Geologically Constrained Optimization of Horizontal Well and Fracture Design in Tight Sandstone Reservoirs: Insights from the Chang 7 Member, Ordos Basin

Abstract

Efficient development of tight reservoirs in shallow-water delta-front environments is often constrained by misaligned horizontal well design and the underlying geological architecture. To address this, a quantitative optimization workflow is proposed, integrating 3D architectural characterization of single sandbodies with reservoir simulation. Using the Chang 7 Member of the Ordos Basin as a case study, three dominant sandbody types—isolated channels, vertically stacked channels, and mouth bars—were characterized in terms of geometry, stacking pattern, and internal permeability anisotropy. High-resolution geological models incorporating stratigraphic cyclicity and heterogeneity were constructed. Local grid refinement around wellbores and fracture networks was implemented to improve simulation fidelity. Sensitivity analyses identified optimal values for horizontal section length, fracture stage, and fracture half-length for each sandbody architecture. The results indicate that production response is highly sensitive to sandbody geometry and heterogeneity, with diminishing returns observed beyond critical design thresholds. Field validation with three horizontal wells confirmed that optimized parameter sets aligned with geological architecture resulted in significantly improved and more stable oil production. To support application in similar reservoirs, a dimensionless design chart was developed using ratios of horizontal well length to sandbody length (L_h/L_s) and fracture length to sandbody width (L_f/W_s). This empirical tool enables rapid pre-drill assessments and informs well planning strategies aligned with sandbody architecture. By emphasizing the integration of geological and engineering disciplines, the approach offers a scalable framework for optimizing horizontal well design in geologically complex tight formations.

1. Introduction

Over the past two decades, tight oil and gas resources have become critical supplements to conventional hydrocarbons [1,2,3]. This shift is largely driven by advances in hydraulic fracturing and horizontal drilling, which have unlocked previously inaccessible reserves [4,5]. As a result, the large-scale and economically viable development of these resources has become a strategic priority for the global energy sector. Countries such as the United States and China have responded with heavy investment in unconventional resource development [6]. According to China Energy News, in 2024, China’s tight gas output surpassed 60 billion cubic meters, comprising over 60% of the country’s unconventional natural gas production [7], and positioning China as the global leader in tight gas output [8].

Horizontal drilling combined with multi-stage hydraulic fracturing has been pivotal in unlocking tight oil and gas reserves [9,10,11,12,13]. The successful execution of horizontal well depends critically on real-time monitoring and control during drilling and fracturing operations. Smart drilling technologies have enabled more precise assessment and mitigation of downhole vibrations and shock loads, which are particularly important in heterogeneous tight formations [14,15]. In many unconventional plays across North America, horizontal wells now contribute over 70% of total hydrocarbon production [16]. However, their performance depends not only on reservoir potential but also on the optimal configuration of engineering parameters—such as horizontal section length, fracture spacing, and stage count [17,18]. Poorly optimized designs can lead to inefficient reservoir stimulation and accelerated production decline [19]. Therefore, parameter optimization is essential for maximizing the recovery and long-term productivity of tight reservoirs [20].

Conventional approaches to horizontal well optimization fall into two main categories: empirical methods and numerical simulations [21]. Empirical approaches use field production data to build statistical correlations between design parameters and well performance [22]. While simple and cost-effective, they often lack geological specificity. Numerical simulations, typically conducted using finite element platforms, offer more detailed modeling but often assume homogeneous reservoir properties [23,24,25]. This simplification limits their applicability in tight reservoirs characterized by strong heterogeneity and complex internal architecture [26]. As a result, there is growing recognition of the need to incorporate detailed architectural data—such as sand body geometry and stacking patterns—into horizontal well optimization workflows. Recent advancements in horizontal well design have increasingly emphasized the role of reservoir architecture [27]. Aligning well trajectories with the orientation of high-quality sandbodies enhances reservoir contact, while fracture placement guided by internal connectivity improves drainage efficiency and stimulation coverage [28]. Architecture-constrained optimization offers superior geological realism compared to conventional homogeneous models and has shown promising results in improving production from tight formations.

However, most existing applications of architectural integration focus on large-scale depositional units, such as sand groupings or stratigraphic sequences [29,30]. These approaches often overlook the internal variability and flow behavior governed by individual sandbodies—the smallest independent units of hydrocarbon storage and flow [31,32]. This mismatch can lead to suboptimal stimulation strategies and poor production performance. Because single sandbodies exhibit distinct geometry, stacking patterns, and internal heterogeneity, their accurate characterization is essential for designing effective well trajectories and fracture networks [33]. Bridging this gap requires a workflow that explicitly couples fine-scale geological architecture with engineering design to optimize horizontal well parameters in tight reservoirs.

In response to these limitations, this study develops an integrated geo-engineering workflow that explicitly couples fine-scale reservoir architecture with horizontal well design. A high-resolution three-dimensional (3D) geological model was constructed based on the quantitative characterization of individual sandbodies, capturing their geometry, spatial stacking patterns, and internal heterogeneity. This geologic framework was incorporated into a series of numerical simulations to evaluate the sensitivity of key engineering parameters—namely, horizontal section length, number of fracturing stages, and fracture half-length—across different architectural scenarios. By establishing quantitative relationships between sandbody architecture and production performance, the proposed approach enables architecture-constrained optimization of horizontal wells in tight reservoirs. This work offers a scalable methodology for improving well placement and stimulation design in sedimentologically complex, tight reservoirs.

2. Geological Settings

The Ordos Basin, located on the North China Platform, is the second-largest petroliferous basin in China, spanning approximately 370,000 km² [34,35]. Structurally, it is an asymmetrical, north–south-trending depression bounded by a gentle eastern flank and a steeper western margin. The basin comprises six secondary tectonic units, among which the Yishan Slope in the central–western region serves as the focus of this study. Within the Yishan Slope, the Chang 7 Member exhibits a gentle structural dip (<1°) with subtle nose-shaped uplifts that promote favorable hydrocarbon migration and trapping conditions [36]. These structural features, coupled with the spatial distribution of sandbodies, create stratigraphic-structural traps conducive to tight oil accumulation (Figure 1a) [37].

The Chang 7 Member belongs to the Upper Triassic Yanchang Formation and was deposited under a shallow lacustrine setting dominated by delta-front processes [38]. During this period, sedimentation was primarily controlled by progradational deltaic activity, leading to the formation of sand-rich facies such as underwater distributary channels and mouth bars [39]. Stratigraphically, the Chang 7 Member is subdivided into three sub-members—Chang 7₁, Chang 7₂, and Chang 7₃—which are further divided into six thin layers (Figure 1b) [40]. The maximum lacustrine transgression occurred during Chang 7₃, depositing widespread organic-rich shales at the base and mouth bar sandstones above [41]. In contrast, the Chang 7₂ sub-member records a regressive phase with enhanced delta-front progradation and the development of underwater distributary channel systems [42]. The final stage, Chang 7₁, marks a waning transgression, characterized by mixed deposition of delta-front and deep-water turbiditic sandbodies [38].

3. Methodology

3.1. Sandbody Architectural Classification

3.1.1. Types and Stacking Relationships of Sandbody

The delta-front subfacies of the study area are characterized by sand-rich depositional elements, primarily consisting of single underwater distributary channels, mouth bars, and sheet sands. These architectural elements vary in geometry, continuity, and stacking styles, influencing reservoir heterogeneity and production potential. Detailed layer-by-layer correlations based on high-density well log data enabled identification of two dominant sand body types within the Chang 7 Member: underwater distributary channels and mouth bars (Figure 2).

Based on spatial arrangement and vertical stratigraphic relationships, six representative sandbody architectural styles were classified: (1) isolated distributary channel (R-I), (2) vertically stacked distributary channel (R-II), (3) laterally amalgamated distributary channel (R-III), (4) isolated mouth bars (MB-I), (5) laterally amalgamated mouth bars (MB-II), and (6) channel-scoured mouth bars (R-MB) (Figure 3a). These classifications serve as the architectural basis for numerical modeling and optimization of horizontal well parameters (Figure 3a).

3.1.2. Target Area Selection for Horizontal Wells

Spatial mapping of sandbody architecture types was conducted through integration of single-well architectural descriptions and sedimentary microfacies maps. The lateral and vertical distribution of dominant architectural elements was identified across each thin layer, enabling precise delineation of horizontal well target zones.

Horizontal well planning in tight reservoirs requires sand bodies that satisfy specific engineering constraints: (1) thickness greater than 6 m to ensure reliable well landing [43], (2) lateral continuity exceeding 800 m to accommodate sufficient horizontal section length, and (3) location within structurally stable intervals with minimal faulting and gentle dip (<1°) to reduce drilling risk.

The study area meets these criteria, exhibiting simple structure and negligible fault development. Based on architectural analysis, the Chang 7₁¹, Chang 7₂¹, and Chang 7₃¹ were selected as priority zones. Within these intervals, laterally continuous underwater distributary channels and amalgamated mouth bars were identified as the most favorable targets for horizontal drilling (Figure 3b).

3.1.3. Quantitative Architectural Parameters

To support model construction and parameter optimization, statistical analysis was conducted on the 3D geometrical attributes of representative sand bodies, including length, width, and vertical thickness. Only bodies with thicknesses ≥ 6 m were included to ensure relevance to horizontal well planning. The measurement methodology is illustrated in Figure 4.

Due to its limited presence in the well pattern, the laterally stacked channel architecture (R-III) was excluded from statistical modeling. The channel-scoured mouth bar type (R-MB), in which the mouth bar is the dominant component, was grouped with the isolated mouth bar (MB-I) type for analysis. A total of 96 sandbody penetrations were used for statistical characterization. The sand bodies were categorized into four architectural classes—R-I, R-II, MB-I, and MB-II—accounting for 22.9%, 10.4%, 18.8%, and 47.9% of the total occurrences, respectively (

Table 1. Three-dimensional architectural parameters of the main sandbody types in the Chang 7 Member.

Type	Sandbody Architecture	L (m)			W (m)			H (m)			Frequency (%)
		Min	Max	Average	Min	Max	Average	Min	Max	Average
A	R-I	820	1400	1060	200	350	286	6.0	8.8	7.0	22.9
B	R-II	900	1100	1010	270	350	317	6.8	11.0	8.9	10.4
C	MB-I	1400	1500	1467	600	1100	866	6.5	9.5	8.6	18.8
D	MB-II	/	/	3500	/	/	1400	/	/	7.8	47.9

Note: L = planar length of the sandbody architecture; W = planar width of the sandbody architecture; H = vertical thickness of the sandbody architecture.

).

3.2. Physical Properties of Sandbody

3.2.1. Experimental Methods

To characterize pore structure and reservoir quality, Mercury Intrusion Porosimetry (MIP) was performed on 12 sandstone samples using the AutoPoRE IV 9500 mercury intrusion meter, following the industry standard of rock capillary pressure [44] measurement. The maximum mercury injection pressure reached 208 MPa, corresponding to a pore-throat radius of 0.003 μm. Pore-throat size distribution were derived using the Washburn model based on the pressure-volume data.

Furthermore, porosity and permeability measurements were conducted on 27 core samples, including 12 from underwater distributary channels and 15 from mouth bar sandbodies. Testing was performed using an AP-608 automatic porosity–permeability system based on the gas pulse decay method, consistent with SY/T 5336-2006 core analysis procedures widely used in the Chinese oil and gas sector.

3.2.2. Pore-Throat Size Distribution and Permeability

The MIP results indicate that sample displacement pressure average approximately 1 MPa, reflecting relatively tight pore systems (Figure 5a). Median pressure for several samples exceeds 10 MPa, corresponding to pore-throat radii of 0.07 μm, and suggesting pronounced intra- and inter-layer heterogeneity. Pore-throat radii across the dataset ranges from 0.01 to 1 μm, with a dominant peak observed between 0.1 and 0.4 μm. A weak bimodal distribution is observed in several samples, with tailing features near 0.01 μm, indicative of multi-scale pore networks and non-uniform pore systems (Figure 5b).

Well log interpretations based on gamma ray (GR), spontaneous potential (SP), and acoustic transit time (AC) curves reveal distinct depositional signatures for the two primary sandbody types. Underwater distributary channels display box- or bell-shaped log motifs, associated with upward-fining grain size trends and abrupt basal contacts (Figure 6a). Lithologically, these intervals consist of medium to fine-grained sandstone with occasional mud pebble lags at the base. Permeability measurements from 12 core samples show increasing values with depth, confirming a positive rhythmic sequence (P-rhythm) and improving reservoir quality toward the base.

In contrast, mouth bar deposits exhibit funnel-shaped or compound funnel-shaped log responses, reflecting upward-coarsening trends due to progressive sand supply (Figure 6b). These intervals are composed of siltstone to fine-grained sandstone with sedimentary structures such as cross-bedding, wavy bedding, and parallel lamination. Permeability profiles from 15 core samples reveal decreasing trend from the top to bottom, consistent with a reverse rhythm (R-rhythm) sequence and enhanced reservoir properties near the upper and middle sections.

Broader petrophysical characterization across 240 core samples shows that maximum permeability remains below 10 mD, with most values concentrated below 1 mD (Figure 7a). This confirms the tight nature and strong heterogeneity of the reservoir. A positive correlation between porosity and permeability is observed, with permeability increasing exponentially as porosity rises (Figure 7b). The average matrix porosity is 8.9%. Due to the low intrinsic permeability, large-scale hydraulic fracturing is essential for enhancing effective drainage and achieving commercial productivity.

3.3. Geological Model Development

3.3.1. Sandbody Geometry and Model Setup

Based on representative architectural types in the study area, three conceptual geological models were constructed to simulate reservoir behavior under varying internal geometries: Type A (isolated distributary channels), Type B (vertically stacked distributary channels), and Type C (isolated mouth bar). Type D, representing laterally amalgamated mouth bars, was approximated using Type C due to their similar depositional geometry and internal rhythm.

Each model incorporated realistic geometries derived from statistical analysis of well log and core data. Type A was configured with a flat top and convex base, characteristic of single-channel fills. Type B was generated by vertically stacking two Type A elements to simulate compound channel architecture. Type C was modeled with a convex top and flat base, consistent with upward-coarsening mouth bar deposits. All models exhibit maximum thickness along their central axes, thinning gradually toward the margins in accordance with depositional patterns observed in the field.

Average quantitative dimensions (length, width, and thickness) for each sandbody type were used to define the model geometry (

Table 2. Grid parameters used in the conceptual models of different sandbody architectures.

Type	Sandbody Architecture	La (m)	Wa (m)	Ha (m)	Planar Resolution (m)	Vertical Resolution (m)	Porosity (%)	Permeability (mD)	Total Number
A	R-I	1060	286	7	X = Y = 10	Z = 1	8.9	0.001~1.6	18,548
B	R-II	1010	317	5				0.001~1.6	19,696
				4				0.001~1.3
C	MB-I	1467	866	8				0.001~2.2	558,312

Note: L_a = planar length of the modeled sandbody; W_a = planar width of the modeled sandbody; H_a = vertical thickness of the modeled sandbody.

). To balance geological resolution and computational efficiency, the horizontal grid spacing was set to 10 × 10 m, and vertical resolution to 1 m. These discretization settings ensure accurate representation of vertical rhythmic layering and lateral architectural transitions necessary for subsequent flow simulation and well design optimization.

3.3.2. Model Heterogeneity

To realistically simulate flow behavior within each architectural type, spatial heterogeneity was incorporated in both the vertical and lateral directions of the permeability field. Laterally, permeability was modeled to decrease from the central axis of the sandbody toward the margins, constrained by well log correlations and depositional facies models. This approach represents a simplified constructive assumption intended to capture the primary effects of lateral heterogeneity while enabling manageable numerical simulation. This lateral trend reflects the decreasing grain size and compaction typically observed away from the main sedimentary conduit.

Vertically, permeability heterogeneity was assigned based on rhythmic stacking patterns directly supported by permeability measurements and log interpretation. Each architectural type exhibits a distinct internal permeability rhythm (Figure 8):

• Type A (R-I): Positive rhythm (P-rhythm), with permeability decreasing from base to top;
• Type B (R-II): Composite positive rhythm (CP-rhythm), represented by two vertically stacked P-rhythm sequences;
• Type C (MB-I): Reverse rhythm (R-rhythm), with permeability decreasing from top to bottom.

These rhythmic patterns were uniformly applied within each conceptual model, ensuring a consistent representation of internal flow anisotropy (Figure 9). The vertical permeability contrast between rhythmic layers was derived from averaged core measurements, allowing the models to preserve realistic geological variability while enabling sensitivity testing of engineering parameters under different architectural scenarios.

3.3.3. Local Grid Refinement

Capturing near-wellbore flow dynamics and fracture–matrix interactions required targeted grid refinement along the horizontal wellbore and fracture planes (Figure 10a,b). This was implemented using Petrel RE software (version 2009.1, Schlumberger, Houston, TX, USA). The grid surrounding the well path was refined to 3.3 × 3.3 m, while cells intersected by fracture planes were subdivided at twice that resolution. Vertical resolution remained at 1 m to preserve stratigraphic layering and rhythmic heterogeneity (Figure 10c).

A structured orthogonal grid framework was used to maintain computational stability and ensure alignment with geological layering. While actual fracture apertures are on the millimeter scale, refined grid blocks were configured to capture equivalent hydraulic conductivity and fracture volume within the limits of numerical resolution. This refinement strategy enabled detailed simulation of fluid exchange between fracture networks and matrix under varying architectural scenarios.

3.4. Engineering Design of Horizontal Well Parameters

3.4.1. Horizontal Section Length Scenarios

The design of horizontal section length was constrained by the lateral extent of individual sandbodies. For Type A and Type B architectures (distributary channels), feasible lengths were generally limited to less than 1000 m, while Type C architectures (mouth bars) permitted lateral lengths up to 1400 m. In all models, the horizontal section was placed along the central axis of the sandbody in the plan view to maximize reservoir contact. The orientation of the wellbore followed the principal depositional direction—parallel to the channel axis in Types A and B, and aligned with the long axis in Type C.

Simulations were conducted under constant-rate production over a 20-year period using a non-proppant fracturing model. This assumption was adopted intentionally as a simplified baseline to isolate the geometric effects of fracture half-length, fracture spacing, and horizontal section length from the additional uncertainties associated with proppant transport, placement efficiency, embedment, crushing, and time-dependent conductivity degradation. Because the primary objective of this study is to establish architecture-constrained relationships between sandbody geometry and optimal engineering parameters, the non-proppant assumption enables a clearer evaluation of fracture geometry effects across different architectural scenarios. In field applications, proppant-supported fractures may maintain higher conductivity and improve the effectiveness of distal fracture segments, which could modestly alter the preferred fracture spacing and effective fracture half-length. However, these effects are strongly completion-dependent and were therefore not explicitly included in the present baseline optimization framework. Horizontal section lengths were varied to capture performance sensitivity. Specifically, six scenarios were modeled for Types A and B (500–1000 m at 100 m increments), and eight scenarios for Type C (500–1200 m at 100 m increments), reflecting its greater lateral continuity.

3.4.2. Fracturing Stage Count Scenarios

Following the optimization of horizontal well length, the number of fracturing stages was evaluated for each sandbody architecture (Figure 11a). A key constraint was that fracture half-lengths remain within the width of the sandbody to avoid out-of-zone stimulation. Fractures were assumed to be uniformly spaced with a fixed half-length of 80 m.

• Type A: 4, 5, 6, 8, and 10 stages;
• Type B: 5, 6, 7, 9, and 11 stages;
• Type C: 6, 7, 8, 10, and 12 stages.

3.4.3. Fracture Half-Length Sensitivity Scenarios

Fracture half-length is a critical factor in stimulation effectiveness and was evaluated across different sandbody architectures following the optimization of horizontal length and stage count (Figure 11b). The design was constrained by the maximum lateral extent of each sandbody to prevent fracture breakthrough beyond the reservoir body. And this limit was derived from the average sandbody width statistics presented in Table 1, with a safety margin to account for geological uncertainty. The allowable fracture half-lengths were determined as follows:

• Type A: <140 m;
• Type B: <150 m;
• Type C: <400 m.

For each architecture, five scenarios were simulated with fracture half-lengths of 40 m, 60 m, 80 m, 100 m, and 120 m. This range was selected to assess production sensitivity and determine optimal fracture dimensions within geological boundaries.

4. Simulation Results and Analysis

4.1. Influence of Horizontal Section Length on Production

Pressure depletion around the horizontal wellbore is strongly influenced by the lateral boundaries of the sandbody. In Types A and B (underwater distributary channels), the proximity of sandbody edges results in more rapid formation energy loss (Figure 12a,b). In contrast, Type C (mouth bar) sandbodies, characterized by broader lateral continuity, experience a slower pressure drop and more sustained reservoir support (Figure 12c).

The relationship between production time and cumulative oil production (CQ) reveals that longer horizontal sections extend the stable production period across all architectures (Figure 12d–f). However, at later production stages, well performance declines, and the CQ growth rate flattens. For the same horizontal length, Type C sandbodies consistently yield higher ultimate recovery and longer stable flow durations due to their larger volume and sandbody connectivity.

The CQ response over 10-year and 20-year periods shows linear growth with horizontal section length up to a critical point (Figure 12g–i). Beyond this threshold, incremental gains decrease, indicating diminishing returns [45]. Taking into account drilling cost and marginal productivity, the recommended optimal horizontal section lengths are:

• Type A: ~600 m;
• Type B: ~700 m;
• Type C: ~900 m.

4.2. Effects of Fracture Spacing and Half-Length

Figure 13a–c shows that cumulative oil production (CQ) increases linearly with the number of fracturing stages up to architecture-specific thresholds: 6 stages for Type A, 7 stages for Type B, and 8 stages for Type C. Beyond these points, the CQ gain deviates from linearity and begins to flatten, particularly in narrower channel-type sandbodies. Type C demonstrates less pronounced tapering due to its larger effective stimulation area. Similarly, fracture half-length exhibited a linear relationship with CQ up to critical values—80 m for Types A and B, and 100 m for Type C (Figure 13b–d). Longer fractures yielded minimal additional benefit and risked extension beyond reservoir boundaries.

Considering both stimulation efficiency and economic returns, the following optimized designs are recommended:

Fracturing stages and spacing:

• Type A: 6 stages (~100 m spacing);
• Type B: 7 stages (~100 m spacing);
• Type C: 8 stages (~110 m spacing);

Fracture half-length:

• Type A and B: ~80 m;
• Type C: ~100 m.

These results provide a data-driven basis for optimizing hydraulic fracture parameters across different architectural scenarios (

Table 3. Optimized fracturing parameters for horizontal wells in different sandbody architectures.

Type	Sandbody Architecture	Sandbody			Horizontal Well
		La (m)	Wa (m)	Ha (m)	Horizontal Section Length (m)	Number of Fracturing Stages	Fracture Spacing (m)	Fracture Half-Length (m)
A	R-I	1060	286	7	600	6	100	80
B	R-II	1010	317	9	700	7	100	80
C	MB-I	1467	866	8	900	8	110	100

).

4.3. Field Engineering Validation

To validate the simulation-based design recommendations, three horizontal wells targeting Type C (mouth bar) sandbodies were selected from the study area.

• Well D1 intercepted the intended sandbody but employed a horizontal section significantly shorter than the recommended length. This limitation restricted the stimulated reservoir volume, resulting in average daily oil production of 4.5 tons (Figure 14a).
• Well D5 deviated out of zone during drilling, leading to poor sandbody contact and a reduced stimulation volume. Its productivity was markedly lower, averaging only 0.5 tons per day (Figure 14b).
• Well D12, in contrast, maintained lateral continuity within the effective sandbody and followed optimized design parameters. It achieved the highest and most stable performance, producing 5.2 tons per day on average (Figure 14c).

Production analysis (Figure 14d,e) confirms that alignment with architectural geometry is critical to maximizing horizontal well output and sustaining long-term productivity.

5. Discussion

5.1. Coupled Architecture–Flow Modeling Approach

This study presents an integrated methodology that combines high-resolution architectural characterization, core-calibrated petrophysical data, heterogenous geological modeling, and dynamic flow simulation to optimize horizontal well parameters in tight Chang 7 Member reservoirs of the Ordos Basin. Field validation, such as the performance of Well D12, confirms the effectiveness of architecture-constrained well design in achieving improved and stable production outcomes.

For the first time, this approach systematically incorporates three-dimensional architectural attributes—geometry, dimensions, stacking styles, and permeability anisotropy—into a dynamic simulation framework, using core-derived rhythmic sequences to reflect internal heterogeneity. Compared to conventional homogeneous models or those relying solely on porosity–permeability inputs, the coupled architecture–flow model offers enhanced geological realism and predictive capability. It also explains spatial variation in engineering optimization—for instance, why the optimal horizontal length may be 600 m in one sandbody and 900 m in another.

Nonetheless, the idealized radial flow assumptions and simplified internal structures in the current model may limit accuracy in forecasting long-term depletion behavior. It should also be noted that the optimization results were derived from conceptual models parameterized using representative sandbody dimensions and average permeability contrasts constrained by core measurements and logging interpretation. While this modeling framework captures the dominant architectural characteristics of each sandbody type, it inevitably simplifies the influence of local-scale heterogeneity. As shown in Figure 7, the permeability data exhibit considerable dispersion, indicating that actual reservoir conditions may deviate from the mean-case scenarios represented in the simulations. Where permeability contrast is stronger or lateral continuity is poorer than assumed, the effective stimulated reservoir volume may be reduced, and the optimal horizontal section length and fracture half-length may shift toward the lower end of the recommended ranges. Conversely, in intervals with more uniform permeability distribution and better connectivity, slightly larger design values may remain effective before diminishing returns become evident. Therefore, the optimized parameters proposed in this study should be regarded as architecture-constrained reference values for preliminary design, and further site-specific calibration is recommended when more detailed petrophysical and geomechanical data are available. Further refinement of flowfield geometry and internal facies complexity is necessary for enhanced predictive robustness.

5.2. Flow Behavior Regulated by Sandbody Architecture

Efficient development of tight reservoirs mainly relies on hydraulic fractures enhancing matrix permeability and connectivity. Initially, sparsely spaced fractures access unstimulated volumes, significantly boosting production [46]. However, as fracture density increases, overlapping drainage areas and stress interactions between closely spaced fractures reduce stimulation efficiency and can cause fracture redirection or reactivation [47,48].

Fracture behavior is further constrainted by reservoir architecutre. In narrow sandbodies, shorter fractures remain confined within high-permeability zones, contributing effectively to flow. In contrast, longer fractures may propagate beyond architectural boundaries or into low-permeability regions, resulting in ineffective extensions and reduced conductivity. This scenario is particularly problematic when the architectural configuration is not considered, as overstimulated non-reservoir zones can act as hydraulic short-circuits and accelerate production decline (Figure 14b). In low-permeability formations, the effective drainage radius is limited. As fracture half-length increases, marginal reservoir contribution declines due to pressure drop and poor reservoir quality at distal fracture tips [49].

These effects are quantitatively evident in the simulation outputs. Channelized architectures (Types A and B) achieve optimal performance with horizontal lengths of 600–700 m and fracture half-lengths around 80 m, while mouth bar systems (Type C) accommodate longer horizontal sections (~900 m) and fracture half-lengths up to 100 m, leveraging their broader continuity and reverse rhythmic heterogeneity.

Flow patterns are also architecture-specific. Underwater distributary channels (Type A) display upward-fining permeability vertically and lateral decline from axis to margins (Figure 15a). Meandering point bars show lateral permeability decrease from bar center to channel margin, influenced by lateral accretion surfaces (Figure 15b). In contrast, mouth bars (Type C) exhibit reverse rhythmic sequences with decreasing permeability from top to base (Figure 15c), while beach bars typically show more uniform lateral permeability (Figure 15d).

Consequently, depositional architecture fundamentally governs spatial reservoir quality, stimulation efficiency, and horizontal well productivity. Matching fracture and well designs to architectural flow regimes is thus essential for maximizing recovery in tight formations.

5.3. Principles for Engineering—Architecture Matching

A major outcome of this study is the recognition that each sandbody architecture corresponds to a distinct flow regime, requiring tailored horizontal well designs. Optimization of engineering parameters—such as horizontal length, fracture stage, and fracture length—must be adaptively matched to the underlying flow unit defined by the sandbody’s geometry and anisotropy. In these reservoirs, permeability is typically anisotropic, typically with K_// > K_⊥, making wellbore and fracture alignment critical for maximizing effective drainage.

For channelized architectures (Types A and B), which act as tubular flow conduits, the optimization focus is to align the horizontal wellbore with the principal flow axis and to match both the horizontal section length and fracture half-length to the effective transport scale of the high-permeability zone (Figure 15a). Exceeding this width leads to marginal returns. In contrast, Type C (mouth bar) architectures resemble blanket-like flow systems with reverse rhythmic or uniform vertical permeability (Figure 15c). Here, optimization emphasizes longer lateral sections and more fracture stages to connect deeper and distal intervals and form a three-dimensional drainage network.

To support application in similar geological settings, an empirical design chart was developed using dimensionless ratios: L_x = L_h/L_s and L_y = L_f/W_s. The chart delineates optimal design windows across architecture types (Figure 16). Recommended ranges are:

L_x = L_h/L_s (horizontal length ratio)

• Type A: 0.5–0.7;
• Type B: 0.6–0.8;
• Type C: 0.5–0.7.

L_y = L_f/W_s (fracture length ratio)

• Type A: 0.5–0.6;
• Type B: 0.45–0.55;
• Type C: 0.2–0.3.

The applicability of this chart has been validated through field development practices in the tight Chang 7 Member reservoirs of the Ordos Basin. Grounded in the sedimentary context of shallow-water delta-front systems and supported by a dataset of 240 core-based experimental measurements, its use is primarily confined to reservoirs exhibiting matrix permeability below 10 mD and also subject to several constraints beyond permeability. First, the chart is valid for sandbodies with thicknesses of at least 6 m, as thinner bodies were excluded from the statistical analysis due to horizontal well landing requirements. Second, the chart assumes the presence of sufficient stress contrast or bounding layers to contain hydraulic fractures vertically within the sandbody. In cases where fractures are likely to grow out of zone—due to low stress contrast, thin sandbodies—the recommended L_f/W_s ratios may overestimate the effective stimulated volume. Third, the chart was developed using permeability anisotropy patterns characteristic of shallow-water delta-front deposits; its application to other depositional environments (e.g., deep-water depositions) should be undertaken with caution and, ideally, with local calibration. In formations with more favorable petrophysical properties, the marginal flow effects governed by architectural constraints become less pronounced or even negligible, thereby limiting the relevance and predictive value of the chart in such settings.

6. Conclusions

This study presents an integrated architecture-constrained optimization framework for horizontal well design in the tight Chang 7 Member sandstones of the Ordos Basin. Key conclusions are as follows:

• A coupling methodology between geological architecture and engineering design was established, integrating quantitative sandbody characterization, core-calibrated petrophysical data, and high-resolution 3D modeling. This framework improves design precision by aligning horizontal well parameters with subsurface heterogeneity.
• Distinct architectural types impose different flow behaviors and optimal stimulation strategies. Channelized systems (Types A and B) favor shorter horizontal sections and controlled fracture spacing, while broader mouth bar architectures (Type C) benefit from extended well lengths and multi-stage fracturing. These differences reflect sedimentary controls on permeability architecture, drainage geometry, and fracture efficiency.
• A dimensionless design chart was developed using sandbody-scale ratios (L_h/L_s and L_f/W_s) to guide preliminary parameter selection. This tool enables scalable application of the optimization strategy across geologically similar tight reservoirs.
• Field validation confirmed the effectiveness of the architecture-constrained design approach, with wells adhering to optimized parameters (e.g., Well D12) demonstrating superior and more stable production.

These results provide a data-driven foundation for integrating sedimentary architecture into stimulation design and offer practical guidelines for improving recovery efficiency in unconventional deltaic sandstones.