Reservoir Modeling of Braided River Reservoirs Based on Geological Knowledge Database: A Case Study of P₁x Formation of the Daniudi Gas Field, Ordos Basin, China

Establishing and expanding quantitative geological knowledge databases for several reservoirs have attracted considerable attention in recent decades [1–6]. Several strategies have been used in studies on establishing geological knowledge databases via quantitative geology, such as the fine anatomy of reservoirs in dense well network development areas [7, 8], field outcrop anatomy [9–11], modern sedimentation anatomy [12], tank simulation experiments [13], 3D digital outcrop model [14–16], 3D seismic data analysis [3, 17], Google Earth satellite image analysis [18–21], and remote sensing image analysis [22]. Several studies have focused on the geological information database of braided river reservoirs in the Ordos Basin, for example, Zhang et al. [8], Shi et al. [7], and Xu et al. [19] for the Sulige gas field reservoir and Zhang et al. [23] for the Hangjinqi area, combined modern sedimentation based on Google Earth measurements, outcrop data, and dense well network anatomy to establish a quantitative geological knowledge database of braided river reservoirs. They established a microfacies model using the multiple-point geostatistical modeling method based on 3D training images of quantitative characterization and received rich results. However, most are limited to using traditional geological dissection methods in the process, and the geological knowledge databases associated with the Daniudi gas field have received less attention in previous research.

A sandy braided river is a vital terrestrial sedimentary system. However, due to its reservoir’s substantial heterogeneity, the sand body’s spatial stacking pattern and distribution are complex, making it challenging to establish a highly reliable 3D geological model of the braided river reservoir [24]. Academics have extensively studied modeling braided river reservoirs. Li et al. [25] used a plane facies-controlled method to establish a 3D geological model of the braided river, analyzed the development pattern of nonpermeable layers inside the reservoir, and predicted the distribution of interbeds. Ma et al. [26] integrated dynamic and static information to demonstrate the scale of different sand body levels through sand connectivity analysis to establish a 3D geological model. Chen et al. [27] restricted the braided river reservoir geological model by collecting genetic and evolution information from well and seismic data and identifying genetic unit types and features. Wu et al. [28] established a spatial spreading model of a braided river reservoir consistent with architectural knowledge by identifying and predicting the constitutive unit’s scale and using a hierarchically restricted embedded modeling technique.

In the Lower Shihezi Formation of the Daniudi gas field in Ordos Basin, it develops a braided river sedimentary system, making it challenging to portray this area’s braided river sand body and bringing uncertainty to the reservoir’s 3D geological modeling because of its frequent migration and oscillation of braided channels and its large well spacing. Therefore, we integrate multisource and multiscale information from similar modern deposits, digital outcrop profiles in the field, and tank simulation experiments to establish a quantitative geological knowledge database for the braided river reservoir in the Lower Shihezi Formation H1 member of the Daniudi gas field. We rely on multiscale data to synthesize the empirical formulas for the scale of various microfacies units and on subsurface well data from the research area to compare the present with the past. The object-based method is employed to establish a 3D training image that conforms to the study area’s geological characteristics based on the knowledge database of quantitative parameters. Furthermore, the multiple-point geostatistical modeling method is applied to establish a sedimentary facies model that conforms to geological understanding laws and deposition patterns, reducing model uncertainty and laying a solid foundation for forecasting the remaining gas in the Daniudi gas field.

The Ordos Basin is in North-Central China (Figure 1(a)). The Ordos Basin is a large multirotational craton-margin stacked basin with depressional migration, subsidence, and tectonic simplicity [29]. It is divided into six primary tectonic units: the Yimeng Uplift to the north, the Weibei Uplift to the south, the Western Thrust Belt and the Tianhuan Depression to the west, the Jinxi Flexural Belt to the east, and the Yishaan Slope in the center [30]. The Daniudi gas field’s regional tectonics is in the north-eastern part of the Yishaan Slope, a secondary tectonic unit of the basin [31] (Figure 1(b)). The surrounding sedimentary outcrop is approximately 130 km east of the Daniudi gas field and tectonically to the north of the Jinxi Flexural Belt in the basin [32] (Figure 1(c)). Regarding the target reservoir, depositional environment, and reservoir features, the Fugu Tianshengqiao profile is comparable to the Daniudi gas field. The section’s primary body is northeast-southwest trending, with good overall outcrop conditions that are easy to observe, describe, and sample for analysis.

The Permian Lower Shihezi Formation is divided into three members from the bottom to the top: H1 (P₁x¹), H2 (P₁x²), and H3 (P₁x³). H1 is the primary gas-bearing section with undeveloped fractures in the target layer and gentle tectonics—a gentle high northeast and low southwest monocline [31, 34]. The local development of a nose uplift with a near east–west trend does not constitute a larger tectonic enclosure. The Shihezi Formation of the Daniudi gas field has braided river deposits with a floodplain and braided river subfacies. In the braided river system, the braided channel subfacies are the primary facies of sandstone deposition, with two microfacies evolved, namely, the channel bar and the braided channel between the channel bars (Figure 2). Sand conglomerate, pebbly sandstone, fine sandstone, siltstone, and mudstone dominate the lithology, and the burial thickness is 2540–2840 m.

This study focuses on the H1 member of the Shihezi Formation in the Daniudi gas field, one of the most gas-rich areas. Currently, approximately 296 wells have been drilled, covering approximately 188 km². The density is 0.635 km²/well, and the distance between the wells is approximately 600 m. We obtained the data related to exploration and production, such as core, logging curve, mud logging, seismic, sand interpretation, and other data types, from the Research Institute of Exploration and Development of North China Petroleum Bureau of Sinopec.

First, 9 typical sandy braided rivers were measured in planimetry using Google Earth software, and 10 sets of planimetric data were measured separately for each braided river, including the channel bars’ lengths and widths, the braided channels’ widths, and the braided rivers’ widths. Second, 13 sets of profile data were measured for the target layer H1 using the digital outcrop dissection of Fugu Tianshengqiao, including the channel bars’ and braided channels’ widths and thicknesses. Finally, the results of 28 tank simulation experiments conducted at the Key Laboratory of Tank Simulation at Yangtze University enabled the counting of 84 groups of stable channel bar length and width data and 82 groups of detectable braided channel width data on a plane. From the 28 tank simulation experiments, 86 sets of channel bar width and thickness data and 91 sets of braided channel width and thickness data were statistically recorded in 12 tangential source profiles at various positions from the source.

Tank simulation experiments were completed in the Key Laboratory of Tank Simulation at Yangtze University. The research object is the sandy braided river deposits developed in the Daniudi gas field, Ordos Basin. The experimental site is 10 m long, 2.2 m wide, and 2 m deep. It primarily comprises a testbed, sediment screening equipment, constant speed water supply, sand supply equipment, a 3D laser scanner, and video recording equipment. Considering the depositional law of a braided river reservoir, experimental purposes, and measured data from the mainstream of the Yellow River’s Huayuankou section with comparable depositional dynamics, it is demonstrated that the depositional slope of the braided river is 0.0267 m/10 m, the median grain size of the sediment is 0.2 mm, and the average velocity is 0.4–0.8 m/s. The primary tank simulation parameters were determined using the hydrological data from the Yellow River’s Huayuankou section. The slope was 0.1 m/10 m, and the median grain size of the sediment was 0.3 mm (the sand composition comprised green and yellow sand in a 3 : 1 ratio). The average velocity was 0.25 m/s.

The geological knowledge database-based braided river modeling proposed in this paper is based on a combination of similar modern sedimentation, field outcrop profiles, and tank simulation experiments to analyze the quantitative relationship between the microfacies of braided channels and channel bars in planes and profiles. Furthermore, the different microfacies thicknesses were divided in the wells in the study area. Establishing a quantitative geological knowledge database of the braided river reservoir’s microfacies was guided and applied to a 3D geological model of the H1 member to identify simulation parameters. The 3D training images were simulated using the object-based method, and a sedimentary facies model was built using the multiple-point geostatistical modeling method. The study area is divided into $25×25$ m grids in the plane and 0.5 m grids vertically, totaling 89,478,400 grids.

Google Earth software was used to screen nine typical modern sandy braided rivers with similar depositional backgrounds to the study area, including the Ningxia part of the Yellow, Lena, and Jamuna Rivers (Figure 3), and 10 sets of channel bar length and width and braided channel width data were measured for each river (Table 1). According to the statistics, the widths of sandy braided river channel bars are below 1500 m, with 16.67% exceeding 1500 m, and the average width is approximately 675 m (Figure 4(a)). The channel bar lengths are primarily less than 4000 m, with 18.89% over 4000 m, and the average length is approximately 1965 m (Figure 4(b)). The active braided channel widths are typically less than 300 m, with 20% surpassing 300 m, and the average width is approximately 195 m (Figure 4(c)).

Field outcrop profiling is the primary technical method for obtaining geometric shapes and scale data of braided channels and channel bars [36]. Even though outcrops are restricted and cannot provide a complete characterization of sedimentary structures (e.g., braided river width), they include a wealth of information on the architectural elements of braided river reservoirs [37]. The Fugu Tianshengqiao outcrop has the same target formation, depositional environment, and reservoir characteristics as the Daniudi gas field in this subsurface research region. The outcrop sections evolve from the bottom up in the Ordovician Majiagou, Carboniferous Benxi, Taiyuan, Permian Shanxi, and Lower Shihezi Formations [33]. The average elevation of the surface is 813 m, the height of the outcrop is 30–100 m, the primary body of the outcrop is trending northeast–southwest, the stratigraphic tendency is 280°, the dip angle is 4°–10°, and the sand body is visible in different directions, convenient for sand body lateral tracing and comparison and fine internal dissection.

The Fugu Tianshengqiao architecture’s anatomy is based on the digital outcrop model (Figure 6(a)), microfacies division (Figure 6(b)), and precise measurement of the channel bar and braided channel sands on the profile (Figures 6(c)–6(j)). For the target layer H1, 13 sets of channel bar width and thickness and braided channel width and thickness data (Table 2) were measured individually (the scale of the incomplete sand body was estimated for the outcrop). According to the outcrop data statistics, single minor channel bars are 30–300 m wide, whereas major composite channel bars are 300–1000 m wide and 6–21 m thick, with an average thickness of approximately 11 m, and only 7.69% of channel bars are more than 18 m thick (Figure 7(a)). The braided channels were 30–200 m wide and 4–14 m thick, with an average thickness of approximately 7.5 m, and only 15.38% of the braided channels were deeper than 12 m (Figure 7(b)).

Tank simulation experiments provide various reservoir sand models of varying depositional types [38–40]. These models resemble outcrops, modern sediments, and subsurface sediment bodies. The primary benefits include simple measurements (you can slice and sample at will), detailed records of depositional processes, and clear genesis mechanisms [41]. It is significant for sedimentology research and determining the reservoir’s macroscopic distribution pattern parameters. We simulate the deposition of a braided river in the Lower Shihezi Formation of the Daniudi gas field with the same parameters of sediment grain size, hydrodynamic conditions, and slope drop as the reservoir in the research area and a comparable depositional environment based on the principle of the present being a key to the past. The results of the tank simulation experiments in phase 18 (Figure 9(a)) were detrended to emphasize the shape and spreading features of the braided channel and the channel bar (Figure 9(b)) by integrating the 28-phase tank simulation process. As the experiments increased, the channel bar gradually changed from a long elliptical shape with a single style to a wide-short elliptical composite channel bar with a complex internal structure (Figures 9(b) and 9(c)). The parameters of a stable channel bar and braided channel in the experiments simulated after phase 11 were measured (Figure 9(b)), and 84 channel bar length and width data and 82 channel bar to braided channel width ratio data were statistically recorded on the plane (Table 3). Figure 9(d) shows the vertical accretion process of the composite channel bar, and Figure 9(e) shows the migration process of the braided channel. In total, 86 sets of channel bar width–thickness ratio data and 91 sets of braided channel width–thickness ratio data were statistically recorded (Table 4). In the profile, data on the composite channel bar and braided channel width–thickness were collected in phase 28 at 12 cut source locations.

According to the statistics of the tank simulation experimental planes and profiles, the stable channel bar’s width is between 20 and 60 mm, with an average width of 32 mm, and only 17.86% is wider than 40 mm (Figure 10(a)). The channel bar’s length was between 60 and 220 mm, with an average length of 107 mm, and only 15.48% were greater than 140 mm (Figure 10(b)). The active braided channel’s width was 2–9 mm, with an average braided channel width of 4.6 mm, and only 9.76% were wider than 7 mm (Figure 10(c)). The channel bar thickness was 2–14 mm, with an average of 6.3 mm, and only 10.47% exceeded 7 mm (Figure 10(d)). The active braided channel thickness was 4–20 mm, with an average braided channel thickness of 8.8 mm, and only 15.38% were more than 12 mm (Figure 10(e)).

Here, $lb3$ is the length of the tank simulation experiment’s channel bar (m), $wb3$ is the width of the tank simulation experiment’s channel bar (m), $wc3$ is the width of the tank simulation experiment’s braided channel (m), $tb3$ is the thickness of the tank simulation experiment’s channel bar (m), and $tc3$ is the thickness of the tank simulation experiment’s braided channel (m).

The primary development member of the Daniudi gas field, H1, is a sandy braided river system with box-shaped and bell-shaped gamma-ray (GR) logging curves produced in the primary part of the braided river (Figure 12). The bell-shaped GR log curve, with the bottom in abrupt contact with the underlying strata, relates to bottom scouring, and the top is in gradual contact with mudstone, reflecting progressively weakening hydrodynamic characteristics, and its corresponding sedimentary microfacies is a braided channel. The top and bottom of the smooth box-shaped GR log curve abruptly contact the mudstone, reflecting strong and stable hydrodynamic characteristics of sufficient material sources. The jagged box-shaped GR log curve reflects strong but unstable hydrodynamic conditions, with frequent alternations of strength and weakness, and its corresponding sedimentary microfacies is a channel bar. The channel bar’s thickness is 3–30 m, and the thickness of the braided channel is 1–22 m, according to the measurement statistics of the underground core well data of the Daniudi gas field.

Popular methods in reservoir geological knowledge database investigations include modern sedimentation, outcrop anatomy, and dense well network dissection. A comparison of empirical formulas for modern sedimentation, outcrop anatomy, and tank simulation data reveals that planimetric data based on modern sedimentation has the best correlation; however, measurement statistics have poor correlations due to the outcrops on the exposed surface affected by denudation, tank simulation experimental sites, and other factors. Therefore, the results of modern sedimentation, outcrop surveys, and tank simulation experiments are corroborated to establish a geological knowledge database for reservoir parameter prediction. All measurements were combined, and correlations between modern sedimentation and tank simulation experiments on planes (Figures 13(a) and 13(b)) and field outcrops and tank simulation experiments on profiles (Figures 13(c) and 13(d)) were fitted, respectively.

The measured data from the tank simulation experiments on planes and profiles are compared to planimetric data from modern sedimentation measurements and profile data from field outcrop observations. When relating the tank simulation experiments to modern deposition data on a plane, the quantitative relationship between the channel bar length and width is $lb=3.2083wb0.983$ (9), with correlation $R2=0.9432$—the results are highly correlated (Figure 13(a)). The quantitative relationship between the braided channel and channel bar width is $wc=0.1931wb1.0759$ (10), with a correlation of $R2=0.8458$⁠, which correlated well (Figure 13(b)). When relating the tank simulation experiments to field outcrop anatomical data on the profile, the quantitative relationship between channel bar thickness and width is $tb=0.0839wb0.8064$ (11), showing a good correlation of $R2=0.8407$ (Figure 13(c)). The quantitative relationship between river thickness and width is $tc=0.1938wc0.8201$ (12), a decent correlation of $R2=0.8279$ (Figure 13(d)). The correlations of all four data sets increased, indicating that the quantitative relationships obtained after integrating the data from modern sedimentation, field outcrop, and tank simulation experiments are real and reliable and can guide establishing a quantitative knowledge database of sandy braided river reservoirs in the Daniudi gas field.

Based on the anatomical understanding of the reservoir sand thickness in the Daniudi gas field, Ordos Basin (Figure 12), combined with a similar modern deposition of the braided river, field outcrop, and tank simulation experiments, a quantitative equation of parameters was fitted (Figure 13). Each sedimentary microfacies unit’s scale of parameters in the study area, such as thickness, width, and length, was determined, and the quantitative geological knowledge database of the sandy braided river in the Lower Shihezi Formation of the Daniudi gas field was established (Table 5). The braided channel has a thickness of 1–22 m and a width of 7–320 m. The single minor channel bar has a thickness of 3–9 m, a width of 80–330 m, and a length of 240–1000 m, and the major composite channel bar has a thickness of 9–30 m, a width of 330–1500 m, and a length of 1000–4200 m.

The multiple-point algorithm can realistically depict the morphology of complex geological bodies while remaining faithful to the original data [42]. The training image is used instead of the traditional variation function to express the correlation between multiple points in space when using the multiple-point geostatistical modeling method for simulation. The training images can show the real reservoir structure, geometry, contact relationships, distribution patterns, and geological knowledge database information in 3D space and reproduce geologists’ geological understanding of the research area. Creating training images includes hand-drawn, object-based simulation; 3D seismic information extraction or transformation; prototype-based models; process-based simulations; and 2D image methods [43]. However, the reliability of the training images limits the accuracy of the simulation results, with the primary issue being a scarcity of 3D training images appropriate for the research area.

A realistic and credible 3D training image is constructed using a quantitative approach that integrates modern sedimentation, outcrop profile, and tank simulation experimental database simulations with drilling data and using the object-based method (Figure 14). According to the statistics of subsurface well data from the Daniudi gas field, the training images include the floodplain (19.02%), the braided channel (36.15%), and the channel bar (44.83%). The generated training images can realistically reproduce a 3D morphology of the reservoir by describing the morphology of the sand body, such as its length, width, thickness, and quantitative correlations. 3D training images show the migrating superposition variants of braided channels and channel bars in the plane and longitudinal directions.

The established geological model can only be close enough to the real characteristics of subsurface sedimentary bodies by applying the quantitative geological knowledge database of braided rivers established in the previous stage to a 3D geological modeling of the Lower Shihezi Formation in the Daniudi gas field. Through the creation of 3D training images, the multiple-point geostatistical modeling method defines the model’s reservoir distribution and sand body structure and achieves results that characterize the actual geometry and quantitative relationships of the subsurface sediment bodies (Figure 15). The results of the model simulated using multiple-point geostatistical method are compared with the hard data in the well, and the well conditioning can reach 100%.

• (1)

  The frequency and distribution intervals of various sedimentary microfacies parameters of a braided river reservoir were analyzed using modern statistical sedimentation, field outcrop, and tank simulation experiments. By fitting the relationship between the parameters, the length and width of the channel bar, the width of the braided channel and channel bar, the thickness and width of the channel bar, and the thickness and width of the braided channel positively correlated to exponential power relationships

• (2)

  A quantitative knowledge database of braided river sedimentary microfacies applicable to the Lower Shihezi Formation of the Daniudi gas field was established, with single-phase braided channel thicknesses of 1–22 m, widths of 7–320 m, and channel bar thicknesses of 3–30 m, widths of 80–1500 m, and lengths of 240–4200 m

• (3)

  For the Daniudi gas field, a 3D geological modeling method is proposed that is compatible with the advantages of object-based and multiple-point methods. The established 3D training images characterize the depositional features of the braided river reservoir and the results of geological knowledge database research, allowing the final geological model results to characterize the real geometry and quantitative relationships of subsurface sedimentary bodies

No additional data are available.

The authors declare no conflict of interest.

This work was supported by the National Natural Science Foundation of China (No. 42172172 and No. 41802123).