The HFTS-1 site, which is operated by Laredo Petroleum, is located in the Midland Subbasin. It is divided into two phases: late 2015 to mid-2018 and mid-2018 to late 2020. HFTS-1 implements various techniques, such as different spacings, cluster numbers, proppant particle sizes, and liquid sand strengths. It employs several methods to gather direct information on artificial fracture and proppant placement. Additionally, cores from artificial fractures were collected to enhance the understanding of artificial fracture initiation, proppant behavior, and the correlation between fracture initiation and proppant propagation (Ciezobka and Reeves, 2020; Reeves et al., 2020; Courtier et al., 2017).

The HFTS-1 experimental well group consists of 2 wells, 11 newly drilled horizontal wells, and 2 large scientific wells. The 11 newly drilled horizontal wells are situated in small layers A and B of the Wolfcamp Formation, with a staggered well spacing of 660 feet, an inclined well spacing of 400 feet, and a vertical distance of 2 in layers A and B (80 feet). These wells are completed with 5-inch half-casing, fracturing 450 sections, and zipper fracturing in groups (Li et al., 2019) (refer to Figure 2). After the entire well group is fractured, it is put into production. After 3 months of production, a large-slope scientific well was drilled to obtain core samples from the artificial fracture distribution area, resulting in 850 feet of the core. All the cores were subjected to CT scans to manually identify the fracture types. Additionally, microscopic examinations of all the materials in the cores were conducted to determine the proppant (Ciezobka et al., 2018; Maity and Ciezobka, 2022; Maity and Ciezobka, 2019b).

HFTS-1 employs a variety of monitoring methods, including horizontal toe-end manometers, permanent optical fibers, downhole microseismic inclinometer sensor arrays, eight groups of pressure gauges for large-inclination scientific wells, three types of tracer delivery and testing of oil/water/proppant, and wellbore logging and production logging (Ciezobka et al., 2018; Kumar et al., 2020). Moreover, during the production of the well group, experiments are carried out to examine artificial fracture spreading, proppant deposition, interwell disturbance, and connectivity. The valuable insights obtained from these experiments challenge the prevailing notion of unconventional oil and gas fracturing technology, including artificial fracture propagation, proppant placement effectiveness, and their changes over time (Stegent and Candler, 2018; Rassenfoss, 2018).

Based on the understanding of the HFTS-1 results and unresolved issues, HFTS-2 began operation in November 2018 in Block 55 in Lowean County. Occidental and shell are the operators of the HFTS-2 (Ciezobka, 2021).

The HFTS-2 experimental well group comprises 2 existing wells, 8 newly drilled horizontal wells, and 2 scientific wells, with a spacing of 660 feet between each horizontal well. The technical measures and analysis tests implemented in HFTS-2 have been enhanced compared to those implemented in HFTS-1. These improvements include drilling a scientific well with 540 feet before fracturing and drilling a large, deviated scientific well. After fracturing, permanent fiber optic tubes were installed in 2 horizontal wells and 1 scientific well. Additionally, permanent pressure sensors were installed in 2 scientific wells (8 groups and 12 groups) and 2 scientific wells to improve the understanding of the effectiveness and influencing factors of fracturing measures (Bessa et al., 2021).

A significant difference between HFTS-2 and previous experiments is the installation of advanced permanent fiber monitoring array technology (FO array) in three wells (two horizontal wells and one vertical well) for the first time in North America. This FO array enables the monitoring of artificial fracture propagation and reservoir production in both spatial and temporal dimensions (Zhao et al., 2021). The FO array is instrumental in characterizing the SRV and reservoir production, calibrating subsurface models, and enhancing monitoring of artificial fracture opening, propagation, and proppant placement (Tan et al., 2022).

Based on full-diameter core CT scanning and modeling, HFTS-1 experiments were used to reconstruct the artificial fracture network within the coring range of large-slope scientific wells. When all 11 newly drilled horizontal wells are well arranged in the azimuth direction, the extension direction of most artificial fractures is not parallel to the direction of the minimum horizontal principal stress (SH_min). Instead, they extend in two directions with positive/negative acute angles to the SH_min. The first group of fractures, with a total of 197 fractures, was the most common and trended NE‒SW. The second group of fractures, with a total of 112 fractures, trended WNW‒ESE. Figure 4A shows the two sets of fractures in Section 25 of Well 6U, while Figure 4B shows the fracture azimuth measurements of all the fracturing sections of the 11 wells obtained by microseismic methods. The orientation of the main fractures was consistent in all wells at N76°E, and the secondary fracture orientation was clear in most sections of the wells, which was approximately N46°W (Maity and Ciezobka, 2021a). A similar situation was observed in the cores obtained from the HFTS-2 experiment (Shahri et al., 2021; Fu et al., 2020).

The results of hydraulic fracturing experiments in different basins of the ConocoPhillips SRV, HFTS-1, and HFTS-2 in North America show that the artificial fractures formed by hydraulic fracturing mostly exist in the form of dense, parallel, or subparallel artificial fracture clusters. The characteristics of small spacing, parallel extension, and cluster clustering are commonly observed in hydraulic fracturing in shale reservoirs (Raterman et al., 2017; Fu et al., 2020; Ugueto et al., 2022; Maity et al., 2018). Additionally, there is a strong correlation between the significant change in the trend of artificial fractures and the significant change in rock properties. Local stress interference can also cause deviations (Guerrero et al., 2022; Maity and Ciezobka, 2020; Zhao and Gray, 2021; Cao et al., 2022).

In the ConocoPhillips SRV experiment, a total of 2,466 artificial fractures were discovered in the S3 well and its three sidetracked horizontal wells along the P3 direction, with an average density of 2.6 feet per well. It should be noted that due to the coring of the four wells, including S3, around the P3 well, some fractures may have been repeatedly cored and observed.

In the HFTS-1 experiment, manual observation revealed 361 artificial fractures in the 600-foot core obtained from the high-inclination well, with an average density of 1.6 feet per piece. Similarly, in the HFTS-2 experiment, 500 artificial fractures were identified through manual observation of the acquired 948-foot core, with an average of 1.9 feet per piece (Joseph et al., 2022). The average spacing between these artificial fractures ranged from 1 to 2 feet. Simulation studies also indicated a significant correlation between the distribution of artificial fractures and perforated holes, which was previously unexpected (Maity, 2018). Moreover, despite the significant differences in reservoir conditions and natural fractures at the three experimental sites, the consistent results of the three experiments suggest that high-density parallel artificial fractures may be common in hydraulic fracturing (Fu et al., 2020; Joseph et al., 2022).

The effectiveness of artificial fractures and seepage after fracturing in Well P3 was identified by the ConocoPhillips SRV experiment. The decrease in reservoir pressure was primarily influenced by the distance of the nearest main fracture, although the frequency and location of these fractures were not uniform. Over the course of 4 years of production, most manometers exhibited a downward trend, except for one manometer located near the main fracture, which showed an increase over time, indicating a disconnection from the production well (Raterman et al., 2019).

Although a large number of artificial fractures were discovered in the ConocoPhillips RV experiments, only a small number of these fractures were effective for oil and gas seepage. During 2 years of production, the downhole pressure gauges in the production and observation wells showed significant variations in pressure drop at different positions within the reservoir, indicating severe unevenness in reservoir production. This finding aligns with the results obtained regarding the effectiveness of artificial fractures.

The results of the HFTS-1 experiments indicate that the key factors influencing the interaction of natural fractures are their direction, opening degree, texture and composition of the cemented filling, and filling degree relative to artificial fractures. An analysis of in situ stress and fracture properties along the coring section revealed that approximately 52% of the natural fractures in section A of the core remained closed and did not interact with artificial fractures. In contrast, approximately 25% of section B and 36% of section C interacted with artificial fractures. This can be attributed to the greater number of opened natural fractures in the B and C cores under stress, which aligns with the observed distribution of tetragonal calcite (Maity and Ciezobka, 2019a), as shown in Figure 5.

Although reservoir coring allows direct observation of artificial fractures after fracturing, determining the initiation and propagation of near-wellbore (NWB) fractures, such as the cluster efficiency and distribution of perforating hole fluid, is challenging. Optical fiber monitoring technology, applied in the HFTS-1 and HFTS-2 experiments, provided real-time temperature and acoustic measurement data near the wellbore, enabling accurate determination of the fluid distribution in the NWB perforated holes (Liu et al., 2021a). Analysis of the microseismic data revealed that the number and spacing of different perforation clusters had a significant effect on the NWB fracture geometry of the experimental well (Stegent and Candler, 2018). As shown in Figure 6, the growth behavior and final geometry of the fractures were consistent across the horizontal section of well 4U, regardless of the cluster design and number of perforations.

The ConocoPhillips SRV experiments employed cross-linked jelly for sand transport, but post-fracturing coring revealed that the farthest core sample with proppant was within 75 feet from well P3. Proppant was detected in 76% of the cuttings in the STO3 section (closest) coring barrel, which was situated 56 feet from well P3. However, only 5% of the fractures contained proppant, whereas 5% of the cuttings in the STO1 coring barrel were located 100 feet above well P3 (Raterman et al., 2017).

The HFTS-1 experiments provided insights into the vertical distribution depth of proppants in the UWC formation. The majority of proppants were located approximately 30 feet above and below the horizontal section of well W4 and extended laterally up to 80 feet. The distribution of proppants exhibited abrupt jumps and rapid declines (Elliott and Gale, 2018). In cases where the artificial fracture overlapped with lithological changes, the geometric morphology of the fracture wall underwent significant changes, resulting in a peak of proppant settlement. This demonstrates the influence of geological, lithological, fracture, and stress changes on the proppant distribution, as illustrated in Figure 7.

In relation to the proppant distribution after fracturing in horizontal wells, the ConocoPhillips RV, HFTS-1, and HFTS-2 experiments consistently demonstrate that the proppant distribution is highly uneven in artificial fracture systems. As observed in the core tube and core samples, the amount of proppant in artificial fractures dramatically decreases with increasing distance from the horizontal wellbore. Proppants are rarely observed in core barrels and cores located 100 feet away from horizontal wellbores (Raterman et al., 2017; Raterman et al., 2019; Maity and Ciezobka, 2021b; Kumar et al., 2021). Moreover, the distribution range of small particle-sized (100 mesh) proppant is significantly larger than that of medium particle-sized (40/70 mesh) proppant, with small particle-sized proppant dominating the longitudinal distribution of fracturing wells. Additionally, the concentration of the proppant distribution within segments presents a distinct heel-end advantage (Maity and Ciezobka, 2022; Wood et al., 2018).

In the HFTS experiment, a systematic analysis of artificial fractures and associated proppants was conducted. The main factors affecting proppant transportation and laying include artificial fractures, changes in lithology, and stress variations in the reservoir (Cao et al., 2022). The presence of artificial fractures in areas with lithology and stress changes often results in variations in fracture surface roughness and flatness. These changes can lead to proppant blockage and settlement in localized areas, preventing further transport to distant locations (Elliott and Gale, 2018; Maity and Ciezobka, 2021c). The experiments revealed a strong causal relationship between fracture surface roughness and the proppant distribution. Over time, artificial fractures located far from rough areas of the fracture wall may contribute less to productivity due to poor connectivity caused by local proppant settlement, resulting in unsupported fractures and limited conductivity for complete production (Gallardo Giozza et al., 2020).

The HFTS-2 experiment was conducted in the vicinity of Well B3H. The proppant placement, as observed through distributed fiber optic monitoring of the fracturing operation, correlated well with the proppant distribution observed in the core. These results were consistent with the pressure drop observed by a large-slope scientific manometer (Gale et al., 2021; Stegent et al., 2019). Therefore, it can be suggested that “primary” artificial fractures, connected to the main artificial fractures and filled with proppant, provide better and greater production over the lifespan of shale wells than “secondary” artificial fractures.

HFTS-1 was used to investigate the effect of proppant dosage on the EUR in a single well during fracturing construction. The EUR of a single well showed a significant positive correlation with the increase in proppant intensity used in layer A of the Wolfcamp Formation when the proppant amount reached a certain value. However, the increase in proppant dosage in layer B of the Wolf Formation did not affect the EUR of a single well (Li et al., 2019). This suggests that the strength of the proppant should match the reservoir’s characteristics and be determined experimentally for different reservoirs (Ruud et al., 2020).

The ConocoPhillips SRV experiment utilized pressure changes to assess the degree of reservoir utilization. Reservoir production was found to be severely heterogeneous (Raterman et al., 2017; Raterman et al., 2019; Wang et al., 2019).

In HFTS-1, 8 sets of high-precision manometers were installed in the 6 TW high-inclination coring well, and pressure data were recorded during the 18-month production period of the experimental well group, as shown in Figure 8. The trends of the G5 and G6 manometer data aligned with the toe-end manometer data of production wells 6U and 6 M, with slight differences due to the vertical height variation. The G7 manometer data exhibited a similar trend but with a smaller decrease. The G8 manometer, located farthest from the production well, showed relatively stable pressure. However, pressure gauge G4, situated between the 5U, 6U, 5 M, and 6 M production wells, exhibited a lower pressure drop rate in the reservoir than did the G5 and G6 manometers, with different trends (Li et al., 2019). The manometer data from well 6 TW demonstrated that the pressure decline trend varied at different positions within the reservoir, highlighting the level of impact of artificial fractures and the production distance from the fracturing well reservoir (Liu Xinghui et al., 2021; Gale et al., 2019).

For HFTS-2, 12 sets of high-precision manometers were installed in the high-inclination coring well B6S. These manometers recorded the pressure data of the experimental well group during the 2-year production period, and the reservoir pressure changes obtained were found to be similar to those observed in HFTS-1.

After collecting the monitoring data and employing digital simulation technology for analysis, the monitoring results of reservoir production following fracturing and production experiments were consistent. The key findings from this analysis are as follows.

Despite the lengthy production time (2 years) and successful production of the densely superimposed tridimensional development horizontal well group, there were notable variations in the pressure drop among the different reservoir areas. These differences in reservoir production indicate that the artificial fracture coverage rate and effectiveness have not been fully optimized. Therefore, further study is required to improve the reservoir production rate (Raterman et al., 2017; Maity and Ciezobka, 2022; Courtier et al., 2017; Bessa et al., 2021; Zhao et al., 2021).

The pressure data revealed that the depletion rate increases significantly when the reservoir pressure drops below the bubble point pressure. Additionally, the complexity of the contribution of artificial fractures to production was evident from the pressure data. The pressure in the reservoir area located far from the root end fluctuated multiple times, while the pressure drop was relatively stable in the region near the root end. This suggests that liquid production in each region can be discontinuous and dynamic [(Zhao et al., 2021); (Muralidharan and Esmaili, 2021); (Yu et al., 2021); (Stegent et al., 2019)].

The layered recovery factor decreases with increasing vertical distance from the horizontal wellbore. Therefore, carefully selecting the wellbore trajectory and understanding the reservoir stress profile are crucial factors for maximizing the degree of production in thick shale oil reservoirs (Muralidharan and Esmaili, 2021; Akash et al., 2020; Morris et al., 2019; Rafiee and Grover, 2017).

The exploitation coefficients of each small layer, as determined by the numerical model, varied based on the artificial fracture network and reservoir properties. Moreover, vertical artificial fractures and pressure drops do not align with the rectangular shapes set in the numerical model. Thus, employing uniform geometry for fracture modeling differs significantly from the actual situation. This incomplete understanding of reservoir production and the challenges in optimizing the oilfield development scheme are consequences of these modeling discrepancies (Baig et al., 2021; Cao et al., 2017).

The deployment of monitoring wells is considered the most effective method for understanding reservoir production levels. This is due to the correlation between well pressure logging data and microseismic interpretation. It is also a cost-effective means of predicting reservoir production (Kumar Abhash et al., 2018; Wicker et al., 2018; Wang et al., 2018).

Microseismic event clouds are often perceived as the “outer boundary” of the “stimulated reservoir volume” (SRV). However, analysis of the HFTS-1 data revealed that microseismic clouds do not accurately reflect the range of artificial fractures but rather indicate the extent of naturally reactivated fractures (Gross and Dershowitz, 2023). Further analysis of the HFTS-2 data suggests that the microseismic event cloud represents only a subset of the SRV (Tan et al., 2021; Chen et al., 2022; Shrivastava et al., 2018; Weddle et al., 2018).