition, SPE, 2022.
[29] B. Hagemann, et al., Hydrogenization of underground storage of natural gas: [57] H2tools. https://h2tools.org/hyarc/hydrogen-data/basic-hydrogenproperties, 2023.
impact of hydrogen on the hydrodynamic and bio-chemical behavior, Comput. [58] W. Owens, D. Archer, The effect of rock wettability on oil-water relative
Geosci. 20 (2016) 595–606. permeability relationships, J. Pet. Technol. 23 (07) (1971) 873–878.
[30] A. Sáinz-García, et al., Assessment of feasible strategies for seasonal underground [59] T. Geffen, et al., Experimental investigation of factors affecting laboratory relative
hydrogen storage in a saline aquifer, Int. J. Hydrog. Energy 42 (26) (2017) permeability measurements, J. Pet. Technol. 3 (04) (1951) 99–110.
16657–16666. [60] E. Spiteri, et al., Relative permeability hysteresis: trapping models and application
[31] W.T. Pfeiffer, S. Bauer, Subsurface porous media hydrogen storage–scenario to geological CO2 sequestration, in: SPE Annual Technical Conference and
development and simulation, Energy Procedia 76 (2015) 565–572. Exhibition, OnePetro, 2005.
[32] W. Pfeiffer, et al., Porous media hydrogen storage at a synthetic, heterogeneous [61] S. Iglauer, M. Ali, A. Keshavarz, Hydrogen wettability of sandstone reservoirs:
field site: numerical simulation of storage operation and geophysical monitoring, implications for hydrogen geo-storage, Geophys. Res. Lett. 48 (3) (2021) p.
Environ. Earth Sci. 75 (2016) 1–18. e2020GL090814.
[33] W.T. Pfeiffer, C. Beyer, S. Bauer, Hydrogen storage in a heterogeneous sandstone [62] L. Hashemi, et al., Contact angle measurement for hydrogen/brine/sandstone
formation: dimensioning and induced hydraulic effects, Pet. Geosci. 23 (3) (2017) system using captive-bubble method relevant for underground hydrogen storage,
315–326. Adv. Water Resour. 154 (2021), 103964.
[34] K. Luboń, R. Tarkowski, Numerical simulation of hydrogen injection and [63] S. Higgs, et al., In-situ hydrogen wettability characterisation for underground
withdrawal to and from a deep aquifer in NW Poland, Int. J. Hydrog. Energy 45 (3) hydrogen storage, Int. J. Hydrog. Energy 47 (26) (2022) 13062–13075.
(2020) 2068–2083. [64] W. van Rooijen, et al., Microfluidics-based analysis of dynamic contact angles
[35] D.S. Mahdi, et al., Hydrogen underground storage efficiency in a heterogeneous relevant for underground hydrogen storage, Adv. Water Resour. 164 (2022),
sandstone reservoir, Adv. Geo-Energy Res. 5 (4) (2021) 437–443. 104221.
[36] M. Lysyy, M. Fernø, G. Ersland, Seasonal hydrogen storage in a depleted oil and gas [65] A. Hassanpouryouzband, et al., Geological hydrogen storage: geochemical
field, Int. J. Hydrog. Energy 46 (49) (2021) 25160–25174. reactivity of hydrogen with sandstone reservoirs, ACS Energy Lett. 7 (7) (2022)
[37] M. Kanaani, B. Sedaee, M. Asadian-Pakfar, Role of cushion gas on underground 2203–2210.
hydrogen storage in depleted oil reservoirs, J. Energy Storage 45 (2022), 103783. [66] A. Aftab, et al., Geochemical integrity of wellbore cements during geological
[38] E.R. Okoroafor, S.D. Saltzer, A.R. Kovscek, Toward underground hydrogen storage hydrogen storage, Environ. Sci. Technol. Lett. 10 (2023) 551–556.
in porous media: reservoir engineering insights, Int. J. Hydrog. Energy 47 (79) [67] J. Killough, Reservoir simulation with history-dependent saturation functions, Soc.
(2022) 33781–33802. Pet. Eng. J. 16 (01) (1976) 37–48.
[39] M. Delshad, et al., Hydrogen storage assessment in depleted oil reservoir and saline [68] B. Plohr, et al., Modeling hysteresis in porous media flow via relaxation, Comput.
aquifer, Energies 15 (21) (2022) 8132. Geosci. 5 (2001) 225–256.
[40] S.P. Dutton, T.N. Diggs, Evolution of porosity and permeability in the lower [69] R.E. Gladfelter, S.P. Gupta, Effect of fractional flow hysteresis on recovery of
Cretaceous Travis Peak Formation, East Texas, AAPG Bull. 76 (2) (1992) 252–269. tertiary oil, Soc. Pet. Eng. J. 20 (06) (1980) 508–520.
[41] M. Panfilov, Underground and pipeline hydrogen storage, in: Compendium of [70] E. Braun, R. Holland, Relative permeability hysteresis: laboratory measurements
hydrogen energy, Elsevier, 2016, pp. 91–115. and a conceptual model, SPE Reserv. Eng. 10 (03) (1995) 222–228.
[42] Computer Modeling Group, GEM Manual. www.CMG.Ca. [71] J. Larsen, A. Skauge, Methodology for numerical simulation with cycle-dependent
[43] F.M. Carlson, Simulation of relative permeability hysteresis to the nonwetting relative permeabilities, SPE J. 3 (02) (1998) 163–173.
phase, in: SPE Annual Technical Conference and Exhibition, OnePetro, 1981.

17