When using CT scanning technology, due to the different substances corresponding to different CT values, different fluid saturations in the core can be identified. The calculation formula of core oil saturation is as follows:

In this paper, the core displacement experiment was carried out using a self-developed online CT scanning system (Fig. 1). The scanning equipment is a LIGHTSPEED 8-slice spiral CT scanner developed by General Electric Company of the United States, with the maximum transmission power per tube of 53.2 kW. The CT image analysis software (CCTAS) developed by PetroChina Research Institute of Petroleum Exploration & Development was used for data processing. Also, a special core holder with a shell made of polyether ether ketone (PEEK) material was used to hold the core position. The X-rays could penetrate the core and reduce the scanning error caused by the ray hardening effect, and the on-line CT scanning for air flooding process can be carried out. The QUZIE5200 pump was used as the injection pump, and the ISCO pump was used to accurately control the confining pressure.

Three shale samples were parallelly taken from the Permian Lucaogou Formation in the Jimsar Sag, Junggar Basin, and used to carry out on-line CT scanning experiment on air flooding shale oil. The development effects of shale oil by starting injecting air at different depletion pressures (28, 21, 13 MPa) were simulated. The basic physical parameters of the cores are shown in Table 1.

The experimental steps are as follows: (1) Dry scanning. The oil and salt in the core were washed out with organic solvent, and the core was dried. The dry core was scanned after setting the scanning parameters as tube voltage of 120 kV and tube current of 130 mA. (2) Wet scanning. The simulated oil with a viscosity of 10.4 mPa·s under formation conditions was prepared by using crude oil and white oil in the study area, and the simulated oil was used as experimental oil. The experimental temperature was raised to 85 °C, and the shale core was vacuumed, saturated with oil and aged. The core saturated with oil (wet core) was scanned at the parameters used for dry scanning. (3) Simulation to depletion development. The depletion development was carried out by setting the back pressure to 28 MPa, recording the oil production, and calculating the recovery degree in the depletion stage. (4) Gas flooding simulation. After depletion development, gas flooding was carried out by injecting air into the core samples with a high-pressure displacement pump. An on-line CT scanning test was conducted on the core every 5 min, and relevant experimental data were recorded. The experiment ended when oil was no longer produced. (5) CT image processing. Imported the CT scan image into the CCTAS software, selected regional parameters, and obtained the saturation image of the whole process of air flooding. (6) Steps (1)-(5) were repeated on two other core samples by setting the depletion pressure at 21 MPa and 13 MPa, respectively.

It has been proved mathematically that the attenuation curve related to pore size is a single exponential function, and the attenuation coefficient is positively correlated with pore size. The molecular surface force of the fluid in a large pore is small, and the T₂ value is large. Conversely, the molecular surface force of the fluid in a small pore is large, and the T₂ value is small. That is, the T₂ value can represent the proportion of pores with different sizes to the total pores ^[18-19]. The relationship between pore size and T₂ value can be expressed as Eq. (2). In addition, a large number of scholars have proved that the pore throat distribution measured by mercury intrusion experiment has a good correlation with the pore throat distribution measured by NMR. Therefore in this study, the conversion coefficient (C) in Eq. (2) was determined based on the experimental results of high-pressure mercury injection on shale samples. First, a small sample was cut from the experimental core to carry out the high-pressure mercury injection experiment. The experimental results (Fig. 2a) show that the pore throat radius of shale reservoirs in this area presents a typical bimodal distribution, which can be divided into four orders of magnitude: nanopores, sub-micropores, micropores, mesopores and macropores. The pore throat radius is mainly distributed in 0.01-10.00 μm, and a large number of nano-scale pore throats are developed. The main contribution to permeability is from sub-micropores and micropores, and the pore throat radius is concentrated at 0.15-1.75 μm. In addition, the shale samples saturated with crude oil were tested by NMR, and the conversion coefficient was calculated to be 0.02 μm/ms. Thus, the NMR pore size conversion results shown in Fig. 2b can be obtained.

The on-line NMR experimental system for shale oil by air flooding was developed by combining the high-temperature and high-pressure core displacement physical simulation system and NMR technology (Fig. 3). The system mainly consisted of three parts: (1) A constant- temperature oil bath heating cycle control system. The experimental temperature is controlled by circulating fluorine oil around the core holder, and the experimental temperature and pressure could reach 90 °C and 50 MPa, respectively by special core holder and bath temperature sensor. (2) Fluid displacement physical simulation system, which consists of constant-speed constant-pressure pump, core device, gas-liquid separation device, etc. (3) On-line NMR experimental system, MacroMR12, with resonance frequency of 12.798 MHz, magnet strength of 0.3 T. The shortest echo time of the device is 0.1 ms, which could capture the fluid signals of the nano-scale pore throats in shale samples, and can accurately observe the change of fluid signal in the core pore throats at different displacement times. Before each experiment, the parameters of the NMR T₂ spectrum are adjusted, and central frequency calibration, electronic shimming, and hard pulse amplitude adjustment are carried out according to the specific samples.

Ten shale samples of the Lucaogou Formation in Jimsar Sag were selected to carry out on-line NMR experiments on air flooding shale oil, and the influencing factors on air flooding were analyzed. The basic physical parameters of the cores are shown in Table 2.

The experimental steps are as follows: (1) The air tightness of the experimental system was checked. The system pressure was maintained at 37 MPa, and the pressure change within 24 h was less than 5%. The oil and salt in the core were washed out via organic solvent extraction, the cores were dried and weighed, and the T₂ spectrum curve of original dry core was tested. (2) The experimental temperature was raised to 85 °C, and the core was vacuumed, saturated with oil and aged. (3) The core saturated with oil was put in the holder, the T₂ spectrum curve of core saturated oil was tested, and the saturated oil volume and porosity were calculated using the gravimetric method and NMR method. (4) The air was injected into the core by a high-pressure pump, and then NMR tests were performed at different injection volumes (0.2, 0.4, 0.6, 0.8, 1.0, 1.2 PV (multiples of pore volume)) under constant temperature and pressure. The gas and oil samples were collected at the outlet for chromatographic analysis. The experiment ended when the NMR spectrum did not change. (5) Steps (1)-(4) were repeated on the different core samples, and the effects of oxygen content, permeability and fractures on air flooding oil were analyzed. (6) Steps (1)-(3) were repeated on the different core samples. The gas injection pressure was changed to 34, 36, 38 MPa, and the NMR tests were carried out at different displacement times (2, 4, 6, 8, 10, 12 h), respectively. Then the influence of gas injection pressure on air flooding was analyzed. The standard sample was used to correct the NMR signal for each scan test, which ensures that the experimental results are comparable.

Fig. 4 shows the CT images of the core after injecting air at different depletion pressures. The red areas correspond to the oil in the core, and the blue-green represents the injected air. It can be observed that the crude oil saturation is sufficient in the initial state. As air flooding begins, the gas phase invades into the inlet of the core sample, and the oil saturation at the inlet begins to decrease. There is no obvious change in the middle and the outlet end, and the outlet end continues to produce oil, and the oil production in this stage accounts for a large proportion. In the middle stage of air flooding (the injection volume of 0.6 PV), the overall oil saturation of the core sample first decreased when the depletion pressure was 28 MPa. When the pressure dropped to 21 MPa and 13 MPa, respectively, the oil saturation at the inlet and the middle of the core decreased greatly, and both gas and oil were produced. The air sweeping area was about 1/3 of the core sample, and the gas-to-oil ratio gradually increased. In the later stage (injection volume of 1.0 PV), after gas breakthrough, no oil was produced, and the oil saturation at the inlet decreased significantly, but the oil saturation at the outlet changed little. When the injected volume reached 1.2 PV, the CT image no longer changed. At that time, the oil saturation was the final residual oil saturation, and the residual oil was left at the outlet. According to the experiment results, the oil saturation of the core sample changed most after starting injecting air at 28 MPa, indicating the crude oil production degree is the highest. The production degree at 21 MPa is in the middle, and that at 13 MPa is the lowest. The results show that when the depletion pressure is higher, injected air can contact with crude oil many times, and low-temperature oxidation (LTO) effect may appear at high pressure. The core oil saturation decreases more and the oil displacement effect is better. Furthermore, after the air flooding process, some black colloids were observed near the large pores at the outlet of the core on the CT image. The chromatographic analysis of the oil sample inside the core showed that heavy components such as asphaltene increased, indicating that LTO reaction caused the black solid in the air flooding process.

Fig. 5 is the recovery degree curves of air flooding under different depletion pressures calculated according to the oil saturation of the core. When the pressure depleted to 28, 21, 13 MPa, the recovery degree was 2.15%, 2.37%, and 3.16%, respectively. The lower the pressure is, the higher the recovery degree is. The total recovery degree of air flooding after depletion is 43.28%, 41.39%, and 39.27% respectively, which is 41.13, 39.02, and 36.11 percentage points higher than that of depletion development. The lower the depletion pressure, the lower the recovery degree of air flooding after depletion development. It can be seen that air injection after depletion development can greatly improve the development effect of shale oil, but there are some differences in oil displacement efficiency at different injection points. In addition, the lower the depletion pressure, the earlier the gas breakthrough (Fig. 5). Taking gas flooding after depletion to 21 MPa as an example, the recovery degree before gas breakthrough increased from 2.15% to 39.28%, and the recovery degree after gas breakthrough increased from 41.65% to 41.39%, only increasing by 1.89 percentage points, it can be summarized from above results that the contribution of air flooding to oil recovery is mainly concentrated before the gas breakthrough. Once gas breakthrough occurs, dominant channels will be formed inside the core, and the gas sweeping scope will decrease.

The variation characteristics of oil saturation at different positions of the core sample during air flooding were further analyzed by taking the gas flooding after depletion to 21 MPa as an example. The CT values at different times were obtained by CT scanning, and the distribution of oil saturation at the corresponding stages was calculated as shown in Fig. 6. It can be observed that the oil saturation in the swept area of the gas flooding front was significantly decreased, and an oil front appeared in the sweeping zone, moving to the outlet end, like a piston pushing. Additionally, it can be seen that the oil saturation in the inlet and middle of the core sample decreased obviously, and the crude oil in these areas was swept thoroughly by the gas phase, while there was a considerable amount of crude oil “accumulation” at the outlet end. It can be concluded that the gas phase is a non-wetting phase in the core. Under the action of pressure difference, the gas can not only enter the main seepage channel, but also occupy the micro-pore space where oil is difficult to produce. As a result, the residual oil after depletion development is forced to “start-up” and move to the outlet end. The residual oil in different zones of the core can be “coalescent” to form an “enrichment block” during the migration process, which is pushed by the injected gas to the outlet end, resulting in higher oil saturation near the outlet end. At the end of the experiment, some oil flowed out when the pressure was released. When the core holder was opened, the oil saturation at the core outlet was higher than that at the inlet, which could prove the above phenomenon.

After the end of gas flooding, according to the CT value, the remaining oil saturation at 28, 21, 13 MPa is 40.3%, 41.6%, and 43.3% respectively, which are close to the experimental results of oil saturation based on the measurement data of oil and gas output at the outlet. The error between them is within 1%, indicating that it is feasible to evaluate the variation characteristics of core oil saturation via the on-line CT scanning experiment method.

Fig. 7a shows the NMR T₂ spectrum of sample J6 after air flooding. In order to facilitate the analysis and description of the producing characteristics and migration laws of crude oil in different sizes of pores, the pore classification criteria of the International Union of Pure and Applied Chemistry (IUPAC) and previous research conclusions were used for reference, and the pores were divided into four categories: nanopores (r<0.1 μm), sub- micropores (0.1 μm≤r<1.0 μm), micropores (1.0 μm≤r<5.0 μm), mesopores and macropores (5.0 μm≤r<10.0 μm), corresponding to T₂<5 ms, 5 ms≤T₂ <50 ms, 50 ms≤T₂< 250 ms, 250 ms≤T₂<500 ms, respectively. According to the T₂ spectrum of samples saturated with oil, it can be seen that shale oil is mainly distributed in the pores of 0.005-10.000 μm, and shows obvious bimodal distribution. Nanopores are generally occupied by oil, accounting for about 30% of the total pores, which is consistent with the results of mercury intrusion experiments, indicating that the accuracy of NMR experimental results is high and the crude oil saturation is sufficient.

Fig. 7a presents that with the increase of air injection, the peak area of the T₂ spectra gradually decreased, and the oil content of the core gradually decreased, indicating that the crude oil in different pore throats of samples can be effectively produced by air flooding. When air injection was 0.2 PV, the peak signals on the left and right sides decreased, but the right peak signal decreased greatly, indicating that air first displaced crude oil in micropores, mesopores and macropores. When air injection was 0.4 PV, the crest of the left peak changed little, and the crest of the right peak continued to decline and the decrease was large, indicating that with continuous increase of air injection, oil production from mesopores and macropores slowed down, while sub-micropores and micropores became dominant oil producers. When air injection was 0.8 PV, the crest of the left peak decreased, the right side of the envelope shifted slightly to the left and the width narrowed, indicating that oil production from nanopores increased. When air injection was 1.2 PV, the spectral line coincided with that when air injection was 1.0 PV, and the width of the envelope line changed slightly, indicating that the lower limit of the movable fluid was reached, and no more oil would produce from the core. In conclusion, the peak zone decreased faster in the initial stage, reflecting higher oil recovery in the stage; and the peak zone changed slowly in the later stage, indicating decreasing oil recovery rate. The oil flooded by air is mainly in the pores with T₂ values of 0.5-500 ms, corresponding to the pores of 0.01-10.00 μm. The oil in the pores with T₂ values less than 0.01 μm was hardly produced. At the end of air flooding, the oil in the pores of 5-10 μm was completely displaced, leaving residual oil mainly in the nanopores and sub-micropores.

Fig. 7b shows the relationship between oil recovery degree and injected volume for different sizes of pores according to the NMR T₂ spectra of shale oil after air flooding. It can be seen that there are obvious differences in the production characteristics of crude oil in different sizes of pores. The crude oil in the mesopores and macropores can be preferentially produced under a small injection volume, and the recovery degree initially increased to the maximum and then remained constant. The recovery degree of sub-micropores and micropores gradually increased with the injection volume, but the increase amplitude gradually decreased, while the recovery degree from nanopores increased significantly under the injection volume of 0.8 PV, which is consistent with the NMR T₂ spectra. It can be obtained that this is related to the sweeping position and migration path of high-pressure air in shale pores. High-pressure air first swept the mesopores and macropores with small flow resistance, and then accumulated and circulated there. The sufficient contact between air and crude oil greatly increased the production from the mesopores and macropores. In the later stage, the pressure was gradually transmitted to the nanoscale pores. Under the action of pressure difference and LTO reaction, the crude oil in the nanoscale pores, which are difficult to be produced via water flooding, flew into larger pores and migrated to the outlet through the macropores. Therefore, it can be summarized from the above results that the oil production from macropores shows a trend of initially fast and then slow, but the oil production from nanopores increases almost linearly. When the air injection volume is 0.6 PV, it is notable that the recovery degree curves of different pores rose slowly, the T₂ spectral line decreased slowly, and the bimodal distribution trend of the residual oil is more obvious. It can be speculated that it may be caused by gas breakthrough which reduced the increase of the recovery degree.

Based on the differences in air flooding sweeping mode and production characteristics of shale oil, the core can be divided into reaction trailing edge area, reaction front edge area, and unswept area. Among them, the reaction front edge area can be driven by pressure difference and dissolution. The main flooding mechanism of differential pressure is that high-pressure air enters the matrix pore throats under pressure and displaces the crude oil by differential pressure. The main oil displacement mechanism of dissolution is LTO reaction between oxygen in the air injected and crude oil. The CO, CO₂, ether, aldehyde and ketone produced by the reaction can expand the volume of crude oil, reduce the viscosity and enhance the fluidity. Meanwhile, the mass exchange between oil and gas makes the reaction products further dissolve into the crude oil, which can effectively supplement the formation energy. With the decrease of depletion pressure, the gas dissolved in the crude oil separates and forms dissolved gas flooding. Moreover, the light components evaporated from the oil, together with the CO and CO₂ etc. can play a role similar to flue gas flooding. The flue gas continuously extracts the low-carbon components in the crude oil, which makes the injected gas continuously enriched and close to the properties of the oil. Meanwhile, the crude oil produces a slow component mass transfer under the action of extraction, and gradually depletes and closes to the properties of the injected gas. As a result, the interfacial tension of oil and gas decreases, and the injected gas, light and medium hydrocarbon gas, and crude oil act as nearly miscible or partially thermal miscible flooding under high temperature and high pressure. Therefore, under various oil recovery mechanisms, during air flooding, LTO reaction induces a high-temperature displacement front which is similar to the bulldozing effect proposed by Gutierrez et al. ^[20]. This high-temperature displacement front pushes crude oil forward, like a bulldozer, which can greatly improve the oil recovery. Jiang et al. ^[21-22] proposed that the EOR by air flooding is the collaboration of N₂ flooding, flue gas flooding, LTO reaction and heating expansion via numerical simulation and experimental demonstration. The heat generated by the LTO reaction can accelerate carbon bond stripping effect and induce fire flooding, which has good miscible, gasification/distillation volatilization effects. In this case, thermal flooding, steam flooding, miscible flooding, and gas flooding are main mechanisms of EOR.

Notably, it is extremely important to select the appropriate injection and production timing for shale oil by air flooding. When the depletion pressure is low, the reservoir energy cannot be replenished in time, so that the pressure in pores and fractures decreases rapidly, which easily leads to a sharp increase in effective stress. The reservoir will elastically deform, resulting in a decrease in the cementation degree among the micro- and nano-scale mineral particles. These mineral particles may slip under pressure, and their contact mode and arrangement will change. Meanwhile, nano-micron pore throats will be easily filled, which will result in a decrease in reservoir seepage capacity and poor air injection effect. In shale reservoirs, in addition to primary intergranular pores, there are also complex low permeability channels such as cement micropores and intragranular pores that haven’t undergone strong compaction. CT images show that micropores and microfractures are easy to close under high stress, indicating that the micro-and nano-pores in shale are difficult to support overlying pressure when the effective stress increases. Even a small pressure change will lead to a significant change in the extremely small seepage channels in shale reservoirs. If injecting air at this time, it is easy to cause stress sensitivity and make air injection effect worse.

Air oxygen content is an important factor affecting the development effect of shale oil by air flooding. Fig. 8 shows the NMR T₂ spectra of air flooding at different oxygen contents. It’s found that as the oxygen content in the air increases from 5% to 21%, the recovery factor increases from 41.62% to 45.71%, indicating that the increase of oxygen content is favorable for improving oil recovery. Oil produced from nanopores increases from 24.8% to 27.4%, that from sub-micropores increases from 27.1% to 31.4%, and that from micropores increases from 36.4% to 38.2%, but oil from mesopores and macropores decreases from 11.7% to 3.0%. It can be found that the small number of mesopores and macropores in shale cores leads to less oil and less contribution to total recovery factor. In comparison, the amount of crude oil in nanopores, micropores, and sub-micropores is large and the production characteristic is more obvious, indicating that the total oil production of air flooding mainly comes from nanopores, sub-micropores, and micropores, and the oil recovery from the three kinds of pores increases with the increase of air oxygen content.

From the above results, it can be concluded that when the oxygen content is low (5%), the air composition is mainly N₂, and the injected air plays a role similar to N₂ flooding. N₂ molecules are small and can evenly enter mesopores and macropores, playing a role in increasing pore pressure and supplementing formation energy. When the reservoir pressure decreases, the gas dissolved in crude oil can be released, and dissolved gas flooding can occur under the action of elastic energy and pressure difference. Then, the crude oil in large pore throats can be displaced and the recovery degree increases, but the production effect on the crude oil in micro- and nano-pores is poor. When the oxygen content is high (21%), with the increase of air injection, LTO reaction between air and crude oil in micro- and nano-pores is mainly oxygen addition reaction and decomposition reaction. From the above mechanism analysis of EOR by air flooding, CO₂ and aldehydes produced by LTO reaction at high oxygen content plays a good role in viscosity reduction by expansion for crude oil in micro- and nano-pore (Fig. 9a). The resulting flue gas flooding can further enhance the recovery degree from micro- and nano-pores. This is consistent with the LOT model proposed by Ren et al. ^[23] after experimental demonstration of the LTO process of crude oil in the core. Meanwhile, as the viscosity decreases, light and medium components in the crude oil and flue gas can form a supercritical and miscible front zone under the action of thermal expansion and gasification/distillation, to significantly enhance oil recovery by “thermally miscible flooding”. It is worth noting that the high oxygen content produced higher oxygen pressure in the experiment, which promoted LTO reaction, as a result, the temperature in the core holder increased by about 2-3 °C, measured at the oil-gas contact front. This proves that the thermal effect produced by LTO reaction can increase reservoir temperature. Limited by short contact time between crude oil and gas in the core under experimental conditions, the LOT effect is not sufficient, therefore the oxygen content of the air injected can be increased in the field within the safety range to enhance oil recovery. Of course, at this point, the residual amount of oxygen after reacting with crude oil becomes larger, and the probability of oxygen concentration in production wells may be risky if exceeding the explosion limit, so an appropriate air oxygen content should be selected.

The chromatographic analysis of produced oil at different oxygen content during air flooding is shown in Fig. 9b. The composition of crude oil after air flooding showed a trend of decrease in zone I (light component C₅-C₁₆ in crude oil), an increase in zone II (heavy component C₁₇-C₂₈ in crude oil), and a slight increase in zone III (C₂₉-C₃₅). The higher the oxygen content, the more sufficient the LTO reaction is, and the more obvious the trend of each zone. This is because the expansion and extraction of dissolved gas in oil can extract the light components in crude oil into the gas phase, and some light components are converted to heavy components. The more sufficient the LTO reaction, the smaller the proportion of light components in crude oil is. Meanwhile, the thermal effect produced by LTO reaction also causes the light components in the oil to evaporate, the proportion of medium and heavy components increases, and the density of the crude oil gradually increases. The medium and heavy components have larger relative molecular mass, more branched chains, and more complex spatial three- dimensional structures which can play a role in blocking the dominant channels in the core, and slow down the phenomenon of fingering and channeling ^{[12,23 -24]}.

In general, the evaluation index of physical properties of high-permeability core is high, and the air has good seepage characteristics in the core during gas flooding development. When the core permeability is low, the injection-production pressure difference is difficult to be effectively established and transmitted to the outlet end, which will cause high injection pressure at the inlet. Fig. 10a and Fig. 10b are NMR T₂ spectra after air flooding of shale samples with different permeabilities. Compared with the low-permeability core (0.074×10⁻³ μm²), the high-permeability core (0.919×10⁻³ μm²) has more large- scale pore throats and higher initial oil content. The NMR signal values of nanopores, sub-micropores, micropores, mesopores, and macropores decreased significantly. Combined with Fig. 10c and Table 3, it can be seen that the recovery degrees of pores with different sizes in high permeability core are 33.6%, 31.1%, 32.5% and 2.8% respectively, which are significantly higher than those in the low permeability core. The final recovery factors of the low-permeability core and the high-permeability core is 17.81% and 46.37% respectively. This means that high permeability means strong fluid flow ability and better pore throat connectivity. Not only the crude oil in large-scale pore throats can be produced, but also the oil from nano-scale pore throats can be produced to a higher degree. In addition, the T₂ spectral signals in the 0.01-0.05 μm pores of the low-permeability core increases slightly and then decreases during the gas flooding, which is caused by the tight lithology of the low-permeability core. Some crude oil is carried by air and transferred to smaller pores during air flooding, which makes the residual oil redistributed. In the later stage, with the increase of gas injection, the air is an excess phase relative to crude oil, which leads to the change of physical properties of the crude oil in micropores, so the oil is produced under the action of air displacement.

It is notable that during the gas flooding, air first sweeps the oil in high-permeability zone of the low-permeability core, and the oil content in high-permeability zone decreases significantly. However, due to the poor pore throat connectivity, there is a “bypass effect” in the low-permeability zone, so the oil production is poor ^[25]. Gas is easy to enter the high-permeability dominant channels during gas flooding in low-permeability cores, resulting in the rapid breakthrough of displacement pressure and gas channeling, and leaving oil in some non-dominant channels. This is reflected by relatively low oil production from nanopores and sub-micropores and high oil production from mesopores and macropores in the low-permeability core. In field application, if the reservoir seepage capacity can be improved by fracturing before gas injection, the shale oil recovery will be greatly improved.

The injection pressure is a key factor influencing air injection development of shale oil. Injection pressure mainly affects the stability of the oil-gas front interface and thereby affects the microscopic pore-throat swept area of the core. Al-Mudhafar et al. ^[26] reported that the injection pressure has an important influence on the recovery factor. When the injection pressure is greater than the critical injection pressure, the stability of the flooding front is poor, which is mainly manifested as unstable flooding and low recovery. When the injection pressure is less than the critical injection pressure, high recovery and stable flooding are obtained. As shown in Fig. 11a, with the increase of flooding time, the T₂ spectra gradually moves down, indicating that crude oil is effectively produced under a certain pressure difference. When the injection pressure is 36 MPa, the crude oil from the pores of 0.05-10.00 μm is mainly produced. When the injection pressure increases to 38 MPa, the crude oil in pores of 0.01-10.00 μm can be produced. When the injection pressure decreases to 34 MPa, the crude oil in pores of 0.08-10.00 μm is producible. In general, the T₂ spectrum moves left and the signal amplitude moves down with the increase of injection pressure. This shows that when the injection pressure increases, the lower limit of oil production of pore throats decreases from 0.08 μm corresponding to 34 MPa to 0.01 μm corresponding to 38 MPa. It can be concluded that the interfacial tension between air and crude oil decreases continuously under higher injection pressure, and the resistance of air entering small pore throat decreases, thereby the crude oil in small pore throats can be effectively produced ^[27-28].

According to the principle of NMR, the occurrence of pore fluid is positively correlated with the signal amplitude, the corresponding signal amplitude will decrease when the occurrence of crude oil in the smallest pore decreases; that is, the pore size is the lower limit of pore throat production. In porous media, fluid flow with Reynolds number between 1×10⁻⁶ and 0.2 is Darcy flow; when the Reynolds number is higher than 0.2, it is high-speed nonlinear flow; when the Reynolds number is less than 1×10⁻⁶, it is low-speed nonlinear flow. The crude oil viscosity of the shale reservoir in the study area is 10.41 mPa·s, and the crude oil density is 0.886 g/cm³. Combined with the daily liquid production per well, fluid flow in the shale matrix is mostly low-speed nonlinear flow. The lower limit of pore throat production corresponding to this flow regime is about 0.01 μm, which is close to the result calibrated by the on-line NMR experiment. Therefore, the target for air flooding in the study area should focus on the pore throat space with a certain flowing capacity and larger than 0.01 μm.

Fig. 11b shows the variation curves of crude oil recovery degree with injection pressure, calculated by T₂ spectral area. It can be seen that the crude oil recovery degree is 37.66%, 43.81%, and 39.46% respectively at the injection pressure of 34, 36 and 38 MPa. As the injection pressure increases, the recovery degree increases and then decreases, which is related to the sweeping mode and migration position of air in the shale pore throats. When reservoir temperature and injection pressure are suitable, air can slowly sweep small pores such as nanopores and sub-micropores, and the crude oil in small pores can sufficiently contact with air. When the internal action energy of air is close to the activation energy of crude oil in small pores, according to similar solubility, the increase of solubility of air in crude oil can improve the total recovery factor. When the injection pressure is higher, air can enter some nanopores under the action of high pressure difference, which increases the production degree of crude oil in nanopores (Fig. 11c). However, air will also accumulate inside the core to form a continuous gas phase zone, increasing the gas viscosity. When the viscous force is greater than the capillary pressure, the fingering phenomenon may occur, and result in earlier gas channeling, and consequently significant decreases in contact area and contact time between oil and gas, and premature breakthrough. A large proportion of crude oil occupies the small pores such as sub-micropores in shale. When gas channeling occurs, the decrease of swept small pores will result in a decrease in crude oil production, and then a decrease in total recovery degree at high injection pressure. In field gas flooding operation, suitable injection pressure is necessary to avoid gas channeling.

The indoor experimental results show that the air injection capacity is significantly enhanced when the shale core contains fractures. Combined with digital core images and logging data, it can be found that there are obvious micro-fractures such as small-angle bedding fractures, fracturing fractures, diagenetic shrinkage fractures, structural fractures and slip fractures in shale samples, which are arranged in light and dark layers. The horizontal layer connectivity is better, but the vertical connectivity is poor. Micro-fractures such as bedding fractures provide shale reservoir space, and improve shale permeability. They are effective channels for shale oil to flow out of matrix pores, and greatly improve the efficiency of air displacing crude oil. Fractures have a significant influence on shale oil recovery by air flooding.

Cores J12 and J13 in Table 2 were selected to analyze the effect of fractures on EOR by air injection in shale oil. Core J13 has a microfracture with a width of about 0.1 mm, and permeability of 0.794×10⁻³ μm², which can represent the fractured core. Core J12 is a matrix core with a permeability of 0.307×10⁻³ μm². Fig.12 shows the NMR T₂ spectra of matrix and fractured cores under different gas injection volumes. It can be seen that the matrix core mainly develops nanopores, with some sub-micropores, and micropores, but the development degree of mesopores and macropores is low. The fractured core not only develops nanopores, sub-micropores, micropores, mesopores and macropores, but also has two T₂ spectral peaks at pore sizes greater than 12 μm. The signal amplitude decreases with the air injection volume, which is caused by the change of residual oil saturation in core fractures during air flooding. Therefore, pores with pore size greater than 12 μm are defined as fracture. Compared with the matrix core, the NMR signals of all sizes of pores in the fractured core decreased significantly during air flooding, indicating that the fracture effectively improved the connectivity of the core.

Fig. 13 is the NMR sagittal plane image of the fractured core under different injection volumes. Large signal intensity represents high oil saturation, and small signal intensity represents low oil saturation. In the early stage of air flooding (injection volume of 0.4 PV), air first enters the high-permeability fracture, and the image color at the fracture becomes lighter (Fig. 13b), indicating that the crude oil from fracture is produced, but the production degree of crude oil in the matrix is low, and the oil mass exchange between the fracture and the matrix is not obvious. With the increase of injection volume, the air in fracture and macropores gradually enters into the core matrix. When the injection volume is 0.6 PV, the NMR T₂ spectrum envelope shifts left and the T₂ spectrum peak of the oil in small pores decreases significantly, indicating that the crude oil containing air in the matrix migrates to the fracture, that is, the matrix supplied oil to the fracture (Figs. 12b, 13c). At this time, the crude oil in the core matrix can be attributed to the following three aspects: (1) Local oxidation reaction occurs when crude oil from the matrix contacts with the gas in the fracture, and the thermal effect causes the matrix crude oil to expand. The resulting thermally miscible flooding can reduce the viscosity of the crude oil in the matrix and improve its fluidity. (2) Gas flows toward the upper part of the core under gravity, the micro- and nano-pores that are difficult to produce can be affected, and LTO reaction between air and crude oil occurs under the action of cross flow. The generated gaseous light hydrocarbon compounds can contact with the crude oil in pores at “dead end” of the matrix, resulting in mass exchange, thereby reducing the interfacial tension of oil and gas, and greatly reducing the seepage resistance of the matrix crude oil to the fracture. (3) The CO₂ (temperature higher than 31.06 °C, pressure higher than 7.39 MPa) generated by the LTO reaction of oil and gas is in a supercritical state and is easily dissolved in matrix crude oil of micropores. Therefore, miscible or near-miscible flooding can occur in the narrow area near the shale matrix/fracture interface, and the matrix crude oil can be produced under the action of desorption and extraction, and the main flow channel is still the fracture. Furthermore, from the change of core oil saturation, it can be seen that the recovery degree of crude oil from matrix in the fractured core is greatly improved in the later stage of air flooding, and the upper part of the matrix is the main residual oil enrichment area (Fig. 13d). The T₂ spectra indicate that the residual oil mainly occurs in the sub-micropores which are the target of further production.

In essence, when shale fractures are developed, crude oil in the macropores around the fractures can be produced in the early stage of air injection, by the differential pressure that plays a major role. With the increase of air injection volume, the pressure will be transmitted to the matrix, and the air in the macropores and fractures sweeps the small pores, and the dissolution contributes more to oil recovery. The crude oil in the macropores first flows rapidly to the fractures, and then the crude oil in the small pores flows through the large pores to the fractures as the seepage distance and resistance decrease. Consequently, the total recovery increases. In addition, under the indoor experimental conditions, the back pressure corresponding to the injection pressure was established at the outlet of the fractured core, the stable and effective displacement pressure difference can be established inside the core, and the possibility of gas channeling is reduced, thereby improving the oil displacement effect. The ultimate oil recovery factor can be calculated by dividing the difference between the gray value of the NMR image at initial oil saturation and that after air flooding by the gray value of the NMR image at initial oil saturation. The ultimate recovery factor of the matrix core and fractured core is 34.54% and 57.94%, respectively. The fracture can effectively increase the contact area between gas and crude oil and the range of displacement front, reduce the seepage resistance and distance of matrix crude oil, and improve the sweeping coefficient and matrix oil drainage area. In field operation, comprehensive factors should be considered to select gas injection parameters. Under the premise of appropriately increasing the injection pressure, it is necessary to select a reasonable pressure difference to prolong the contact between gas and crude oil and delay gas breakthrough. Meanwhile, the reservoir can be adjusted by combining the anti-channeling system such as air foam, which can decrease the risk of gas channeling and improve the recovery degree.