Simulation of CO 2 injection in a depleted gas reservoir: A case study for Upper Miocene sandstone, Northern Croatia

Carbon capture and storage (CCS) technology is a bene ﬁ cial greenhouse gas mitigating strategy carried out in the last 20 years. Depleted gas reservoirs are promising candidates for the storage of carbon dioxide (CO 2 ). Therefore, a depleted gas reservoir in the Upper Miocene sandstone located in Northern Croatia was taken as an example. The purpose of this study was to compare CO 2 storage capacity obtained with two analytical equations to total storage capacity obtained through the simulator, in order to validate the equations. The ﬁ rst equation takes the average reservoir pressure and available production data into account, while the other one is more general and includes produced volume, CO 2 density and formation volume factor of the original ﬂ uid. The tools used for these calculations were Schlumberger PVTi soft-ware, in which the equation of state was obtained, and ECLIPSE (E300 Module) which is a reservoir engineering simulator used for reservoir behaviour prediction. The results con ﬁ rmed analytical solutions, indicating that, depending on the depth, the mass of the CO 2 that can be injected is twice as big as the mass of CH 4 produced. The results of analytical solutions, 16.7 × 10 6 m 3 and 14.6 × 10 6 m 3 , are in accordance with the results obtained by the simulation of CO2 injection in depleted reservoirs - 16.2 × 10 6 m 3 . Based on this, a conclusion is derived that these analytical solutions can be used as a ﬁ rst approximation of injection in a depleted gas reservoir.


Introduction
There are several possibilities for carbon dioxide (CO 2 ) sequestration, among which injection into oil and gas reservoirs for enhanced recovery is the most feasible one. This is due to the possibility of recovering additional quantities of oil and gas. However, considering that a vast number of oil and gas fi elds in Croatia are experiencing a signifi cant decline (Velić et al., 2016), injection into abandoned/depleted oil and gas reservoirs is also an acceptable form of CO 2 reduction. In this case, no additional value is created. Gaurina-Međimurec et al. (2018) stated that hydrocarbon reservoirs are considered one of the most favourable options for CO 2 disposal. Gas reservoirs are the most reliable potential storage locations since produced gas can be considered as an indication that the reservoir would be impermeable for the same volume of injected CO 2 (Vulin, 2010). Additionally, Novak et al. (2013a) report that the reservoirs in the Croatian part of the Pannonian Basin System (CPBS) are at a suffi cient depth and of older age, which means that they are consolidated and not prone to tectonic activity. Some of the advantages of depleted reser-voirs, compared to storage in aquifers and other geological formations, are well-characterized reservoirs with known properties, proven traps capable of retaining fl uids for a long time, small exploration costs and a possibility of reusing some of the existing infrastructure (URL1). The advantage of sequestration into depleted reservoirs, compared to enhanced oil/gas recovery (EOR/EGR) processes, is that CO 2 is not produced but trapped underground. There are several trapping mechanisms that hinder CO 2 from escaping the geological formations and these mechanisms depend on the type of formation in which CO 2 is stored. These mechanisms are structural or stratigraphic trapping assured by the cap rock, residual gas trapping connected with the drainageimbibition process in the post-injection phase -hysteresis, solubility/dissolution trapping in an aquifer, and mineral trapping, which is a reaction of CO 2 with rock minerals to create precipitates (Temitope et al., 2016;Raza et al., 2017). Structural trapping is dominant in the fi rst period after injection stops, with negligible mineral trapping. Later, after 50-150 years, residual trapping becomes more signifi cant. Solubility trapping has an almost constant share but becomes more important at later times (500+ years). Storage security increases with time (Ampomah et al., 2015). Raza et al. (2017)  for short times, residual trapping is most effi cient. Novak (2015) examines the effect of mineral trapping and fi nds that 2-5% of CO 2 injected could be permanently stored with this mechanism.
An additional advantage of sequestration in depleted gas reservoirs is geothermal exploitation but is valid only for high-temperature reservoirs. It is based on the fact that supercritical CO 2 has 1.5 times higher heat capacity compared to water Cui et al. 2016). It is possible to store more CO 2 in a depleted gas reservoir having the same hydrocarbon pore volume than in an oil reservoir. This is due to a higher ultimate recovery of gas reservoirs compared to oil reservoirs, and higher compressibility of the gas (Lawal and Frailey, 2004;Stein et al., 2010). While injection into oil reservoirs, which are depleted or near the end of production, is usually accompanied by the EOR process, depleted gas reservoirs should be used as storage only. In the case of injection into depleted gas reservoirs, CO 2 could contaminate the remaining natural gas. Therefore, it is recommended to inject CO 2 only in reservoirs that would not become economic with an increase in gas prices. The study of IEA Greenhouse Gas R&D Programme (URL1) estimated that the global potential for CO 2 storage in disused gas reservoirs is 797 × 10 12 kg and in disused oil reservoirs 126 × 10 12 kg. The cost of sequestration into oil and gas reservoirs could be less than $2.5/kg for 746 × 10 12 kg. When it comes to gas reservoirs, only around 105 × 10 12 kg of CO 2 could be stored at a cost of $0.6/kg, while the storage of an additional 575 × 0 12 kg could be achieved at a cost of $0.8-1.4/kg. The rest of 117 × 10 12 kg could be stored at a price from $1.7 to 8.3/kg (URL1). It also showed that only 75-80% of void space left after primary gas recovery could be used for CO 2 storage, mainly due to water inclusion and the fi eld edge effect. However, the same study, along with Cui et al. (2015), indicates that due to the higher compressibility of CO 2 compared to methane, the amount of CO 2 that could be injected is higher than the amount of natural gas produced. This ratio highly depends on the injection depth. At 1000 m, a depleted gas reservoir can withstand 3 times the standard volume of CO 2 compared to a standard volume of CH 4 , and at 3000 m, this ratio is only 1.5 (URL1). During the injection, the seal integrity or impermeable caprock is crucial since CO 2 is injected in supercritical condition, which means it tends to migrate upwards, relative to water present in the pores, after the gas depletion took place. However, geomechanics and seal stability were not considered and are out of the scope of this paper. It was assumed that the caprock is proven to be reliable and no signifi cant effect on geomechanics would be evident due to a relatively small amount of CO 2 injected. Sequestration in oil reservoirs is better when no previous water injection or enhanced oil recovery is performed (Pawar et al., 2004). When it comes to gas reservoirs, low gas saturations are desirable, i.e., it is recommended to have secondary recovery (Raza et al., 2018). In the same study, it was proven that at low residual gas saturation, structural trapping is the main mechanism and at high residual gas saturation (30%), capillary trapping dominates over dissolution and structural trapping. It was also stated that after 1500 years, only 10% of injected CO 2 is dissolved in immobile reservoir water. The effect of residual gas saturation was also studied by Raza et al. (2018). It was concluded that previous research showed a negative infl uence of residual gas saturation on injectivity and storage capacity since the dissolution of gas mixtures in supercritical CO 2 reduces the density and viscosity of gas mixtures. In addition, it was found that in the case of EGR, stabilization of production occurs in the early years and production decline starts earlier in a high residual gas saturation reservoir. Finally, it was concluded that high residual gas saturation impacts the relative permeability of gas, thus infl uencing the recovery factor. Low residual fl uid reservoirs make better storage locations.  (2005) observed the effects of CO 2 diffusion and solubility in water and concluded that gas diffusion is important in mixing CO 2 with gas present in the reservoir. A CO 2 breakthrough in the EGR/ EOR process is delayed by the dissolution of CO 2 in water. Loeve et al. (2014) investigated the propagation of temperature during the injection of CO 2 . Carbon dioxide is injected at a minimum of 12 °C. The research showed that after 5 years, a radius of up to 100 m from the borehole has a signifi cantly lower temperature than the rest of the reservoir in the case of high brine saturation. If the brine saturation is low, the cold front does not exist.
Galic et al. (2009) described building a model of CO 2 injection in a depleted gas reservoir through Integrated Production Modelling Petroleum Experts (IPM PETEX) software, with the main focus on the impact of the pipeline surrounding temperature on bottom hole injection temperature and infl uence of injection manifold pressure change. Arts et al. (2012) used Shell's compositional simulator, MoReS, to simulate the injection of CO 2 , which was supposed to be collected at the point source of coal power plant, in a depleted offshore gas fi eld in the Netherlands. The simulation was done for a period of 5 years, at the end of which the original pressure was not achieved. The study showed that injection through a previous production well could be prolonged, to the point of reaching the original pressure. The paper also showed a detailed geological setting of the gas reservoir as well as the monitoring plan that was to be set during and after CO 2 injection as a means of controlling CO 2 and reservoir behaviour. Luo et al. (2013) investigated the infl uence of the reservoir heterogeneity and well placement on CO 2 storage with the conclusion that CO 2 is transported faster in the heterogeneous reservoir. Furthermore, if the wells (producer and injector) are located in the lower permeability layer, storage is improved. In this paper, isothermal conditions are assumed, but there are a number of studies that deal with the change of the temperature and Joule-Thomson Cooling effect ( 2015), calculation of the CO 2 storage capacity can be divided into three groups: volume-based, production-based, and numerical method. The volume-based method is simple and convenient but due to heterogeneity and geological uncertainty, this method is not reliable. The production-based method for CO 2 storage capacity, introduced by Tao and Clarens (2013), is based on historical and projected CH 4 production data. The numerical method is complex, sitespecifi c and the required data must be very detailed. Valbuena et al. (2012) developed an algorithm for estimating cumulative CO 2 storage capacity where 1.023 pore volumes of CO 2 were injected. Injection rate and a number of wells determine only the duration of injection and CO 2 storage capacity depends on saturation, temperature, pressure differential and fl uid composition and characteristics.
Several studies deal with the material balance of CO 2 injection in depleted gas reservoirs (Lawal and Frailey, 2004;Stein et al., 2010. Although calculation in the IPM PETEX MBAL (Material Balance) module is somewhat simplifi ed, Lawal and Frailey (2004) obtained a p/z vs. G p curve for hydrocarbon gas and for CO 2 and showed that the volume of CO 2 that could be injected is less than double the volume of hydrocarbon produced. The p/z vs. G p method is used for volumetric dry gas reservoirs, in which the ratio of pressure and z-factor yields a straight line when plotted versus cumulative gas production. This straight line is extrapolated to a value of p/z to zero for initial gas in place estimation and can also be used to estimate the ultimate recovery at a selected p/z value. Iogna et al. (2017) simulated injection of CO 2 in a depleted gas fi eld using the Black Oil module in ECLIPSE, but the paper mainly dealt with EGR. The obtained results were a bit more conservative than results from a fully compositional model. It was found that reservoir properties, the conversion sequence, well positioning and the faults transmissivity effect enhanced gas recovery.
There are a number of studies regarding seal stability (Orlic, 2009 2012), presenting a geological sequestration of CO 2 without either form of enhanced recovery. However, the data presented in the paper is insuffi cient for the work to be reproducible and applicable to the case of Croatian sandstone and the simulation was done in a different simulator than this research. Various studies dealing with different CO 2 storage issues have been published over the last twenty years (Orlic, 2009 Therefore, previously developed simulation models almost always contained a production well and calculated the additional recovery (Iogna et al., 2017). A simulation model of this case study for Croatian sandstone contains only one well which was used as a production well until production ended and then the same well is used for CO 2 injection with the aim of permanent CO 2 storage. As mentioned, geological sequestration without the enhanced recovery practically could not be found in the literature, except for one paper (Arts et al., 2012), and the relevance of this work can be found in the applicability of simulation for a typical Croatian gas reservoir. An example of a small depleted gas reservoir was considered in terms of the gas production history of the fi eld, reservoir pressure and CO 2 that could be injected to replace the produced gas. Simulation in the Schlumberger ECLIPSE (E300) Package was made to check the cumulative injection of CO 2 compared to published, theoretical amounts and also in order to monitor the reservoir pressure so it does not exceed the fracturing pressure.
Carbon capture and storage (CCS) is a methodology which includes the capture, transport, and storage of CO 2 (Gaurina-Međimurec et al., 2018). Considering the emission quantities, Croatia is not a big emitter of CO 2 -around 20 × 10 9 kg in 2015, according to European Commission (URL2). Therefore, the need for storage is accordingly small so quantities of CO 2 stored obtained from the simulation are realistic and would mean an elimination of around 0.18% of CO 2 emitted in Croatia. The research presented in this manuscript deals with the storage process of CO 2 in a depleted gas reservoir, while the capture and transport processes were not considered.

Methods
There are several published analytical methods presented by Lai (Lai et al., 2015), Bachu (Bachu et al.,  for estimating CO 2 storage capacity in depleted gas reservoirs and two of them were used for comparison with simulated results. Lai et al. (2015) found that by reverting pressure to the initial reservoir pressure, the total volume of CO 2 stored could be 1.4 times larger than that of the gas production. In this work, Equation 1 was used for calculating the theoretical amount of CO 2 that could be injected in the reservoir MALM. The material balance equation (MBE) is used for estimating the original gas in place (OGIP) from the available production data and average reservoir pressure. It was assumed that the hydrocarbon pore volume of the gas reservoir was unchanged during gas production and CO 2 injection and that reservoir volume of the OGIP should be equal to the reservoir volume of the mixture of remaining gas in the reservoir and injected CO 2 . The liquid production was eliminated from the MBE and the formula for CO 2 storage in the depleted dry gas reservoir was given (Lai et al., 2015):

2007) and Schuppers
(1) Where = cumulative CO 2 injected at s.c. (m 3 ), G p = cumulative gas production at s.c. (m 3 ), G i = original gas in place volume at s.c. (m 3 ), p r = reverted pressure of gas reservoir with a mixture of gas and CO 2 (bar), z mix = gas deviation factor of the mixture of natural gas and CO 2 (dimensionless), z i = gas deviation factor (z-factor) at initial reservoir condition (dimensionless), p i = initial reservoir pressure (bar) Schuppers et al. (2003) gave this general formula for CO 2 mass injection calculation: (2) Where = mass of CO 2 that can be injected (kg), = CO 2 density at reservoir conditions (kg/m 3 ), B gi = formation volume factor of natural gas (m 3 /m 3 ) Water production is neglected in Equation 2, which is applicable in this case since water production was either not reported properly or negligible. Equations 1 and 2 represent the theoretical capacity for CO 2 sequestration and in order to obtain the effective capacity, some other factors have to be taken into account, such as the mobility ratio of CO 2 and reservoir fl uid, the production mechanism, reservoir heterogeneity, water saturation and other effects (Novak, 2015). The novelty of this work is an attempt to incorporate all of the stated effects in the simulation.
Regarding the pore pressure and overburden pressure, the injection bottom hole pressure of 200 bar was taken, since Chen et al. (2015) recommended injection pressure below lithostatic. Due to the normal pore pressure and overburden (geostatic) pressure, the CO 2 injection pressure of 200 bar was taken. The perforation depth is 1569 m. According to Eaton (1972), normal pore pressure for sandstones equals 0.10519 bar/m which gives a pressure of 165.04 bar, while the overburden gradient for sandstones equals 0.22621 bar/m which gives a pressure of 354.92 bar.

Geological setting of the reservoir
There are around 4500 exploration, production or development oil and gas wells altogether in Croatia (URL3). Of that number, there are around 950 exploration wells which are located in the Croatian part of the Pannonian Basin System (CPBS) (Velić et al., 2002). The CPBS is located in the southwestern parts of the larger unit of the Pannonian Basin (PBS). It is divided into four main depressions: Sava, Drava, Mura, and Slavonija-Srijem. In continental Croatia, hydrocarbon reservoirs are found within these main depressions. The CPBS was formed during the Neogene, with three megacycles of sedimentation (Velić et al., 2002). The sediments are usually of Quaternary and Tertiary origin, and they overly crystalline bedrock or Mesozoic sedimentary rocks (Velić et al., 2012). The reservoirs are usually found within the sediments of the 1 st and 2 nd megacycle which are mainly represented by sandstones, breccias, basal conglomerates, turbiditic lobes, sand sheets. etc. (Saftić et al., 2003). Figure 1 shows chronostratigraphic and lithostratigraphic units for each of the depressions with assumed development of the lithostratigraphic facies. The thickness of Neogene and Quaternary sediments varies within depressions. The thickness of sediments in the Drava depression reach up to 7000 m in the thickest parts, in the Sava and Mura depression around 5000 m and up to 4000 m in the Slavonija-Srijem depression (Saftić et al., 2003; Velić et al., 2012). The chosen depleted gas reservoir is located in the Upper Miocene formation, with a general geographical location in Northern Croatia. The Upper Miocene sediments of the 2 nd megacycle, are mainly comprised of sandstones of turbiditic or deltaic origin (Saftić et al., 2003). They are important hydrocarbon reservoirs. Figure 2 shows the locations of the PBS and the CPBS, with sediment thickness and major oil, gas and condensate fi elds. Northern Croatia is a geographical area, which spans over the Mura depression, partly over the Drava depression and in a small part over the Sava depression and its approximate area is marked by a red rectangle in Figure 2.

Reservoir properties
In this case study, an example of a typical small depleted and abandoned gas reservoir was used. The reservoir is located in the CPBS in Northern Croatia, in the Upper Miocene fi ne-grained quartz mica sandstones. The caprock is a relatively thin marl deposition (10 m) which represents a good isolator that hinders hydrocarbon migration. The area of the reservoir is 238 400 m 2 and the volume is 631 680 m 3 . The average net pay (thickness) of the reservoir is 2.68 m. The reservoir was developed by only one well with a total depth of 3000 m. Proved geological reserves of free natural gas are estimated to 15 211 684 m 3 and a recovery factor of around 50% was achieved. Other main reservoir characteristics are given in Table 1. Reservoir natural gas composition is given in Table 2.
The fi rst step in modelling was to characterize the reservoir fl uid in Schlumberger ECLIPSE PVTi. The fl uid composition was imported, and Peng-Robinson (1976) equation of state was applied. The obtained fi le was included in the data fi le for ECLIPSE E300. The model was initialized with a 10×5×1 grid, with the same permeabilities in the x and y-direction and 10 times smaller   permeability in the z-direction. Relative permeabilities were calculated using Corey's exponents. Although certain production decline can be observed for gas reservoir MALM in Figure 3, no clear trend can be approximated by known decline equations. Production varied throughout history with a maximum of 570 000 m 3 /mo. at the beginning and a steep decline towards the end of production history, ending with around 48 000 m 3 /mo. The fi rst and important step in constructing the reservoir model is to calibrate the model against historical production and pressure data. The model must reproduce past reservoir performance accurately before it is used to reliably pre-  Figure 4 presents actual fi eld production and production simulated in ECLIPSE, i.e., history matching is shown. Based on observed past production, history matching was done by gas rate control and with a fl owing bottom hole pressure of 64 bar.

Results and discussion
Production of gas started in October 1984 and it lasted until May 7 th , 1986 and satisfactory history matching was achieved. Simulated bottomhole pressure (BHP) after closing the well was 69.75 bar. Formation pressure dropped down from 160 bar to 82.05 bar. The well was shut from May 8 th , 1986 until September 29 th , 2017. Simulation of CO 2 injection started on September 30 th , 2017 through the same wellbore. The simulation showed that after 1 219 days (3 and a half years) of injection, bottom hole pressure will reach its upper limit and the amount of injected CO 2 will reach its maximum of 16 200 000 m 3 which is 2.191 times more than the produced natural gas. Daily and cumulative CO 2 injection, formation pressure and cumulative gas in place (gas and injected CO 2 ) are given in Figure 5 and Figure 6. The default rate of 27 000 m 3 /day was constant for 14 months after which it started to decrease to reach the formation pressure at 200 bar. The injection of CO 2 lasted for 40 months, during which a total of 16 200 000 m 3 was injected. In Figure 6, cumulative injection and formation pressure show a similar trend. Profi les show CO 2 injection rates and corresponding formation pressure during injection, starting with September 30 th , 2017 and end date of January 31 st , 2021. It should be noted that formation pressure is 199.65 bar and does not exceed the chosen injection bottom hole pressure value of 200 bar.
Injection capacity calculated by Equation 1 was 16 738 479 m 3 . According to Equation 2, available capacity for CO 2 sequestration in reservoir MALM is 26 912 114 kg, which equals to 14 592 444 m 3 .   The relevance of obtained results can be found in confi rmation of analytical solutions validity. The simulation shows that analytical equations are a good fi rst approximation for CO 2 injection capacity of the reservoir. However, Equation 1 depicts higher accuracy compared to Equation 2 since differences between simulated quantities and results obtained by stated formulas are 3% and 10%, respectively, as shown in Figure 7. These accuracies could be confi rmed with actual CO 2 injection in the reservoir in question.

Conclusions
With the increasing need for greenhouse gas mitigation, any attempt of emission avoidance is a case of good industrial practice. That is the main reason why many studies were directed towards CO 2 injection into hydrocarbon reservoirs with subsequent additional recovery. However, those studies considered oil or gas production, which is also a source of CO 2 emissions, while in this research, CO 2 is sequestrated without any of it being emitted. The injection of CO 2 in a small and relatively shallow depleted gas reservoir in Croatia was studied and simulated with ECLIPSE E300. The fi rst step was to get a history match for fi eld production and pressure drop in ECLIPSE E300 and it resulted with high accuracy for production rates, ending with fi eld pressure close to measured. After a conducted history match, a production well is shut and CO 2 injection simulation through it starts. Simulation analysis shows how total injection capacity decreases with an increase in bottom hole pressure during injection. The baseline scenario takes bottom hole pressure of 200 bar, which is 25% higher than the initial reservoir pressure.
The end result of baseline simulation is in accordance with the analytical equations used in this study for the estimation of CO 2 storage capacity. The surface amount of CO 2 that could be injected in the reservoir during the injection period of 40 months is almost two times larger, 16.2 × 10 6 m 3 , than the amount of natural gas, 7.4 × 10 6 m 3 , obtained during a production period of 19 months.
The novelty of this work lies in the fact that a history match was done for a real reservoir, along with CO 2 injection simulation and the fi nal result was compared with two analytical expressions for storage capacity. In the last 15 years, no reproducible study has been carried out in terms of simulation of CO 2 injection into a depleted gas reservoir for Croatian sandstone for the purpose of sequestration only. This study could serve as a benchmark for future CO 2 sequestration in larger depleted gas reservoirs in Croatia and can easily be adjusted for the same types of sandstone in the CPBS. Furthermore, an experimental study should be conducted to verify the equations and simulation results.