The Effect of Hydraulic-Natural Fracture Networks on the Waterflooding Development in a Multilayer Tight Reservoir: Case Study

An integrated hydraulic fracturing followed by water ﬂ ooding was conducted in a heterogeneous sandstone formation in the Northern Shanxi Slop of Ordos Basin in Western China. Water breakthrough quickly occurred, and the underlying mechanism of water breakthrough has not been well understood. Such mechanism needs to be investigated comprehensively from the spatial connectivity of multilayer sand bodies and characterization of hydraulic-natural fracture networks. Here, an integrated approach is proposed to tap the remaining oil in the individual sand layer during the late-stage development of tight sandstone reservoirs. A case study is utilized to demonstrate the applicability of the integrated method. It is found that the six sand layers could be further divided within the target oil layers. These sand layers have a variety of physical and mechanical properties, leading to the asymmetric spatial distribution of hydraulic fractures after performing the integrated fracturing of whole oil layers. The spatial di ﬀ erence of sand bodies conforms to the features of the multiperiod superimposed channel in the sedimentary environment of fan delta front. The natural fractures were generated from the tectonic movement in the Mesozoic period with a dominant orientation of approximately NE 67 ° . The asymmetric hydraulic fractures propagated and connected with the preexisting natural fractures, forming the intricate natural-hydraulic fracture networks. The water breakthrough pattern in each sand layer is primarily ascribed to the spatial distribution of the hydraulic-natural fracture networks and sedimentary microfacies. The refracturing operations based on the remaining oil distribution in sand layers are proven to be e ﬀ ective in further developing the formation. The average oil production of related wells increased from 0.61 t/d to 2.18t/d. This practical development strategy provides insights for further development of likewise heterogeneous tight sandstone reservoirs.


Introduction
In Western China, along with the long-term water injection development, the oil production within pay zones is uneven in the naturally fractured tight sandstone reservoirs and usually accompanied by the increasing water cut [1,2]. The fracture-induced water breakthrough is frequently observed shortly after the integrated fracturing operations targeting the whole oil layers ( [3,4]; Van et al., 2018). The trigger mechanisms of such fracture-induced water breakthrough have not been well investigated, especially in consideration with the spatial connectivity of multistage sand bodies and the intricate hydraulic-natural fracture networks after the integrated fracturing operations [1][2][3]5]. Thus, it is necessary to conduct the comprehensive study to characterize the hydraulic-natural fracture networks based on the multistage sand layers subdivision in these reservoirs. The improve understanding of individual layer-based fracture network could help propose proper strategy to tap the remaining oil in the late stage development of tight sandstone reservoir.
The hydraulic fracturing technology has been widely utilized in the development of the tight sandstone reservoirs. In 1947, the hydraulic fracturing was first performed in field tests in Kansas, USA [6]. Fast et al. [7] proposed the largescale hydraulic fracturing technology in the ultralow permeability Wattenberg gas field. However, although the large-scale hydraulic fracturing is effective in enhancing the oil recovery of some reservoirs, it is not practical in some low permeability reservoirs [4]. Rahim et al. [8] used the finite difference method to simulate the impact of operational and geomechanical parameters on hydraulic fracturing in the layered reservoirs. The results suggested that if the interlayers developed in the pay zones, it was necessary to perform fracturing stimulation for each single oil layer, as the interlayers may function as the barriers to restrain the vertical propagation of hydraulic fractures. Abdideh et al. [9] argued that the favourable oil layer for hydraulic fracturing should be optimized based on the comprehensive study of rock mechanics, in situ stress field, proper density of drilling fluid, and various stress changes caused by fracturing treatments. Wen et al. [10] proposed the integrated fracturing theory for low porosity-low permeability reservoirs, which improved the development effect of integrated hydraulic fracturing in these reservoirs.
However, although the integrated fracturing may be suitable for the early stage of oilfield development, it may not be practical in the middle-late stage oilfield development [4,5]. Instead, the single sand body-based fracturing operations are required in the late stage [11,12]. Wang et al. [11] studied the factors that affected the refracturing effect in the lowpermeability oilfields. The heterogeneity within the oil layers plays an essential role on the stimulation effect of refracturing in low permeability reservoirs. Weng et al. [12] conducted in-depth research on the well and layer selection in the WY Oilfield. The potential natural fractures may pose the significant role on the effect of refracturing in the low permeability reservoirs. Thus, the spatial connectivity of multistage superimposed sand bodies, combined with complex hydraulic-natural fracture networks should be investigated comprehensively to tap the remaining oil effectively via refracturing the individual sand layer.
An integrated approach is proposed to tap the remaining oil in each sand layer during the late-stage development of tight sandstone reservoirs. A case study is utilized to demonstrate the applicability of the integrated method. Based on the division of sand layers and analysis of sedimentary facies, the hydraulic-natural fracture networks in each sand layer are determined by combining the inferred natural fractures with simulated hydraulic fractures. The refracturing measures are then conducted to tap the remaining oil within in the individual sand layer. Figure 1 shows an integrated approach used to tap the remaining oil in the heterogeneous sand layers in the late stage development of tight sandstone reservoirs. Specifically, the multiscale field database is first established by collecting outcrops and core observation, well logging, production, and monitoring data. The sand layer division and sedimentary facies analysis are then conducted by comprehensively analysing the multiscale data. Next, the single sand body-based fracture network is established by combining the inferred natural fractures with simulated hydraulic fractures. A 3D single sand body-based geological model is finally built up. The proper development strategy is proposed accordingly to tap the remaining oil in the individual sand layer. The specific procedures for this integrated approach are the following.

Stratigraphic Subdivision and Sedimentology Analysis.
The availability of outcrops and core data provides a good opportunity for further dividing the possible sand layers within oil layers. The relationship between the core observation results and well logging features is first investigated. The muddy or calcareous interlayers are then distinguished from the sand bodies, which mark the boundary between two adjacent single layers [13]. The sedimentary microfacies at the well sites are interpreted from the core observation and well logging response in each single layer [14]. Based on the well-established sedimentary pattern, the microfacies between the wells are finally determined to characterize the spatial connectivity of each individual sand layer. Moreover, this connectivity could also be corroborated by the related response between the injection wells and production wells.

Determination of Hydraulic-Natural Fracture Networks.
The generation of natural fractures is primarily attributed to the regional tectonic movement. The spatial distribution of natural fractures could be determined by the fractureinduced water breakthrough and tracer test. The natural fracture at the well site is identified by the characteristic image logging and well logging response. The natural fracture properties (i.e., strike, dip, and filling state) could be obtained from the imaging logging analysis of coring wells [15]. The propagation of hydraulic fractures is simulated based on the rock mechanical properties and real-time treatment data. The mechanical properties (i.e., Poisson's ratio and Young's modulus) of rocks are estimated first from the velocity logging data. The real-time treatment data of stimulated wells are then utilized to determine the geometry of hydraulic fractures by history-matching the net operational pressure [16]. Finally, the hydraulic-natural fracture networks are built up to investigate their effect on the waterflooding in the area.
2.3. Tapping the Remaining Oil of Individual Sand Layer. The 3D single sand body-based geological model is established by integrating results from the sand layer subdivision, microfacies analysis, and fracture network realization. The spatial distribution of the remaining oil is then determined within the single sand layer (Lin et al., 2018). Finally, the corresponding measures of reperforation and refracturing could be conducted to further tap the remaining oil within the corresponding individual sand layer.

Results and Discussion
3.1. Field Background. A case study is utilized to demonstrate the applicability of the aforementioned integrated method. The study region is located in the Northern Shanxi Slop of Ordos Basin in Western China ( Figure 2). The main production layers are the C6 1 oil layer group of Yanchang Formation in the Upper Triassic System of the Mesozoic period. Under the sedimentary setting of fan delta facies, the multistage sand bodies in the target reservoir display different superposition patterns and internal structures [2,3]. The C6 1 reservoir in the WY Oilfield is characterized by the 2 Geofluids extra-low permeability, with the average permeability of only approximately 1.2 mD (i.e., 1:2 × 10 −3 μm 2 ). Moreover, the average porosity is about 12.4%, while the average original water saturation is appropriately 45%. Thus, this reservoir is a typical low-porosity and low-permeability reservoir, which needs to perform hydraulic fracturing to enhance its oil production. However, after performing the integrated fracturing on the C6 1 reservoir for oil production, many production wells exhibited a high degree of water cut shortly after the    3 Geofluids fracturing stimulations. In 2016, 146 fracturing production wells were either shut down or switched to the injection wells due to the high water cut. The water breakthrough was also observed within the C6 1 oil layers shortly after fracturing operations of particular wells. Its spatial distribution exhibited several lineaments of approximately NE 67°trending [2,3]. The reservoir entered a stage of high water cut and low recovery degree as a result of this water breakthrough in many fracturing wells. Previous works have demonstrated that such water breakthrough patterns are linked to the intri-cacy of subsurface fracture networks in heterogeneous multistage sand bodies. Therefore, it is essential to investigate the effect of fracture networks on the waterflooding development and propose the corresponding strategy to tap the remaining oil in the reservoir.

Multistage Sand Layer Subdivision and Sedimentary
Microfacies Analysis 3.2.1. Sand Layer Division and Correlation. The C6 1 oil layer group is divided into the C6 1-1 , C6 1-2 , and C6 1-3 oil layer [3,     . Based on the outcrops and core observation, the C6 1-1 oil layer is primarily composed of siltstone and hence not the pay zone. In contrast, the C6 1-2 and C6 1-3 oil layers consist of oil-saturated fine sandstones and contribute to the majority of oil production. Figure 3 shows the observed outcrops of related sand layers. It is found that the C6 1-2 and C6 1-3 oil layers have the features of multistage sand bodies' superposition. Given that the muddy or calcareous interlayers mark the boundary between two adjacent individual sand layers, the C6 1-2 oil layers could be further subdivided into the C6 1-2-1 , C6 1-2-2 , C6 1-2-3 , and C6 1-2-4 sand layer, whereas the C6 1-3 oil layer into C6 1-3-1 and C6 1-3-2 . The thickness of one sand layer is approximately 2~5 m, consistent with the thickness derived from the logging response.

Microfacies Analysis for Each Sand
Layer. The C6 1 oil layers, as a typical shallow-water platform, were deposited in the sedimentary environment of fan delta front [3]. Based on the classic sedimentary pattern, the microfacies at the well

Geofluids
sites are further interpreted from the core observation and logging response. It is shown that the subaqueous distributary channel, interdistributary bay, overbank, and sheet sand developed in the study region in order of decreasing fre-quency ( Figure 4). Overall, the sedimentary datum level of the C6 1 oil layers followed a downward trend, in which the A/S value showed a fluctuate decrease. Specifically, in the period of C6 2 , the isolated channel belt generated with a   6 Geofluids small ratio between the channel and interdistributary bay. In contrast, during the period of C6 1-3 , the distributary channel gradually emerged, and the branch channel began to swing laterally. A notable increase is observed in the deposition scale of the main channel and the branch channel belt. The overbank sediments switched to the silt-fine sandstone from the argillaceous sediments, accompanied by the increasing ratio of the channel to interdistributary bay. The distributary channel continued to grow during the period of C6 1-2 . The ratio of central channel belt to branch channel belt further raised, whereas the interchannel mudstone gradually disappeared. Then, the sand-land ratio reached a local peak value in the C6 1 oil layers. Figure 5 shows the map view of inferred microfacies in the C6 1-2-1 , C6 1-2-2 , C6 1-2-3 , C6 1-2-4 , C6 1-3-1 , and C6 1-3-2 sand layer. It is shown that the distributary channel mainly developed in the C6 1-2-2 and C6 1-2-3 sand layer, in which interdistributary bay microfacies seldom developed. The ratio of sand to formation thickness in these two layers is relatively high in comparison with other sand layers. In the C6 1-2-1 sand layer, two central distributary channels and several branch channels are developed in the study area. The overbank sand deposits on both sides of the channel. In the C6 1-2-2 sand layer, the distributary channel microfacies concentrate in the west area while the overbank mainly develops in the east region. The distribution of microfacies in the C6 1-3-1 sand layer displayed nearly the same pattern as that of the C6 1-3-2 sand layer. Two branches of distributary channel generate in the west and east area in these two sand bodies, while the over banks scattered distribute in the corresponding sand layer. The interlayer usually formulates between the main channel and the overbank microfacies, which would function as a barrier to restrain the spatial propagation of hydraulic fractures during fracturing operations.

Determination of Hydraulic-Natural Fracture Networks
3.3.1. Preexisting Natural Fracture Identification. The natural fractures in the C6 1 reservoir have been ascribed to the tectonic movement in the Yanshan and Xishan period of Mesozoic [17,18]. Based on the previous studies on the associated outcrops, natural fractures generated with an orientation range of NE 66°~85°. This range of orientation agrees with that of the maximum principal stress (S Hmax ) in the WY Oilfield, which is NE 67°according to the comprehensive analysis of the multiscale database [2].
The natural fractures could be recognized from the Micro-Conductivity Imager (MCI) logging, core observation, production-injection response, and field test. Specifically  7 Geofluids pattern in the MCI logging, shown in Figure 6(a). The dip angle of this fracture is interpreted to be 75°. Table 1 shows the statistics of natural fractures properties (i.e., strike and dip) detected from the MCI logging data of four wells. It is shown that natural fractures in this region are mainly NEtrending high-angle fractures, with a dip range of 66.8°to 87.1°and strike range NE 20.1°to NE 114.3°. Moreover, some core samples also exhibit a similar fracture pattern. Figure 6(b) depicts a natural fracture with a height larger than 0.7 m and a dip angle of approximately 80°.
Additionally, the existence of natural fractures is also characterized by the abnormality in the conventional well logging. As shown in Figure 6(c), the magnitude of the gammaray log (GR) is abnormally high at a depth of 1248.9 m in this well. This abnormality is possibly ascribed to the existence of natural fractures because the radioactive uranium in the reservoir could be precipitated inside the fractures, leading to an unusually high magnitude of gamma-ray response. Similarly, given that the drilling fluids flowed into the fractures and reduced the measured formation resistivity, the deep resistivity (AT90) at the same depth is relatively low in comparison with other layers. Moreover, the interpretation of Pulsed Neutron-Neutron (PNN) logging further verifies the existence of natural fracture developed, which is interpreted as the waterflooded state at this depth. The production test for this layer shows the pure water production of 35.1 t per day. These various features indicate the existence of natural fractures at the specific depth of this well (Figure 6(c)).
In addition, the natural fractures are mainly in an unfilled state based on the core observation and MCI logging [17]. Under the high in situ formation pressure (11.2 MPa in 1040 m), the unfilled fractures temporarily closed. However, as the long-term water flooding continued, the formation pressure increases correspondingly, and these fractures would be stimulated and reopened as a high permeable zone [2]. In this scenario, the hydraulic fractures propagate and connect with aforementioned reopened natural fractures, which might account for the water breakthrough phenomenon is frequently observed shortly after the fracturing operations in this area.

Geomechanical Analysis and Hydraulic Fracture
Propagation. The propagation of hydraulic fractures is simulated based on the geomechanical properties and real-time treatment data. Specifically, geomechanical properties (i.e., Poisson's ratio and Young's modulus) are derived from the data of velocity and density logging, using the following expressions [19]: where V s is the S-wave velocity, m/s; V p is the P-wave velocity, m/s; ν is the Poisson's ratio; E is the Young's modulus, GPa; and ρ is the density, kg/m 3 . Well w155 is logged to the target zone and measured the P-wave (V p ) and S-wave velocities (V s ). The regression expression is derived from the relationship between the velocity logging and other conventional logging for this well. Next, the calculation of geomechanical parameters is then extended to other wells, using the following regression expression: where GR is the gamma-ray, API, and RT is the deep formation resistivity, Ω·m.
The changes of in situ stress fields in response to hydraulic fracturing are considered in this work. The induced stress tensors for the random spatial point in proximity to hydraulic fractures are calculated by the following expressions [20]: σ z ′ = −P r a a 2 r 1 r 2 3/2 sin θ sin where σ x ' , σ y ' , and σ z ' are the induced stress tensors in x, y, and z axle; P is the pore pressure within the hydraulic fracture; r, r 1 , and r 2 are the distance between the spatial point to the centre, top, and bottom of the hydraulic fracture; and θ, θ 1 , and θ 2 are the angle between the spatial point to the centre, top, and bottom of the hydraulic fracture with respect where POR is the porosity; PERM is the permeability, μm 2 ; GR ′ is the normalized gamma-ray, which is calculated by ð GR − GR min Þ/ðGR max − GR min Þ; and RT ′ follows the same calculation. Figure 7 shows the detailed view of logging response and interpretation results of well w155. It is shown that the calculated porosity and permeability match well with the physical properties of core samples. The Magnetic Resonance Imaging (NMI) logging displays a feature of the bimodal-pattern, indicating that the pore structures contain both the macropore and micropore structure within the reservoir. Normally, the movable oil is filled in the macropore while the irreducible water is filled in the micropore. Moreover, the interpretation is based on the comprehensive analysis of flooding tests for core samples. The blue color represents the state of severely waterflooded, whereas the red one denotes the oil layer without waterflooded. Thus, the severely waterflooding layer usually has a relatively high porosity and permeability (Figure 7). Table 2 shows the statistics of hydraulic and mechanical parameters for each sand bodies. It is shown that the Poisson's ratio, Young's modulus, porosity, and permeability in six sand layers range from 0.18~0.39, 30.4~58.7 GPa, 0.01~0.222, and 0.09~111:6 × 10 −3 μm 2 , respectively. The 3D view for these parameters is shown in Figure 8. This heterogeneity of hydraulic and mechanical properties in six sand layers further verifies the superposition pattern of multistage channels, indicating the possible asymmetric propagation of hydraulic fractures after performing fracturing operations.
The real-time treatment data are utilized to determine the fracture half-length by history-matching the net operational pressure during the fracturing stimulation. First, the treat-ment datasets are established, including the treatment pressure, slurry rate, and proppant concentration. Next, the heat conduction, formation parameters, and wellbore structure are introduced to simulate the stimulation process under subsurface conditions. The fracturing fluid and proppant types are then determined to calculate the closure pressure and net operational pressure. Finally, the history matching of net operational pressure is conducted to determine the propagation of hydraulic fractures in terms of fracture halflength. Figure 9 shows the simulated results of hydraulic fractures for well wj9061 after performing integrated fracturing for the whole C6 1 oil layers. It is worth noting that the hydraulic fractures fail to connect the C6 1-2-2 with C6 1-2-3 sand layers vertically, which is ascribed to the existence of the muddy interlayer between two layers. Moreover, the fractures propagated laterally with an asymmetrical pattern in each sand layer ( [4]). It is also shown that hydraulic fractures in the C6 1-2-2 and C6 1-2-3 sand layer extended further than that of C6 1-2-3 and C6 1-2-4 layer. This difference is attributed to the geomechanics variety for different sand layers ( Table 2).
The simulated hydraulic fractures can be further calibrated by the analysis of microseismic monitoring datasets. Table 3 shows the statistics of inferred hydraulic fractures, which are interpreted from the microseismic monitoring. It is found that the full length of hydraulic fractures ranges from 102 m to 217 m, with an average value of 178 m. The variety of east and west wing of hydraulic fractures further corroborates the asymmetric pattern. The height of hydraulic fractures has a variety of 4.2 m~16.0 m, indicating the possible irregular distribution of muddy or calcareous interlayers between different sand layers. The spatial propagation of hydraulic fractures is then determined from the comprehensive analysis.

Hydraulic-Natural Fracture Networks and Water
Breakthrough. The fracture networks are first determined by analysing the waterflooded state of coring wells. Figure 7 shows the distinctive waterflooded state in different sand layers of well W155. It is found that the severely waterflooded thin layer (1040.4~1044.6 m) is spatially adjacent to the inferred fractures (1038.7-1039.2 m) in the C6 1-2-3 layer. Depth /m Figure 9: The simulated hydraulic fractures of each sand layer in well wj9061.The asymmetrical hydraulic fractures are observed in these sand layers. Given that the vertical permeable features of each sand layer, this severely waterflooded state is ascribed to the vertical connection of fractures within the sand layers. Moreover, some features of injection-production responses also indicate the growing up of the fracture networks [21,22]. Figure 10(a) shows the water injection and production data of wells W6 and W7, respectively. After 45 months of production, the water cut of well W7 rose rapidly from 35% up to 90%, leading to the shut-in of production operations on 10 January 2012. Simultaneously, the water injection volume of the well W6 showed a sudden increase from 30 m 3 /d to 50 m 3 /d. The sharp increases in the water cut of production wells and injection volume of injection wells are possibly attributed to the fracture network that connect both wells. Specifically, the long-term water injection activates the potential fracture networks, leading to a rapid rise of water cut and a sharp decrease in oil production. Furthermore, the effective reservoir permeability reaches up to 21:1 × 10 −3 μm 2 , based on the interpretation of well test for an injection well (Figure 10(b)). This high permeability, in comparison with the average matrix permeability (1:2 × 10 −3 μm 2 ), indicates the potential existence of fractures developed in this well. The full length of inferred fractures is interpreted to be approximately 442 m, exceeding the potential extent of hydraulic fractures propagation and verifying the involvement of natural fractures. Moreover, the evolution features of the water absorption profile for injection wells illustrate the dynamic behavior of hydraulicnatural fractures. Figure 10(c) shows that the water absorption concentrated at a thin layer of C6 1-2-2 from the onset of water injection. The thickness of water absorption exhibited a decreasing trend along with the long-term water injection. This evolution pattern indicates that hydraulic fractures continued propagating within the sandstone layer until connect with the preexisting natural fractures, forming the hydraulic-natural fracture networks, which further affect the water breakthrough in each sand layer. Besides the network identification for single wells, some field tests could also provide insights into spatial distribution for the hydraulic-natural fracture networks. The tracer monitoring demonstrates the spatial distribution of potential combined fractures based on the concentration and flow velocity of the monitored tracer. Figure 10(d) shows the tracer monitoring results for three injection wells. It is found that the fluid from the injection wells could diffuse over thousands of meters to waterflood the production wells, with a dominant orientation of approximately NE 70°. Overall, the comprehensive analysis is conducted to integrate the results of injection-production response, well test analysis, absorption profile evolution, and tracer tests. Figure 11 illustrates the water breakthrough restrained by the hydraulic-natural fracture networks in the C6 1-2-3 layer.
It is shown that the water breakthrough lineaments follow the orientation of approximately NE 67°with the full length of combined fractures ranging from 292 m to 2,101 m. The water breakthrough originates from the injection wells with a high amount of injection volume. This large injection provides the driving force for the growing up of the combined fractures. Moreover, the water breakthrough usually occurs within the microfacies of distributary channels, probably owing to the relatively high permeability in this kind of microfacies. Overall, the water breakthrough pattern is primarily ascribed to the spatial distribution of the hydraulicnatural fracture networks and sedimentary microfacies in each sand layer.    Figure 11: The water breakthrough controlled by the fracture networks in the C6 1-2-3 sand layer. The light blue circle denotes the cumulative injection volume of injection wells. The short line overlapped the production wells show the hydraulic fractures. The long blue belt illustrates the combined hydraulic-natural fracture. The base map was the microfacies in the C6 1-2-3 sand layer. 12 Geofluids restrained by the distribution of both microfacies and fractures via the geostatistical method [23]. Figure 8 shows the 3D geological model for different properties within the C6 1-2-3 sand layer. The grid size of x, y, and z in the model is set with 5 m × 5 m × 0:1 m. The grid resolution meets the requirement of middle-late oilfield development for the sand layers. The microfacies model is in good agreement with the results of microfacies analysis (Figure 8(a)). Three colors in the map denote the distributary channel (orange), the overbank sand (green), and the interdistributary bay (blue). The porosity model and permeability model, constrained by the facies distribution, are shown in Figures 8(b) and 8(c). It is found that the average porosity and permeability of the distributary channel, overbank, and interdistributary bay are 15.8% and 5:1 × 10 −3 μm 2 , 13.6% and 2:58 × 10 −3 μm 2 , and 9.2% and 0:7 × 10 −3 μm 2 , respectively, in the C6 1-2-3 layer, consistent with the previous statistics ( Table 2). The Young's modulus and the Poisson's ratio are averaged to be 44.1 GPa and 0.245, respectively (Figures 8(e) and 8(f)). The water saturation model conforms to the pattern of combined fracture networks ( Figure 10). The water breakthrough follows the direction of NE 67°, also   13 Geofluids controlled by the spatial distribution of fracture networks (Figure 8(d)).

Single Sand Layer-Based Remaining Oil Development.
Based on the waterflooded state in coring wells (Figure 7), combined hydraulic-natural fracture networks (Figure 11), and the water saturation distribution in the geological model (Figure 8(d)), the spatial waterflooded state is determined within the sand layer, as shown in Figure 12. It is found that the water saturation gradually decreases in the direction perpendicular to the water breakthrough line. In addition, the remaining oil is mainly distributed between the water breakthrough lines, with surrounding water saturation lower than 40%. Therefore, the corresponding measures of reperforation and refracturing could be conducted to tap the remaining oil in the sand layer for particular wells. Figure 13(a) shows the waterflooded state in the vicinity of the well w9061 in the C6 1-2-3 layer. It is worth noting that the remaining oil possibly distributed between the water breakthrough lines of injection wells w607 and w906. The distribution of the remaining oil is attributed to the existence of the muddy or calcareous interlayers between the C6 1-2-2 and C6 1-2-3 layer, which prevent hydraulic fractures from connecting both layers after performing the integrated fracturing operations ( Figure 9). Hence, the refracturing operations in 2014 were performed to tap the remaining oil in the four single layers for the well wj9061. The newly simulated results of hydraulic fractures after the refracturing operations are shown in Figure 13(b). It is noted that the hydraulic fractures in the C6 1-2-3 and C6 1-2-4 layers propagate more distance than before ( Figure 9). Accordingly, the oil production of well wj9061 increased from the 0.97 t/d to 2.49 t/d, indicating that the remaining oil in the single layers was better tapped after the refracturing operations.
Similarly, the refracturing operations were applied to the other seven wells in the study area to tap the remaining oil of particular sand layers. Table 4 shows the comparison of well production performance before and after the refracturing operations. It is worth noting that the average oil production for eight wells increased from 0.61 t/d to 2.18 t/d. Simultaneously, the water cut increased slightly from 41.1% to about 45.5% on average [4,5]. This performance indicates that the remaining oil within the individual sand layer is well devel-oped via the refracturing treatments, which provides insights for further development of likewise tight sandstone reservoirs.

Conclusions
An integrated approach is proposed to tap the remaining oil in the individual sand layer during the late-stage development of tight sandstone reservoirs. A case study is utilized to demonstrate the effective applicability of this integrated method.  14 Geofluids

Data Availability
All data included in this study are available upon request by contact with the corresponding author.

Conflicts of Interest
The authors declare no conflict of interest.