CO 2 , Water and N 2 Injection for Enhanced Oil Recovery with Spatial Arrangement of Fractures in Tight-Oil Reservoirs Using Huff-‘ n-puff

Jie Bai 1,2, Huiqing Liu 1,2,*, Jing Wang 1,2, Genbao Qian 3, Yongcan Peng 3, Yang Gao 3, Lin Yan 4 and Fulin Chen 4 1 State Key Laboratory of Petroleum Resources and Engineering, China University of Petroleum-Beijing, Beijing 102249, China; jaybye@163.com (J.B.); wangjing8510@163.com (J.W.) 2 MOE Key Laboratory of Petroleum Engineering in China University of Petroleum, Beijing 102249, China 3 PetroChina Xinjiang Oilfield, Karamay 834000, China; qgenbao@petrochina.com.cn (G.Q.); pyongcan@petrochina.com.cn (Y.P.); gaoyang@petrochina.com.cn (Y.G.) 4 PetroChina Research Institute of Petroleum Exploration & Development, Beijing 100083, China; linyan@petrochina.com.cn (L.Y.); Chenfulin@petrochina.com.cn (F.C.) * Correspondence: liuhq@cup.edu.cn


Introduction
The key dominant feature difference between a tight reservoir and a traditional reservoir lies in the tight pore.The pore throat of the porous medium of the tight reservoir is very narrow, and its movable fluid saturation is very low.The fluid flow requires a higher-pressure gradient through the pore throat (when <50 nm), and the mobility of fluids is poor [1][2][3][4][5][6], so in the development of tight reservoirs, fracturing is a key technology with an indelible contribution to oil production.The main objective of fracturing is to reconstruct the reservoir seepage formation and to achieve the purpose of the effective expansion and development of the fracture through the mechanical mechanism and provide a more efficient and convenient flow channel for the crude oil [2,7].However, according to the evaluation of the oil production of tight reservoirs both in China and America, the production rate of tight reservoirs decreases rapidly, without a long and stable production period [2,6,[8][9][10][11].Thus, it is difficult to obtain satisfactory oil recovery only relying on the depressurization process even with fracturing.
The tight porous reservoir structure is dense, the pore throat is narrow, and at the nanometer scale, and the fluid flow is also affected by the boundary layer effect.This results in higher-pressure gradients in the flooding process in the matrix that restricts traditional flooding application in tight reservoirs [12][13][14], while with the existence of fractures, the breakthrough of gas/water usually occurs during the tight reservoir flooding process, so the huff-'n-puff method is considered a promising energy increasing method in tight reservoirs over the depressurization mode.
Many scholars have developed studies on effective displacement of different fluids in tight reservoirs, mainly including gas injection and chemical injection in Bakken tight oil formations [15,16].Due to the narrow pore throat in tight reservoirs, gas injection, with its lower viscosity and larger injectivity, is considered to be much easier than water injection.CO 2 is one of the most effective injection gases for enhanced oil recovery, and the use of CO 2 to improve the recovery also serves a role of greenhouse gas storage, so it is attracting increasing attention.In the study of the enhanced oil recovery (EOR) by fluid injection, most of the studies focus on the comparison of the miscible process and immiscible process mechanisms [13,14,[17][18][19][20][21][22].
In the CO 2 huff-'n-puff process, CO 2 is first injected into a producer well, then the well is shut off for soaking, finally, the well is reopened for fluids to be produced [23].Its application to unconventional reservoirs has been investigated by lots of researchers recently [18,[24][25][26][27].The traditional EOR mechanisms of CO 2 are: (1) viscosity reduction, (2) oil swelling, (3) solution gas drive, (4) hydrocarbon extraction by CO 2 .The most important operating parameters in CO 2 huff-'n-puff include injection rate and time, number of cycles, soaking time, and pressure.
Hawthorne et al. [28] explained CO 2 EOR process in tight reservoirs relatively clearly: (1) CO 2 is preferentially injected into the fractures, (2) CO 2 contacts the matrix rock through fractures, (3) CO 2 will further seep into the matrix, dominated by pressure gradients.When CO 2 dissolves in crude oil, the oil swells and its viscosity decreases, (4) oil drains into the fractures under the effect of CO 2 , and (5) when the pressure gradient drops, oil transport is dominated by the concentration gradient across the matrix and the fractures.
A considerable number of scholars have also focused on the study of CO 2 in porous media and crude oil mass transfer mechanisms [15,29].The mechanisms of CO 2 EOR in tight oil reservoirs is considered to be significantly different from those in conventional reservoirs, due to the special petrophysical properties, reservoir fluid thermodynamics, and mass transport mechanisms.Diffusion is considered to be an especially important mechanism affecting gas transportat under tight reservoir conditions, and its effect on enhanced oil recovery of tight oil cannot be neglected based on the core scale experiments.
In addition, experimental results from both 2D core slice and 3D homogeneous core scale models show that diffusion plays a significant role [30].However, their field scale huff-'n-puff simulation model showed that convection is considered the dominating mechanism instead of diffusion.Vega carried out CO 2 injection experiments in miscible conditions [31].The results showed that both diffusive and convective transfer mechanisms were significant.In addition to CO 2 , N 2 was also used to explore the EOR potential in tight reservoirs [32].Cores from Eagle Ford were used in this research.Similar experiments can also be found in Li and Sheng's research [33].The results indicated that oil recovery was improved when the soaking pressure increased.Extending soaking time can also significantly improve oil recovery.In addition, increasing huff-'n-puff cycles improved oil recovery.After six cycles, the recovery reached 30.99%.
Different from traditional steam huff-'n-and puff, cold water injection relies on the mechanism of imbibition for oil displacement in water-wet reservoirs in the presence of fractures [34,35].This requires the rock to be water-wet.To alter the rock wettability, surfactants can be used [36][37][38].
In the extensive study of tight reservoirs, it is also found that the fracture formation process in tight reservoirs is complex and diverse.The fracture form (such as the Figure 1) is mainly influenced by the natural fractures and properties of the reservoir (the parameters of brittleness, modulus of elasticity, etc.) and the distribution characteristics of natural fractures [2, 3,7,10,39,40].The expansion and connection of the fractures occurs both in the plane and three-dimensional direction.The diversity in the fracture expansion and connection leads to the fact that the contact between fractures and matrix is a multi-contact relationship [39].This undoubtedly increases the difficulty of evaluating the development effect of tight reservoirs, because the fractures are often simplified to simple uniform and single contact surfaces.In the fluid injection process, especially in huff-'n-puff mode, the mass exchange between crude oil and injected fluid is unavoidably partially ignored due to the lack of multi-contact relationships between fractures and the matrix.Research on the effect of processing under complex fracture conditions has also progressed.In Molero's research, different types of complex fractures were simulated and the effects of CO 2 huff-'n-puff evaluated.The results confirmed that the appropriate modeling of fracture geometry plays a critical role in the estimation of the incremental oil recovery.In Meng's research about imbibition effects, for fractured reservoirs the area of the water-covered-face between the fractures and matrix also had a significant effect on the oil production process.Therefore, the simulation of fractures is essential when evaluating the oil recovery in both depletion and huff-'n-puff processes [38].
increased.Extending soaking time can also significantly improve oil recovery.In addition, increasing huff-'n-puff cycles improved oil recovery.After six cycles, the recovery reached 30.99%.
Different from traditional steam huff-'n-and puff, cold water injection relies on the mechanism of imbibition for oil displacement in water-wet reservoirs in the presence of fractures [34,35].This requires the rock to be water-wet.To alter the rock wettability, surfactants can be used [36][37][38].
In the extensive study of tight reservoirs, it is also found that the fracture formation process in tight reservoirs is complex and diverse.The fracture form (such as the Figure 1) is mainly influenced by the natural fractures and properties of the reservoir (the parameters of brittleness, modulus of elasticity, etc.) and the distribution characteristics of natural fractures [2, 3,7,10,39,40].The expansion and connection of the fractures occurs both in the plane and three-dimensional direction.The diversity in the fracture expansion and connection leads to the fact that the contact between fractures and matrix is a multi-contact relationship [39].This undoubtedly increases the difficulty of evaluating the development effect of tight reservoirs, because the fractures are often simplified to simple uniform and single contact surfaces.In the fluid injection process, especially in huff-'n-puff mode, the mass exchange between crude oil and injected fluid is unavoidably partially ignored due to the lack of multi-contact relationships between fractures and the matrix.Research on the effect of processing under complex fracture conditions has also progressed.In Molero's research, different types of complex fractures were simulated and the effects of CO2 huff-'n-puff evaluated.The results confirmed that the appropriate modeling of fracture geometry plays a critical role in the estimation of the incremental oil recovery.In Meng's research about imbibition effects, for fractured reservoirs the area of the water-covered-face between the fractures and matrix also had a significant effect on the oil production process.Therefore, the simulation of fractures is essential when evaluating the oil recovery in both depletion and huff-'n-puff processes [38].
In terms of core scale experiments, cutting cores and then re-splicing them is the most common method for fracturing simulation.By cutting a complete vertical core with a saw or cutter, the core is separated in half.The fracture is then simulated by splicing the two core parts.The disadvantages of this method are usually a large fracture width (2 mm) and a loss of core and single shape of fractures.The cores also cannot be reused.Tovar et al. [30] simulated fractures by surrounding tight cores with glass beads.To prevent the loss of the glass beads, two sandstone plugs were mounted at both ends as filters.This method doesn't break the cores and keep them reusable.However, the shape of the fracture is still too single, and the whole surrounding space of the core is treated as fracture, so the rationality is questionable.In addition, the spatial arrangement of fractures still cannot be easily simulated in these methods.On the basis of the established importance of complex fractures for tight oil huff-'n-puff processes, this research aims to mechanistically understand how complex fractures impact EOR in huff-'n-puff for different injected fluids.First, indoor core experiments were carried out, and a new In terms of core scale experiments, cutting cores and then re-splicing them is the most common method for fracturing simulation.By cutting a complete vertical core with a saw or cutter, the core is separated in half.The fracture is then simulated by splicing the two core parts.The disadvantages of this method are usually a large fracture width (2 mm) and a loss of core and single shape of fractures.The cores also cannot be reused.Tovar et al. [30] simulated fractures by surrounding tight cores with glass beads.To prevent the loss of the glass beads, two sandstone plugs were mounted at both ends as filters.This method doesn't break the cores and keep them reusable.However, the shape of the fracture is still too single, and the whole surrounding space of the core is treated as fracture, so the rationality is questionable.In addition, the spatial arrangement of fractures still cannot be easily simulated in these methods.
On the basis of the established importance of complex fractures for tight oil huff-'n-puff processes, this research aims to mechanistically understand how complex fractures impact EOR in huff-'n-puff for different injected fluids.First, indoor core experiments were carried out, and a new fracture simulation method using woven metallic wires was proposed.Then a numerical simulation method was used to perform historical fitting of the experimental results with GMG GEM, which fully considers unconventional fluid flow, composition interaction and the mass transfer mechanism mentioned in fractured tight reservoirs.Three different injecting fluids are used, including CO 2 , N 2 and water.Finally, core scale simulation cases with eight different fracture patterns are used in the study to understand how fractures mechanistically impact EOR in huff-'n-puff for different injecting fluids with a numerical method.Through this research, a more regular understanding of the influence of fracture and matrix distribution relationship on EOR process can be clear.It is of great significance for guiding fracturing construction.When selecting the optimal injected fluids corresponding to spatial arrangement fracture, theoretical support can also be obtained from this study.

Experiments
The core plug used in this experiment was from the J Oilfield with a water-wet wettability.Three cores were selected to perform the huff-'n-puff process with CO 2 , water and N 2 .For the experimental system, a degassed crude oil is selected, so the weighing method is adopted to measure the amount of oil produced.Then the oil recovery of each round is calculated.The length is 5 cm with a diameter of 2.5 cm.The permeability is about 0.0375 mD, and porosity is 17.5%.Three cores are selected and their properties are listed in Table 1.In this paper, a new method for the simulation of core-scale fractures was designed.As shown in Figure 2, woven metallic wires attached to the outer surface of the core create a space between the core holder and core as a high permeable zone, equivalent to a fracture.The shape of the metallic cloths can be adjusted to mimic different fracture geometries and relative positions in space.In order to ensure the same fracture area, the same metal meshes were used and the cropped different shaped metal meshes kept at the same weight.The flow chart is shown in Figure 3.
In a specific experimental process, first, we wash, dry and weigh the tight core.The core is saturated with a vacuum saturate.The weight of the core before and after oil saturation is measured separately to determine the weight of oil.Finally, 3.5 mL oil is saturated in the tight core.Then, the core is placed into the core holder.A confining pressure of 30 MPa was applied to the core holder by the hand pump.Then it was saturated with an extra 1.2 mL of oil when initial pressure reached 30 MPa.The viscosity of the oil is 6 mPa•s at 60 • C. The entire system was placed in an incubator at 60 • C. Thus, the core scale tight reservoir system has been simulated using a lab method.
In this paper, a new method for the simulation of core-scale fractures was designed.As shown in Figure 2, woven metallic wires attached to the outer surface of the core create a space between the core holder and core as a high permeable zone, equivalent to a fracture.The shape of the metallic cloths can be adjusted to mimic different fracture geometries and relative positions in space.In order to ensure the same fracture area, the same metal meshes were used and the cropped different shaped metal meshes kept at the same weight.The flow chart is shown in Figure 3.In a specific experimental process, first, we wash, dry and weigh the tight core.The core is saturated with a vacuum saturate.The weight of the core before and after oil saturation is measured separately to determine the weight of oil.Finally, 3.5 mL oil is saturated in the tight core.Then, the core is placed into the core holder.A confining pressure of 30 MPa was applied to the core holder by the hand pump.Then it was saturated with an extra 1.2 mL of oil when initial pressure reached 30 MPa.The viscosity of the oil is 6 mPa•s at 60 °C.The entire system was placed in an incubator at 60 °C.Thus, the core scale tight reservoir system has been simulated using a lab method.
The injection scheme for the huff-'n-puff process is as follows: first, reduce the core pressure from an initial pressure of 30 MPa to 5 MPa.Then, inject CO2/N2/Water until the BHP reaches 40 MPa.The huff-'n-puff is repeated 12 times.The oil produced in the depletion process and CO2/N2 huff-'n-puff processes can be directly weighed, while in water huff-'n-puff, the water should be dried before weighing the produced oil.
The lab method can only provide practical oil recovery data.Subjected to laboratory conditions, the internal pressure and saturation, as well as other properties of fluids, are difficult to measure during a core-scale huff-'n-puff process.Then based on the experimental results, the parameters of relative permeability and fracture permeability are determined by fitting historical experimental data (shown in Figure 4).Finally, the numerical simulation method is used for more detailed analysis and research in the tight core production process.The injection scheme for the huff-'n-puff process is as follows: first, reduce the core pressure from an initial pressure of 30 MPa to 5 MPa.Then, inject CO 2 /N 2 /Water until the BHP reaches 40 MPa.The huff-'n-puff is repeated 12 times.The oil produced in the depletion process and CO 2 /N 2 huff-'n-puff processes can be directly weighed, while in water huff-'n-puff, the water should be dried before weighing the produced oil.
The lab method can only provide practical oil recovery data.Subjected to laboratory conditions, the internal pressure and saturation, as well as other properties of fluids, are difficult to measure during a core-scale huff-'n-puff process.Then based on the experimental results, the parameters of relative permeability and fracture permeability are determined by fitting historical experimental data (shown in Figure 4).Finally, the numerical simulation method is used for more detailed analysis and research in the tight core production process.The lab method can only provide practical oil recovery data.Subjected to laboratory conditions, the internal pressure and saturation, as well as other properties of fluids, are difficult to measure during a core-scale huff-'n-puff process.Then based on the experimental results, the parameters of relative permeability and fracture permeability are determined by fitting historical experimental data (shown in Figure 4).Finally, the numerical simulation method is used for more detailed analysis and research in the tight core production process.

Reservoir Simulation Model
To investigate the several mechanism effect on oil recovery, a Cartesian grid model is built in CMG GEM [41] based on the core experiments.The governing equation describing the mass balance of component i in the oil/gas phases is given by the following expression [15], including the cumulative term of component i in the rock and fluid phases, and the convection, dispersion and molecular diffusion term of component i in phase j: where φ is the porosity of the tight matrix, ρ s is the density of the tight matrix, ρ j is the density the of j phase, S j is the saturation of phase j, w is is the fraction of component i that precipitates in the tight matrix rock, w ij is the fraction of component i in the phase j. r i is the injection or production rate as a source or sink term, N p is the total number of phases, N c is the total number of components, u j is the flow velocity, which is defined based on Darcy's law as: In particular when both oil and water phases exist, the capillary pressure also has effect on the flow velocity, and the equation is described as: where K is the formation permeability, k rj is the relative permeability of phase j, p j is the pressure of phase j, and µ j is the viscosity of phase j.K ij is the dispersivity coefficient of component i in the phase j,which is defined as: where τ is the tortuosity of the tight matrix, D ij is the diffusion coefficient of component i in phase j, and α l is the dispersity coefficient of fluid j in different directions.
D ij (its units are cm 2 /s) can be calculated based on the Sigmund correlation [42].The diffusion in water phase is ignored.The diffusion coefficient D ik is calculated by: where ρ 0 j D 0 ik is the zero-pressure limit of the product of density and diffusivity, which can be calculated by: where T is the absolute temperature (K), σ ik is the collision diameter (Å), Ω ik is the collision integral of the Lennard-Jones Potential, R is the universal gas constant, and M i is the molecular weight of component i.These parameters can be calculated by the following equations [43]: where ω is the acentric factor, P c is the critical pressure (atm), T c is the critical temperature (K), ε is the characteristic Lennard-Jones energy and κ B is the Boltzmann constant: ρ jr is the reduced density, which can be calculated by: V ci is the critical volume of component i, y ij is the mole fraction of component i in phase j.Nc is the total number of components.The diffusion coefficient of component i in the mixture D ij is defined as: In this paper, the core scale reservoir depletion and huff-'n-puff process were simulated with CMG GEM based on the partial differential equation (PDE) of Equation (1).

Reservoir Model Description
The matrix is rectangular with dimensions of 5 × 2.5 × 2.5 cm (Figure 5).This core is divided into 10 × 5 × 5 gridblocks with dimensions of 0.5 cm × 0.5 cm × 0.5 cm.The matrix is assumed to Energies 2019, 12, 823 8 of 31 be homogeneous and isotropic with a porosity of 0.175 and permeability of 0.03 mD.The relative permeability curve inside the matrix is given in Figure 6.Two layers of grid in each direction with dimensions of 0.1 cm × 0.5 cm × 0.5 cm, 0.5 cm × 0.1 cm × 0.5 cm and 0.5 cm × 0.5 cm × 0.1 cm are added to the periphery of the matrix model to serve as the grid space of the fracture, and thus the total number of model grids is 12 × 7 × 7.
The matrix is rectangular with dimensions of 5 × 2.5 × 2.5 cm (Figure 5).This core is divided into 10 × 5 × 5 gridblocks with dimensions of 0.5 cm × 0.5 cm × 0.5 cm.The matrix is assumed to be homogeneous and isotropic with a porosity of 0.175 and permeability of 0.03 mD.The relative permeability curve inside the matrix is given in Figure 6.Two layers of grid in each direction with dimensions of 0.1 cm × 0.5 cm × 0.5 cm, 0.5 cm × 0.1 cm × 0.5 cm and 0.5 cm × 0.5 cm × 0.1 cm are added to the periphery of the matrix model to serve as the grid space of the fracture, and thus the total number of model grids is 12 × 7 × 7.
The fractures are assumed to be thin highly permeable passages covering the surface of the core sample matrix.These assumptions are experimentally possible as shown in Figure 5. Woven metallic wires attached to the outer surface of the core create space between the core holder and core as a high permeability zone, an equivalent fracture.The shape of the metallic cloths can be adjusted to mimic different fracture geometries and relative positions in space.Validating the model with experimental data is critical to ensure that the simulation results are correct.In the history matching work, the main adjustment parameters are the relative permeability curves and the fracture permeability.The porosity of the fractures is assumed to be 0.5 and permeability 50 mD through the history fitting of the experimental data.The relative permeability inside the matrix and the fractures are shown in Figures 6 and 7.
The injected fluids for huff-'n-puff in this paper are CO2, N2 and water.The initial oil and water saturations are 0.8/0.2.The initial pressure is 30MPa.The oil viscosity is 6 mPa•s.The CO2 diffusion coefficient is set based on published simulated and laboratory measurements [15,30,44,45].Other initial reservoir properties can be found in Table 2.The phase behavior of water/oil, water/oil/CO2, and water/oil/N2 are modeled with the Peng-Robinson equation of state.Inputs for the EOS model is given in Tables 3 and 4. The initial oil components are provided by the J field, and shown in Table 5.The well is located at (12,4,4) as indicated in Figure 5.The injection scheme for the huff-n-puff process are as follows: first, depletion and pressure to 5 MPa.Then, inject CO2/N2/Water until BHP reaching 40 MPa.Huff-'n-puff is repeated 12 times.
The relative arrangement of matrix and several fractures are modeled on a core scale with CMG GEM.Table 6 lists the core-scale tight oil-reservoirs with the same contact area between fractures and cores, but with different fracture spatial arrangements.The fractures are assumed to be thin highly permeable passages covering the surface of the core sample matrix.These assumptions are experimentally possible as shown in Figure 5. Woven metallic wires attached to the outer surface of the core create space between the core holder and core as a high permeability zone, an equivalent fracture.The shape of the metallic cloths can be adjusted to mimic different fracture geometries and relative positions in space.
Validating the model with experimental data is critical to ensure that the simulation results are correct.In the history matching work, the main adjustment parameters are the relative permeability curves and the fracture permeability.The porosity of the fractures is assumed to be 0.5 and permeability 50 mD through the history fitting of the experimental data.The relative permeability inside the matrix and the fractures are shown in Figures 6 and 7.
The injected fluids for huff-'n-puff in this paper are CO 2 , N 2 and water.The initial oil and water saturations are 0.8/0.2.The initial pressure is 30MPa.The oil viscosity is 6 mPa•s.The CO 2 diffusion coefficient is set based on published simulated and laboratory measurements [15,30,44,45].Other initial reservoir properties can be found in Table 2.The phase behavior of water/oil, water/oil/CO 2 , and water/oil/N 2 are modeled with the Peng-Robinson equation of state.Inputs for the EOS model is given in Tables 3 and 4. The initial oil components are provided by the J field, and shown in Table 5.The well is located at (12,4,4)     The relative arrangement of matrix and several fractures are modeled on a core scale with CMG GEM.Table 6 lists the core-scale tight oil-reservoirs with the same contact area between fractures and cores, but with different fracture spatial arrangements.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.

Depletion Mode
By comparing the simulation results of the different fracture mode cases (Figure 8), it is found that in the depressurization period, the existence of fractures is significant for the enhanced oil process.With the same contact area between fracture and matrix, the cases with fractures have approximately 18% higher oil recovery than the case without any fracture.Figure 9 shows the pressure distribution in several cases.Without fractures, the pressure drop is not obvious.The pressure change area in the cases with fracture is wider and the pressure drop is more significant, especially in Case-4.This shows that the fractures supply the more efficient and convenient channels and contact relationships.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.In the depletion mode, crude oil flows mainly by the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.
Case-4 represents the most typical contact relationship between horizontal well fracturing artificial fracture and tight matrix without natural fracture.Case-4 is consistent with the experimental scheme.Figure 9 shows that the fracture type in Case-4 has the best pressure spread effect.In the next study, Case-4 is used as the basis for comparing different fluid using huff-'n-puff.Case-4 represents the most typical contact relationship between horizontal well fracturing artificial fracture and tight matrix without natural fracture.Case-4 is consistent with the experimental scheme.Figure 9 shows that the fracture type in Case-4 has the best pressure spread effect.In the next study, Case-4 is used as the basis for comparing different fluid using huff-'n-puff.

Oil recovery comparison
In this part, Case-4 is used to illustrate the mechanisms of using CO2, N2 and water for the huff-'n-puff process.Figure 10 shows the oil recovery factor on surface conditions by using the three different injecting fluids.CO2 was able to enhance oil recovery by 21% after depletion, water by 13%, and N2 by 10% so CO2 has obvious advantages over water and N2.


Gas/liquid injection volume Firstly, injection fluid volume under the reservoir conditions was selected as a comparison index for the injectivity (Figure 11) to evaluate the EOR effect of different fluids.
From the comparison it can be found that the injectivity of N2 is the highest among the three fluids, but its impact on EOR is the least.CO2 has the secondary injectivity with the best EOR effect.The injected volume of water is the least, but with a moderate oil recovery.

• Oil recovery comparison
In this part, Case-4 is used to illustrate the mechanisms of using CO 2 , N 2 and water for the huff-'n-puff process.Figure 10 shows the oil recovery factor on surface conditions by using the three different injecting fluids.CO 2 was able to enhance oil recovery by 21% after depletion, water by 13%, and N 2 by 10% so CO 2 has obvious advantages over water and N 2 .Case-4 represents the most typical contact relationship between horizontal well fracturing artificial fracture and tight matrix without natural fracture.Case-4 is consistent with the experimental scheme.Figure 9 shows that the fracture type in Case-4 has the best pressure spread effect.In the next study, Case-4 is used as the basis for comparing different fluid using huff-'n-puff.

Oil recovery comparison
In this part, Case-4 is used to illustrate the mechanisms of using CO2, N2 and water for the huff-'n-puff process.Figure 10 shows the oil recovery factor on surface conditions by using the three different injecting fluids.CO2 was able to enhance oil recovery by 21% after depletion, water by 13%, and N2 by 10% so CO2 has obvious advantages over water and N2.


Gas/liquid injection volume Firstly, injection fluid volume under the reservoir conditions was selected as a comparison index for the injectivity (Figure 11) to evaluate the EOR effect of different fluids.
From the comparison it can be found that the injectivity of N2 is the highest among the three fluids, but its impact on EOR is the least.CO2 has the secondary injectivity with the best EOR effect.The injected volume of water is the least, but with a moderate oil recovery.• Gas/liquid injection volume Firstly, injection fluid volume under the reservoir conditions was selected as a comparison index for the injectivity (Figure 11) to evaluate the EOR effect of different fluids.
From the comparison it can be found that the injectivity of N 2 is the highest among the three fluids, but its impact on EOR is the least.CO 2 has the secondary injectivity with the best EOR effect.The injected volume of water is the least, but with a moderate oil recovery.Then, we also found the relationship between the volume of the accumulative injection of fluid and the oil recovery under the reservoir conditions of different fluids (CO2, N2 and water) cannot be satisfactorily correlated (shown in Figure 11a).It shows that that the injectivity is only part of the basis for the EOR effect of fluids.Because of the obvious difference of EOR mechanism among the three fluids, the oil recovery only shows a good linear relationship with cumulative injected volume under reservoir conditions in same fluid cases (shown in Figures 11b,c,d).
The following is a detailed comparison of EOR processes of different fluids, to analyze how distributions of the fractures mechanistically affect the performance of the three different injectants:

Viscosity profile and CO2 concentration in oil phase
To further understand the EOR mechanisms of the three different injecting fluids, the influential factors in oil recovery are examined.One of the most important factors is the viscosity of the oil phase, which is affected by the thermodynamic conditions and oil composition.Figure 12 shows the viscosity of oil phase during the soaking process in the 12th cycle by using the three different fluids.It can be found that the effect of CO2 on crude oil viscosity reduction is very significant, especially the viscosity of crude oil near the well and near the interface of fractures has the most significant change from the initial 6 mPa•s to about 1 mPa•s.However, N2 is also mixed as a gas during the contact with crude oil, but there is no obvious viscosity reduction effect.After the water injection, the viscosity of crude oil may even increase to 7.5 mPa•s.
CO2 is only fluid that is miscible with oil (even at 90% mole fraction with CO2 in the oil phase as shown in Figures 13-15).The solubility of CO2 results in the lowest oil phase viscosity of the three cases.In contrast, N2 mainly exists in the gas phase and is much less soluble in the oil phase compared to CO2.Water does not dissolve in the oil and water mainly enhances oil recovery through capillary force, gravity and elasticity displacement.Then, we also found the relationship between the volume of the accumulative injection of fluid and the oil recovery under the reservoir conditions of different fluids (CO 2 , N 2 and water) cannot be satisfactorily correlated (shown in Figure 11a).It shows that that the injectivity is only part of the basis for the EOR effect of fluids.Because of the obvious difference of EOR mechanism among the three fluids, the oil recovery only shows a good linear relationship with cumulative injected volume under reservoir conditions in same fluid cases (shown in Figure 11b-d).
The following is a detailed comparison of EOR processes of different fluids, to analyze how distributions of the fractures mechanistically affect the performance of the three different injectants:

• Viscosity profile and CO 2 concentration in oil phase
To further understand the EOR mechanisms of the three different injecting fluids, the influential factors in oil recovery are examined.One of the most important factors is the viscosity of the oil phase, which is affected by the thermodynamic conditions and oil composition.Figure 12 shows the viscosity of oil phase during the soaking process in the 12th cycle by using the three different fluids.It can be found that the effect of CO 2 on crude oil viscosity reduction is very significant, especially the viscosity of crude oil near the well and near the interface of fractures has the most significant change from the initial 6 mPa•s to about 1 mPa•s.However, N 2 is also mixed as a gas during the contact with crude oil, but there is no obvious viscosity reduction effect.After the water injection, the viscosity of crude oil may even increase to 7.5 mPa•s.

Fracture Contact
No Fracture Contact Figure 16 shows the produced components of all three cases.CO2 gives the most incremental C19 and C30 compared to depletion.However, in the water case, more C7 are produced, leaving a deprived heavy mass that causes an increase in the viscosity of the crude oil.N2 has the least incremental effect on oil components in all carbon number ranges.CO 2 is only fluid that is miscible with oil (even at 90% mole fraction with CO 2 in the oil phase as shown in Figures 13-15).The solubility of CO 2 results in the lowest oil phase viscosity of the three cases.In contrast, N 2 mainly exists in the gas phase and is much less soluble in the oil phase compared to CO 2 .Water does not dissolve in the oil and water mainly enhances oil recovery through capillary force, gravity and elasticity displacement.Figure 16 shows the produced components of all three cases.CO2 gives the most incremental C19 and C30 compared to depletion.However, in the water case, more C7 are produced, leaving a deprived heavy mass that causes an increase in the viscosity of the crude oil.N2 has the least incremental effect on oil components in all carbon number ranges.Figure 16 shows the produced components of all three cases.CO 2 gives the most incremental C19 and C30 compared to depletion.However, in the water case, more C7 are produced, leaving a deprived heavy mass that causes an increase in the viscosity of the crude oil.N 2 has the least incremental effect on oil components in all carbon number ranges.

• Oil swelling
The mixture of gas and crude oil will cause a volume expansion of crude oil. Figure 17 shows that the volume expansion of crude oil under reservoir conditions is most obvious after CO 2 injection, and N 2 has only a slight expansion effect on oil volume.When injecting water, the oil is compressed and its volume in reservoir condition decreased.Based on the expansion mechanism, CO 2 showed much better displacement performance than N 2 and water.

Oil swelling
The mixture of gas and crude oil will cause a volume expansion of crude oil. Figure 17 shows that the volume expansion of crude oil under reservoir conditions is most obvious after CO2 injection, and N2 has only a slight expansion effect on oil volume.When injecting water, the oil is compressed and its volume in reservoir condition decreased.Based on the expansion mechanism, CO2 showed much better displacement performance than N2 and water.

 Saturation profile and relative permeability profile
Another influential factor is the relative permeability of the oil phase.Figure 18 shows the relative permeability of the oil phase in the same production process in the 12th cycle.Figure 19 shows the oil, water and gas saturations at the same time.It can be seen with similar oil saturation, the relative oil relative permeability in the CO2 and water case is much higher than the N2 case.This is a consequence of the much lower relative permeability to liquid with the existence of gas phase.Based on the gas/liquid relative permeability curve, when two-phase flow occurs, the gas flows more easily than the liquid.Because the gas-liquid two-phase flow area is obviously narrower than the oil-water two-phase flow area (Figures 6 and 7).

Fracture Contact
No Fracture Contact  • Saturation profile and relative permeability profile Another influential factor is the relative permeability of the oil phase.Figure 18 shows the relative permeability of the oil phase in the same production process in the 12th cycle.Figure 19 shows the oil, water and gas saturations at the same time.It can be seen with similar oil saturation, the relative oil relative permeability in the CO 2 and water case is much higher than the N 2 case.This is a consequence of the much lower relative permeability to liquid with the existence of gas phase.Based on the gas/liquid relative permeability curve, when two-phase flow occurs, the gas flows more easily than the liquid.Because the gas-liquid two-phase flow area is obviously narrower than the oil-water two-phase flow area (Figures 6 and 7).
From the saturation distribution of production stage in Figure 19, it can be seen that for gas huff-'n-puff, oil saturation in some areas is even higher than that in water huff-'n-puff.In water huff-'n-puff, because of the wider two-phase flow area of water and oil, the overall saturation decline area is larger.However, influenced by comprehensive mechanisms, water's EOR effect is still not as good as CO 2 with viscosity reduction mechanism, but better than N 2 .
shows the oil, water and gas saturations at the same time.It can be seen with similar oil saturation, the relative oil relative permeability in the CO2 and water case is much higher than the N2 case.This is a consequence of the much lower relative permeability to liquid with the existence of gas phase.Based on the gas/liquid relative permeability curve, when two-phase flow occurs, the gas flows more easily than the liquid.Because the gas-liquid two-phase flow area is obviously narrower than the oil-water two-phase flow area (Figures 6 and 7).From the saturation distribution of production stage in Figure 19, it can be seen that for gas huff-'n-puff, oil saturation in some areas is even higher than that in water huff-'n-puff.In water huff-'n-puff, because of the wider two-phase flow area of water and oil, the overall saturation decline area is larger.However, influenced by comprehensive mechanisms, water's EOR effect is still not as good as CO2 with viscosity reduction mechanism, but better than N2.

 CO2 diffusion
Figure 20 shows that the oil recovery considering the CO2 diffusion effect at 0.005 cm 2 /s is 1.42% higher than the cases without considering it.Figure 21 compares the CO2 global mole fraction distribution profile with and without CO2 diffusion.From the profile, the matrix grids have higher CO2 mole fraction with consideration of CO2 diffusion, which indicates that more CO2 may enter into the tight matrix from fractures to mix with the crude oil.Therefore, in order to accurately enhance the oil recovery effect of CO2 in the tight reservoir simulation model, CO2 diffusion effects cannot be ignored.On the basis of CO2 diffusion to improve CO2 EOR, a reasonable value of diffusion coefficient is critical for the reasonable evaluation of oil recovery in CO2 huff-'n-puff.• CO 2 diffusion Figure 20 shows that the oil recovery considering the CO 2 diffusion effect at 0.005 cm 2 /s is 1.42% higher than the cases without considering it.Figure 21 compares the CO 2 global mole fraction distribution profile with and without CO 2 diffusion.From the profile, the matrix grids have higher CO 2 mole fraction with consideration of CO 2 diffusion, which indicates that more CO 2 may enter into the tight matrix from fractures to mix with the crude oil.Therefore, in order to accurately enhance the oil recovery effect of CO 2 in the tight reservoir simulation model, CO 2 diffusion effects cannot be ignored.On the basis of CO 2 diffusion to improve CO 2 EOR, a reasonable value of diffusion coefficient is critical for the reasonable evaluation of oil recovery in CO 2 huff-'n-puff.

 CO2 diffusion
Figure 20 shows that the oil recovery considering the CO2 diffusion effect at 0.005 cm 2 /s is 1.42% higher than the cases without considering it.Figure 21 compares the CO2 global mole fraction distribution profile with and without CO2 diffusion.From the profile, the matrix grids have higher CO2 mole fraction with consideration of CO2 diffusion, which indicates that more CO2 may enter into the tight matrix from fractures to mix with the crude oil.Therefore, in order to accurately enhance the oil recovery effect of CO2 in the tight reservoir simulation model, CO2 diffusion effects cannot be ignored.On the basis of CO2 diffusion to improve CO2 EOR, a reasonable value of diffusion coefficient is critical for the reasonable evaluation of oil recovery in CO2 huff-'n-puff.

 Capillary imbibition
According to Meng's research, in the presence of fractures the capillary force can play a role of imbibition, which can also promote the drainage process [39].In comparison (Figure 22), oil recovery considering the imbibition effect is 1.3% higher than cases without considering it.From the water saturation profile (Figure 23) the matrix grids have higher water saturation with consideration of imbibition, which indicates that more water in the fracture may enter the matrix, and then impact the oil drainage.

• Capillary imbibition
According to Meng's research, in the presence of fractures the capillary force can play a role of imbibition, which can also promote the drainage process [39].In comparison (Figure 22), oil recovery considering the imbibition effect is 1.3% higher than cases without considering it.From the water saturation profile (Figure 23) the matrix grids have higher water saturation with consideration of imbibition, which indicates that more water in the fracture may enter the matrix, and then impact the oil drainage.

 Capillary imbibition
According to Meng's research, in the presence of fractures the capillary force can play a role of imbibition, which can also promote the drainage process [39].In comparison (Figure 22), oil recovery considering the imbibition effect is 1.3% higher than cases without considering it.From the water saturation profile (Figure 23) the matrix grids have higher water saturation with consideration of imbibition, which indicates that more water in the fracture may enter the matrix, and then impact the oil drainage.

 Gravity
Another observation of the water saturation profiles is that water can be imbibed and is subject to density differences in the water case, which can displace more oil towards the fracture.Water is denser than oil, and N2 is less dense than oil.Aggregation of different phase during the soaking period may affect the performance of huff-'n-puff as well.By setting the Kv = 1, the permeability anisotropy is eliminated, and the difference between Case-4 and Case-7 is dominated by gravity effects.It can be seen from Figure 24 that the gravity effect is almost negligible in the depletion mode and the gas huff-'n-puff mode.In water huff-'n-puff, the oil recovery finally has a difference of 0.95% in Case-4 and Case-7.By comparing the oil saturation profiles of different layers (Figure • Gravity Another observation of the water saturation profiles is that water can be imbibed and is subject to density differences in the water case, which can displace more oil towards the fracture.Water is denser than oil, and N 2 is less dense than oil.Aggregation of different phase during the soaking period may affect the performance of huff-'n-puff as well.By setting the Kv = 1, the permeability anisotropy is eliminated, and the difference between Case-4 and Case-7 is dominated by gravity effects.It can be seen from Figure 24 that the gravity effect is almost negligible in the depletion mode and the gas huff-'n-puff mode.In water huff-'n-puff, the oil recovery finally has a difference of 0.95% in Case-4 and Case-7.By comparing the oil saturation profiles of different layers (Figure 25), there are slight differences between top and bottom layers in CO 2 and N 2 huff-'n-puff, but the differences are not significant.The difference of saturation between top and bottom layers is more obvious in water huff-'n-puff.Thus, the gravity effect can be neglected in core scale model.Whether there is a significant impact on large size will be verified and discussed in the authors' later studies.
Energies 2019, 12, 823 19 of 29 25), there are slight differences between top and bottom layers in CO2 and N2 huff-'n-puff, but the differences are not significant.The difference of saturation between top and bottom layers is more obvious in water huff-'n-puff.Thus, the gravity effect can be neglected in core scale model.Whether there is a significant impact on large size will be verified and discussed in the authors' later studies.

Fracture Characteristics
Now that the fundamental mechanisms of huff-'n-puff by using CO2, N2 and water are clear, it is of interest to understand how the distribution of fractures affects their performance in huff-'n-puff.For different fracture characteristics, the direction of scattered fracture, fracture in end face or lateral face, Scattered pattern of fractures, fracture relative location with well and fracture area are designed to mechanistically understand the impact of fracture characteristics on huff-'n-puff.

Fracture Characteristics
Now that the fundamental mechanisms of huff-'n-puff by using CO 2 , N 2 and water are clear, it is of interest to understand how the distribution of fractures affects their performance in huff-'n-puff.For different fracture characteristics, the direction of scattered fracture, fracture in end face or lateral face, Scattered pattern of fractures, fracture relative location with well and fracture area are designed to mechanistically understand the impact of fracture characteristics on huff-'n-puff.

Scattered Pattern of Fractures
It is found that the dispersion characteristics of fractures has little effect on the recovery in N 2 and water huff-'n-puff.However, it has a great influence on the EOR effect in CO 2 case (Figure 26).When designing a fracture contact relationship, we compared four scattered fracture (Cases-1-4).It is found that the dispersion characteristics of fractures has little effect on the recovery in N2 and water huff-'n-puff.However, it has a great influence on the EOR effect in CO2 case (Figure 26).When designing a fracture contact relationship, we compared four scattered fracture (Cases-1-4).
In the depletion development mode, the final oil recoveries are not very different.In CO2 huff-'n-puff, the oil recovery in Case-4 is the highest.The EOR effect in Case-1 is the worst.It can be seen from the comparison of CO2 that the more scattered and farther in distance, the better the recovery effect of CO2 huff-'n-puff.In N2 huff-'n-puff cases, Case-3 and Case-4 are the best.In water huff-'n-puff cases, Case-4 is the best, followed by Case-3.By comparing the oil viscosity distribution in different fractures (Figure 27), the fracture type in Case-4 can result in much less oil viscosity in the soaking process, and hence more oil can be produced.Therefore, the oil saturation is much lower in the Case-4 profile (Figure 28).In the depletion development mode, the final oil recoveries are not very different.In CO 2 huff-'n-puff, the oil recovery in Case-4 is the highest.The EOR effect in Case-1 is the worst.It can be seen from the comparison of CO 2 that the more scattered and farther in distance, the better the recovery effect of CO 2 huff-'n-puff.In N 2 huff-'n-puff cases, Case-3 and Case-4 are the best.In water huff-'n-puff cases, Case-4 is the best, followed by Case-3.
By comparing the oil viscosity distribution in different fractures (Figure 27), the fracture type in Case-4 can result in much less oil viscosity in the soaking process, and hence more oil can be produced.Therefore, the oil saturation is much lower in the Case-4 profile (Figure 28).By comparing the oil viscosity distribution in different fractures (Figure 27), the fracture type in Case-4 can result in much less oil viscosity in the soaking process, and hence more oil can be produced.Therefore, the oil saturation is much lower in the Case-4 profile (Figure 28).3.2.2.Direction of Scattered Fracture (with Kv = 0.2) In this section, Case-4 and Case-7 are used to illustrate the mechanisms of permeability anisotropy by setting fractures in different directions.Figure 29 shows the oil recovery factor effect on surface conditions by using three different injecting fluids with fractures in different directions.The conclusion of this comparison is that permeability anisotropy plays a significant role in the water case (Case-7 is 4.38% higher than Case-4; but gravity has a negligible impact on CO2 and N2).The cross section of Cases-4 and -7 in k-j direction is specifically selected for analysis (Figures 30 and  31).From the viscosity comparison (Figure 30), the viscosity reduction area in Case-4 is larger than the area in Case-7.In this section, Case-4 and Case-7 are used to illustrate the mechanisms of permeability anisotropy by setting fractures in different directions.Figure 29 shows the oil recovery factor effect on surface conditions by using three different injecting fluids with fractures in different directions.The conclusion of this comparison is that permeability anisotropy plays a significant role in the water case (Case-7 is 4.38% higher than Case-4; but gravity has a negligible impact on CO 2 and N 2 ).The cross section of Cases-4 and -7 in k-j direction is specifically selected for analysis (Figures 30 and 31).From the viscosity comparison (Figure 30), the viscosity reduction area in Case-4 is larger than the area in Case-7.
on surface conditions by using three different injecting fluids with fractures in different directions.The conclusion of this comparison is that permeability anisotropy plays a significant role in the water case (Case-7 is 4.38% higher than Case-4; but gravity has a negligible impact on CO2 and N2).The cross section of Cases-4 and -7 in k-j direction is specifically selected for analysis (Figures 30 and  31).From the viscosity comparison (Figure 30), the viscosity reduction area in Case-4 is larger than the area in Case-7.From the saturation profiles (Figure 31), it can be seen that in Case-4, more water can enter the matrix grid, but in Case-7 less oil can be displaced by water in this type fracture.As in the Kv and gravity effect, in Case-4, the oil recovery is significantly better than in Case-7.The fracture patterns in Cases-4 and -7 represent the typical horizontal well fracture and vertical well fracture.When it is extended to engineering applications, it can be considered that the horizontal well fractures are more effective than the vertical well fracture, and the horizontal fracture well is more suitable for water huff-n-puff.CO 2 huff-n-puff can be used both in horizontal and vertical fracture wells.

Case-4
Case-7    From the saturation profiles (Figure 31), it can be seen that in Case-4, more water can enter the matrix grid, but in Case-7 less oil can be displaced by water in this type fracture.As in the Kv and

Fracture Near Well and Far Well
Due to the high conductivity of the fracture, the crude oil can flow into the well through the fractures, and the difference of oil recovery between the near-well and the far-well cases is small (Figure 32).
Due to the high conductivity of the fracture, the crude oil can flow into the well through the fractures, and the difference of oil recovery between the near-well and the far-well cases is small (Figure 32).
Comparing the EOR effects of near-well and far-well fracture in different development modes (Cases-5,6), it can be found that the results are not consistent either.For the depletion period, the far-well fracture (in Case-6) has an advantage over the near-well case (Case-5) by 1% higher oil recovery.
In the CO2 and N2 huff-''n-puff periods, the oil recovery of the two fracture types are close to each other.In water huff-'n-puff, the fracture in Case-5 has a higher incremental oil content.In the fourth round, the difference in oil recovery can even reach 3.1%, but in the later rounds, the difference is less than 0.74%.

Fracture Contact Area
Comparing the relationship between fracture contact area and oil recovery (Figure 33), it can be found that in the depletion period, the larger the contact area of the fracture, the higher the oil recovery, but when the contact area is greater than 25, the increase of the oil recovery is obviously slowed down.In the CO2 huff-'n-puff, oil recovery is the highest when the contact area is 25.In N2 and water huff-'n-puff, there is still not a clear trend between the contact area and oil recovery.Therefore, the contact area is not sensitive for the huff-'n-puff process based on this research.Comparing the EOR effects of near-well and far-well fracture in different development modes (Cases-5,6), it can be found that the results are not consistent either.For the depletion period, the far-well fracture (in Case-6) has an advantage over the near-well case (Case-5) by 1% higher oil recovery.
In the CO 2 and N 2 huff-''n-puff periods, the oil recovery of the two fracture types are close to each other.In water huff-'n-puff, the fracture in Case-5 has a higher incremental oil content.In the fourth round, the difference in oil recovery can even reach 3.1%, but in the later rounds, the difference is less than 0.74%.

Fracture Contact Area
Comparing the relationship between fracture contact area and oil recovery (Figure 33), it can be found that in the depletion period, the larger the contact area of the fracture, the higher the oil recovery, but when the contact area is greater than 25, the increase of the oil recovery is obviously slowed down.In the CO 2 huff-'n-puff, oil recovery is the highest when the contact area is 25.In N 2 and water huff-'n-puff, there is still not a clear trend between the contact area and oil recovery.Therefore, the contact area is not sensitive for the huff-'n-puff process based on this research.

Components Analysis
Through the previous comparison of oil recovery, viscosity, phase permeability and mobility, the EOR effect of different fluids in different fracture can be examined and compared macroscopically.In order to analyze the microscopic differences in more detail, we also selected Cases-4, 5 and 8 as the comparison experiments, adding the analysis of EOR produced components of different fluids in different fracture modes.
As can be seen from the figure (Figure 34), the mainly produced crude oil components include C7, C19 and C30.The proportion of C1 and C4 components in crude oil is originally small, and their production amount did not differ significantly between different fractures.First, for C7 analysis, the most component C7 can be produced in water huff-'n-puff, followed by CO 2 , with the least amount of C7 produced during N 2 huff-'n-puff.For C19 and C30, CO 2 has a better drainage effect on them, followed by water, and N 2 is still the worst.This shows that different fluids have different drainage effects on different oil components.More interestingly, the fracture condition also has an effect on the composition of the components produced.Taking C19 as an example, C19 can be produced much easier in Case-5 in CO 2 huff-'n-puff.For water, the fracture in Case-8 is more conducive to the production of C19.This indicates that the fracture distribution pattern also has a non-negligible effect on the EOR effect of different fluids.

Components Analysis
Through the previous comparison of oil recovery, viscosity, phase permeability and mobility, the EOR effect of different fluids in different fracture can be examined and compared macroscopically.In order to analyze the microscopic differences in more detail, we also selected Cases-4, 5 and 8 as the comparison experiments, adding the analysis of EOR produced components of different fluids in different fracture modes.

Components Analysis
Through the previous comparison of oil recovery, viscosity, phase permeability and mobility, the EOR effect of different fluids in different fracture can be examined and compared macroscopically.In order to analyze the microscopic differences in more detail, we also selected Cases-4, 5 and 8 as the comparison experiments, adding the analysis of EOR produced components of different fluids in different fracture modes.As can be seen from the figure (Figure 34), the mainly produced crude oil components include C7, C19 and C30.The proportion of C1 and C4 components in crude oil is originally small, and their Finally, it should be noted that the main EOR effect is still due to the fluid.Fractures only provide a more efficient and convenient channel and contact relationship for fluids to exert their EOR effects.Finally, in the fluid selection of tight reservoir EOR huff-'n-puff, it is necessary to accurately and reasonably evaluate the distribution characteristics of fractures, and select the optimal injected fluid for specific fracture morphology.

Conclusions
This paper takes a fractured tight reservoir as the basic research object.Natural tight cores from the J field in China are used to conduct experimental studies on different fluid huff-'n-puff processes.A new experimental core-level fracture simulation method that is easy to operate and reusable is proposed.Three different injecting fluids are used, including CO 2 , N 2 and water.The equivalent core scale reservoir numerical models in depletion and huff-'n-puff mode are then restored by numerical simulation with CMG GEM.Simulation cases with eight different fracture patterns were used in the detail study to understand how fracture mechanistically impact EOR in huff and puff for different injecting fluids including CO 2 , N 2 and water.Finally, the key findings can be summarized as follows: (1) In core-scale fractures, a new experimental core-level fracture simulation method that is easy to operate and reusable is proposed.In this method, woven metallic wires are attached to the outer surface of the core to create a space between the core holder and the core as a high permeability zone, equivalent to a fracture.This avoids the defects of single fracture shape, sanding, and difficulty in reusing the core.Different spatial arrangement of fractures can be set up in core-scale experiments.
(2) Based on the core-scale experiments and numerical simulation analyses, the existence of fractures is significant for the enhanced oil process in tight reservoirs.The presence of fractures not only has a significant effect on the depletion mode, but also has an important effect on the enhanced oil recovery (EOR) performance of the huff-'n-puff method.After comparison, fracture arrangement in space has most impact on CO 2 huff-'n-puff, followed by water and finally N 2 .(3) In the depletion mode, crude oil flows mainly due to the pressure drop caused by elastic energy, and there is no contact between other fluids and crude oil in the matrix through fractures.Therefore, in the case of depletion mode, when fractures have the same contact area with spatial arrangement, the obvious difference in oil recovery cannot be observed.(4) CO 2 has more advantages over water and N 2 in tight reservoir with huff-'n-puff.Through the EOR mechanism analysis, CO 2 is the only fluid that is miscible with oil (even a 90% mole fraction with CO 2 in the oil phase is possible).This solubility of CO 2 results in the lowest oil phase viscosity in the three cases.The CO 2 diffusion mechanism is pronounced.In contrast, N 2 mainly exists in the gas phase and is much less soluble in the oil phase compared to CO 2 , however, it still has a certain swelling effect on crude oil.Water does not dissolve in the oil and water mainly enhances oil recovery through capillary force, gravity and elasticity displacement.(5) When considering the anisotropy, the direction of fracture will have a greater impact on oil recovery for the water huff-'n-puff.The conclusion from this comparison is that permeability anisotropy plays a significant role in the water case (Case-7 is 4.38% higher than Case-4; but anisotropy has a negligible impact on CO 2 and N 2 ).( 6) When it comes to the relative location between well and fracture, for the depletion development mode, the far-well fracture has an advantage over the near well case by 1% higher oil recovery.
In the CO 2 and N 2 huff-'n-puff modes, the oil recovery of the two fracture types are not much different.In water huff-'n-puff, the fractures near the well have a better effect on the EOR process.(7) The fracture contact area is not sensitive for the huff-'n-puff process for N 2 and water cases based on this research.(8) From analysis of the produced components, CO 2 gives the most incremental C19 and C30 compared to depletion.However, in the water case, more C7 are produced, leaving a depleted

Figure 1 .
Figure 1.Nonplanar fracture model in a naturally fractured formation.

Figure 1 .
Figure 1.Nonplanar fracture model in a naturally fractured formation.

Figure 4 .
Figure 4. Comparisons of oil recovery between lab data and huff-'n-puff simulation.Figure 4. Comparisons of oil recovery between lab data and huff-'n-puff simulation.

Figure 4 .
Figure 4. Comparisons of oil recovery between lab data and huff-'n-puff simulation.Figure 4. Comparisons of oil recovery between lab data and huff-'n-puff simulation.

Figure 7 .
Figure 7. Relative permeability curve of the fracture.

Figure 6 .
Figure 6.Relative permeability curve of the matrix.
as indicated in Figure5.The injection scheme for the huff-n-puff process are as follows: first, depletion and pressure to 5 MPa.Then, inject CO 2 /N 2 /Water until BHP reaching 40 MPa.Huff-'n-puff is repeated 12 times.

Figure 6 .
Figure 6.Relative permeability curve of the matrix.

Figure 7 .
Figure 7. Relative permeability curve of the fracture.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 8 .
Figure 8.The comparison of oil recovery in fracture and no-fracture conditions.

Figure 9 .
Figure 9. Pressure comparison between different scattered fractures in the depletion process.

Figure 9 .
Figure 9. Pressure comparison between different scattered fractures in the depletion process.

6 Figure 9 .
Figure 9. Pressure comparison between different scattered fractures in the depletion process.

Figure 11 .
Figure 11.Relationship between injection volume RC and oil recovery in huff-'n-puff mode.

Figure 11 .
Figure 11.Relationship between injection volume RC and oil recovery in huff-'n-puff mode.

Figure 12 .
Figure 12.Oil viscosity of different fluids in Case-4 in soaking process (ij profile).

Figure 13 .
Figure 13.Mole fraction of injection fluid in oil in Case-4 in soaking process (ij profile).

Figure 14 .N2Figure 15 .
Figure 14.Mole fraction of injection fluid in oil in Case-4 in soaking process (ij profile).Fracture Contact (ik)No Fracture Contact (ik)

Figure 15 .
Figure 15.Mole fraction of injection fluid in gas in Case-4 in soaking process (ij profile).

N2Figure 15 .
Figure 15.Mole fraction of injection fluid in gas in Case-4 in soaking process (ij profile).

Figure 17 .
Figure 17.Oil volume in reservoir conditions comparison in Case-4.

Figure 17 .
Figure 17.Oil volume in reservoir conditions comparison in Case-4.

Figure 18 .
Figure 18.Oil relative permeability of different fluids in Case-4 in the production process (ij profile).Figure 18.Oil relative permeability of different fluids in Case-4 in the production process (ij profile).

Figure 18 .
Figure 18.Oil relative permeability of different fluids in Case-4 in the production process (ij profile).Figure 18.Oil relative permeability of different fluids in Case-4 in the production process (ij profile).

Figure 19 .
Figure 19.Oil saturation of different fluids in Case-4 in production process (ij profile).

Figure 19 .
Figure 19.Oil saturation of different fluids in Case-4 in production process (ij profile).

Figure 19 .
Figure 19.Oil saturation of different fluids in Case-4 in production process (ij profile).

Figure 24 .Figure 25 .
Figure 24.Oil recovery comparison with gravity effect.Top layer Middle layer Bottom layer

Figure 26 .
Figure 26.Oil recovery comparison between different scattered fractures.

Figure 26 .
Figure 26.Oil recovery comparison between different scattered fractures.

Figure 26 .
Figure 26.Oil recovery comparison between different scattered fractures.

4 Figure 27 .
Figure 27.Oil viscosity comparison between different scattered fractures in the CO2 soaking process.Figure 27.Oil viscosity comparison between different scattered fractures in the CO 2 soaking process.

Figure 27 . 4 Figure 28 .
Figure 27.Oil viscosity comparison between different scattered fractures in the CO2 soaking process.Figure 27.Oil viscosity comparison between different scattered fractures in the CO 2 soaking process.Energies 2019, 12, 823 21 of 29

Figure 28 .
Figure 28.Oil Saturation comparison between different scattered fractures in CO 2 production process.

Figure 29 .
Figure 29.Oil recovery comparison between different directions of scattered fracture.Figure 29.Oil recovery comparison between different directions of scattered fracture.

Figure 29 .
Figure 29.Oil recovery comparison between different directions of scattered fracture.Figure 29.Oil recovery comparison between different directions of scattered fracture.

Figure 30 .
Figure 30.Oil viscosity of different fluids in soaking process (jk profile).Figure 30.Oil viscosity of different fluids in soaking process (jk profile).

Figure 30 .
Figure 30.Oil viscosity of different fluids in soaking process (jk profile).Figure 30.Oil viscosity of different fluids in soaking process (jk profile).

Figure 31 .
Figure 31.Oil saturation of different fluids in production process.

Figure 31 .
Figure 31.Oil saturation of different fluids in production process.

Figure 32 .
Figure 32.Oil recovery comparison between fractures near well and far well.

Figure 32 .
Figure 32.Oil recovery comparison between fractures near well and far well.

Figure 33 .
Figure 33.Relationship between contact area and oil recovery in different modes.

Figure 33 .
Figure 33.Relationship between contact area and oil recovery in different modes.

Figure 33 .
Figure 33.Relationship between contact area and oil recovery in different modes.

Figure 34 .
Figure 34.Production component comparison in different cases.

Figure 34 .
Figure 34.Production component comparison in different cases.

Table 2 .
Basic reservoir properties in the numerical model.

Table 2 .
Basic reservoir properties in the numerical model.

Table 3 .
Binary interaction parameters Figure 7. Relative permeability curve of the fracture.

Table 2 .
Basic reservoir properties in the numerical model.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 4 .
Component properties and composition.

Table 5 .
Oil composition in the simulation model.

Table 6 .
Settings of fracture contact relationships.

Table 6 .
Settings of fracture contact relationships.

Table 6 .
Settings of fracture contact relationships.

Table 6 .
Settings of fracture contact relationships.