Optimization of Logistics Distribution Centers Based on Economic Efficiency and Sustainability: Data Support from the Hohhot–Baotou–Ordos–Ulanqab Urban Agglomeration

Abstract

This study proposes a nonlinear 0-1 mixed-integer programming model for optimizing the location of logistics distribution centers within the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, integrating transportation costs, carbon emissions, and operational coefficients. The optimization problem is solved using a genetic algorithm (GA), whose robustness is systematically validated through comparative analyses with linear programming (LP) and alternative heuristic optimization methods including simulated annealing (SA) and particle swarm optimization (PSO). Comprehensive sensitivity analyses are conducted on critical parameters—including transportation costs, demand fluctuations, carbon pricing mechanisms, the logistics center capacity, land use impact, and water resource constraints—to evaluate the model’s adaptability under diverse operational scenarios. The research methodology incorporates environmental impact factors, including carbon emission costs, land resource utilization, and water resource management, thereby extending traditional optimization frameworks to address region-specific ecological sensitivity concerns. The empirical results demonstrate that the optimized location configuration significantly reduces logistics operational costs while simultaneously enhancing both the economic efficiency and environmental sustainability, thus fostering regional economic coordination. This study makes several key contributions: (1) developing an integrated decision-making framework that balances economic efficiency and environmental sustainability; (2) systematically incorporating environmental impact factors into the optimization model; (3) establishing calibration methods specifically tailored for ecologically sensitive regions; and (4) demonstrating the potential for the synergistic optimization of economic and environmental objectives through strategic logistics network planning.

1. Introduction

1.1. Research Background

According to the “14th Five-Year Plan for National Economic and Social Development of the Inner Mongolia Autonomous Region and the Outline of Vision Goals for 2035”, a comprehensive “14th Five-Year” Integrated Development Plan for the Hohhot–Baotou–Ordos–Ulanqab region has been formulated. This plan aims to establish an inclusive platform integrating resources and elements both within and beyond the region. The Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, widely acknowledged as the economic “locomotive” of the Inner Mongolia Autonomous Region, has demonstrated consistent growth in both the resident population and economic contribution. As the population and economic carrying capacity continue to expand, this urban cluster has emerged as the primary catalyst for economic growth, ecological initiatives, reform, and innovation within the autonomous region.

Accelerating the integration process is anticipated to facilitate high-quality development within Inner Mongolia, foster a coordinated regional framework characterized by complementary advantages, and enhance ecological protection in the Yellow River Basin. Moreover, this development strategy is expected to advance the region’s integration into both domestic and international markets, supporting the establishment of a new development paradigm.

Regional logistics constitutes a critical component of the integrated development plan and is recognized as a fundamental pillar supporting regional integration. An optimized regional logistics system is instrumental in facilitating industrial cluster development, fostering inter-city specialization, and enhancing regional competitiveness. Furthermore, such a system can significantly reduce transportation and storage costs, improve the logistics efficiency, and lower business operating expenses, thereby enhancing the economic performance and accelerating regional growth. Additionally, an environmentally sustainable logistics system is essential for this ecologically sensitive region, where water scarcity, grassland preservation, and carbon emission reduction represent critical challenges alongside economic development imperatives.

Despite continuous improvements in the regional logistics system, several critical weaknesses persist: incomplete transportation networks, inadequate infrastructure in certain areas, inefficient transportation channels, a limited adoption of modern information technologies, and the insufficient implementation of green logistics technologies. The most significant issue lies in the suboptimal location and configuration of logistics distribution centers, which fails to leverage agglomeration and radiation effects, impedes operational efficiency, and contributes to increased transportation costs through redundant routing.

The primary contributions of this study are fourfold: (1) We develop an integrated framework for logistics distribution center location decision making that balances economic efficiency with environmental sustainability, specifically tailored to the region’s unique characteristics. (2) We systematically incorporate environmental impact factors—including carbon emission costs, land resource utilization, and water resource management—into the location optimization model. (3) We establish parameter calibration methods specifically designed for ecologically sensitive regions. (4) We propose regional logistics network planning strategies based on economic–environmental synergistic optimization. By optimizing the logistics distribution center network, this study addresses issues of disharmony, imbalance, and insufficiency in regional logistics development while offering new research perspectives for the “Hohhot-Baotou-Ordos-Ulanqab integration” development strategy.

1.2. Research Positioning and Methodological Selection

This study focuses on the logistics distribution center network optimization in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, adopting a problem-oriented research approach rather than a methodological innovation orientation. This research does not claim methodological innovation in combining genetic algorithms (GAs) with mixed-integer programming models. Instead, it emphasizes this combination’s applicability and effectiveness in addressing complex location problems in specific regional contexts. The genetic algorithm (GA), as a well-established heuristic optimization method, offers advantages in handling nonlinear constraints, multi-objective optimization, and avoiding local optima, making it particularly suitable for the logistics network optimization problems in ecologically sensitive regions that this research addresses.

The core value of this study lies in constructing a decision-making framework that integrates economic efficiency with environmental sustainability, considering the unique geographical, economic, and ecological characteristics of the Hohhot–Baotou–Ordos–Ulanqab region. By extending traditional logistics location problems through the integration of environmental impact dimensions, this research responds to contemporary sustainable development requirements while providing an analytical paradigm for logistics planning in similar ecologically sensitive regions.

The nonlinear 0-1 mixed-integer programming model employed in this research captures the complex relationships in logistics networks, particularly the nonlinear characteristics of transportation costs that vary with the distance and cargo volume, as well as the interactions between different environmental factors. This modeling approach more closely approximates the complexity of actual logistics operations, enhancing the practical value of the optimization results. Simultaneously, the parameter calibration methods developed in this study, especially for environmental impact factors, provide methodological references for similar research.

Furthermore, this research contributes to bridging the gap between theoretical optimization models and practical implementation challenges in ecologically sensitive regions. By carefully calibrating environmental parameters based on region-specific data and incorporating them into a comprehensive decision-making framework, we demonstrate how sustainability considerations can be operationalized within logistics network optimization. The multi-dimensional sensitivity analysis presented in later sections provides practical insights for policy-makers and logistics planners, enabling them to understand the critical thresholds and interaction effects among economic and environmental parameters. This approach represents a significant advancement beyond traditional single-objective optimization models that fail to capture the complex interplay between economic efficiency and environmental sustainability in real-world logistics systems.

The structure of this paper is organized as follows: Section 1 presents the introduction; Section 2 provides a comprehensive literature review including regional integration theory, the relationship between regional logistics and regional integration, research on the logistics distribution center location, and a systematic review of environmental impacts in logistics center location decisions; Section 3 outlines the problem description; Section 4 details the model construction, including modified basic assumptions, the mathematical model, and parameter calibration; Section 5 describes the algorithm design and analysis, including the implementation of the GA, a comparative analysis, sensitivity analysis, and implementation recommendations; and Section 6 concludes this paper.

2. Literature Review

2.1. Regional Integration

The theoretical foundation of this study is built upon three interconnected research domains: regional integration, regional logistics, and logistics center location optimization. First, regional integration theory provides the fundamental framework for understanding economic, political, and social cooperation between geographically adjacent regions, constituting the theoretical starting point for examining the development of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration. This section reviews the historical evolution and core theoretical perspectives of regional integration research, establishing the foundation for the subsequent analysis.

Regional integration refers to the formation of closer alliances or communities by geographically adjacent countries or regions through enhanced cooperation across economic, political, and social domains. This process encompasses not only trade and investment liberalization, but also multi-level interactions including policy coordination, institutional integration, and cultural exchanges. Studies of regional integration can be traced to the Hanseatic League, considered a representative case of early economic interactions [1]. Research demonstrates that regional integration plays a pivotal role in driving economic growth. First, by fostering economies of scale, it significantly enhances the resource allocation efficiency [2].

Second, by facilitating technology spillovers, it encourages regional technology sharing and dissemination [3]. Third, by stimulating enterprise investment, it improves market competitiveness [4]. Significant similarities exist between the internal driving forces of regional integration and factors driving economic globalization [5]. Both phenomena are primarily attributed to continuous growth in economic demand that propels countries toward more extensive cooperation [6].

Current theoretical research on regional integration primarily converges around two perspectives: analyses rooted in new institutional economics and frameworks derived from new regionalism theory [7]. New institutional economics emphasizes the importance of institutions in regional integration, although institutional factors are not necessarily the sole determinants during early integration stages [8]. However, effective institutional design and policy coordination remain essential for promoting regional convergence and narrowing economic disparities [9]. New regionalism theory regards regional economic development as the core driving force for regional integration, arguing that rapid regional economic growth promotes deeper regional cooperation, with institutional integration emerging as a natural response to strengthening economic ties [10].

In summary, regional integration effectively reduces transaction costs by lowering institutional barriers, eliminating local protectionism, and simplifying inter-regional communication processes. This facilitates the free flow of production factors and ensures optimal resource allocation, not only promoting regional economic integration, but also serving as a driving force for long-term economic development within the region [11].

2.2. An Investigation into the Relationship Between Regional Logistics and Regional Integration

Building on the theoretical foundation of regional integration, this section further explores the relationship between regional logistics and regional integration. As a key element in promoting regional economic integration, the development level of regional logistics directly influences the process and quality of regional integration. By systematically reviewing relevant research, we analyze how regional logistics drives regional integration and the challenges and opportunities encountered in this process.

Studies demonstrate that regional logistics plays a crucial role in the process of regional integration. Regional logistics is defined as the system facilitating material circulation through various transportation modes and logistics networks within a specific region, addressing regional economic integration demands [12,13]. An efficient logistics system not only improves the interconnection between regions, but also enhances regional economic cohesion, serving as a key factor in promoting regional economic transformation.

Improving the logistics infrastructure constitutes a fundamental condition for efficient logistics system operation and provides critical support for promoting regional integration. A high-quality logistics infrastructure not only enhances the logistics operational efficiency, but also strengthens the inter-regional connectivity, providing essential support for regional economic integration [14]. Logistics infrastructure construction and optimization bring significant external effects to regional economic interactions, enhance social welfare, and improve the business environment, thereby strengthening regional economic cohesion. Hulten et al. noted that improving regional logistics infrastructure can significantly enhance regional overall competitiveness [15], accelerating the integration process. However, the high costs of logistics infrastructure construction may also result in social and environmental burdens, such as land-use conflicts and increased carbon emissions [16,17]. Promoting coordinated development between regional logistics and regional economics represents an effective approach to achieving regional integration. Through policy coordination and infrastructure optimization, regional logistics systems and economic development can establish positive interaction mechanisms, significantly enhancing regional economic competitiveness [18,19]. Although the role of regional logistics in promoting regional integration is widely recognized, existing research still shows deficiencies in the environmental impact assessment and social cost analysis. For instance, Ghayebloo et al. pointed out that carbon emissions from logistics activities and long-term impacts of resource consumption on regional environments have not received adequate attention [20,21]. Additionally, current research predominantly focuses on developed regions, with relatively limited studies on logistics models in developing and underdeveloped regions [22,23].

2.3. Research on Regional Logistics and the Location of Regional Logistics Distribution Centers

The preceding analysis of regional logistics and regional integration provides the macroeconomic context for our study. We now narrow our focus to examine the specific research domain of logistics distribution center location optimization, which represents the core methodological foundation of our investigation. This section analyzes the current state of research on logistics distribution center location problems, with particular attention to computational complexity considerations and algorithmic approaches.

Regional logistics represents a complex, multi-level system encompassing goods circulation and distribution across individual, collective, and societal levels [24,25]. From a commercial product flow perspective, its core function lies in redistributing industrial and agricultural resources via logistics activities, realizing the commodity value and facilitating market circulation [26]. Regional logistics operations involve complete processes from the supplier to the recipient, covering key links including transportation, storage, loading and unloading, handling, packaging, circulation processing, distribution, and information interaction [27]. Regional logistics demonstrates notable characteristics including spatial distribution, system independence, and structural integrity [28]. As a highly systematic network, regional logistics addresses diverse regional economic needs through efficient resource allocation and facilitates the coordinated development of both internal and external regional economies. This systematic structure not only provides foundational support for regional economic activities, but also strengthens the region’s overall competitiveness within broader market contexts [29].

Within regional logistics systems, the logistics distribution center location is critical for optimizing and enhancing the regional logistics network efficiency. Researchers have proposed various solutions to address logistics distribution center location challenges, including greedy heuristic algorithms [30] and two types of stochastic programming models [31], including a multi-objective network optimization model with random fuzzy coefficients [32], a chance-constraint model [33], hybrid multiple local search algorithms [34], Lagrangian relaxation heuristic algorithm [35], disk algorithm [36], a two-phase hybrid heuristic algorithm [37], the K-means clustering algorithm [38], scatter search algorithm [39], bilevel programming model [40], heuristic algorithms [41], the branch and bound method [42], fuzzy multi-criteria decision-making methods [43], a three-level joint supply chain network model [44], and so on. These methods have yielded significant results in optimizing logistics distribution center locations.

From a computational complexity perspective, logistics distribution center location problems are classified as NP-hard, presenting significant challenges as problem dimensions increase. Daskin and Owen [45] demonstrated that even basic facility location problems are NP-complete, with solution spaces growing exponentially with problem size. This complexity is further amplified when incorporating multiple constraints, nonlinear cost functions, and environmental factors. Hakimi’s [46] seminal work on network location theory established that for problems involving more than p facilities, exhaustive enumeration becomes computationally prohibitive. Particularly relevant to our research context, Feliciani et al. [47] confirmed in their study, “Exact and heuristics algorithms for screen line problem in large size networks”, that exact algorithms face significant scalability limitations when applied to complex network problems like those in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration. Similarly, Almutairi et al. [48], in “Active Traffic Sensor Location Problem for the Uniqueness of Path Flow Identification in Large-Scale Networks”, demonstrated that meta-heuristic approaches offer a superior performance in balancing solution quality and computational efficiency for large-scale spatial optimization problems. The computational complexity of the logistics distribution center location is particularly relevant when considering the environmental sustainability dimensions. Centobelli et al. [49] demonstrated that the incorporation of nonlinear environmental cost functions transforms even simplified location models into complex non-convex optimization problems. This theoretical understanding of computational complexity provides an important context for algorithm selection in logistics network optimization, particularly for ecologically sensitive regions requiring multi-dimensional sustainability constraints.

However, two significant shortcomings persist in existing research: first, most algorithms lack adequate adaptability to complex scenarios, exhibit limited capabilities in exploring optimal solutions across broader ranges, and are prone to convergence at local optima; second, limited research has integrated carbon emission costs with conventional distribution center costs for comprehensive analysis. Therefore, this study proposes a GA to address complex multi-objective optimization problems while integrating transportation and carbon emission costs from suppliers to distribution centers and from distribution centers to customers, along with variable costs associated with processing, storage, and packaging within distribution centers and fixed construction costs, aiming to optimize total costs using a GA to determine optimal location schemes.

The findings of this study are expected to offer theoretical support for the network planning of the logistics system within the ‘Hohhot-Baotou-Ordos-Ulanqab regional integration’. Through optimizing the regional distribution center system, this study aims to address issues of disharmony, imbalance, and deficiency in regional logistics development, thereby offering a novel research perspective to advance the ‘Hohhot-Baotou-Ordos-Ulanqab regional integration’ development strategy.

In existing research on the logistics center location, while economic factors and network optimization have received extensive attention, the assessment of environmental impacts remains relatively insufficient. As highlighted by Ghayebloo et al. [20] and Li et al. [21], carbon emissions from logistics activities and their long-term impacts on regional environments have not been comprehensively considered. With the concept of sustainable development increasingly permeating the field of logistics planning, exploring the environmental impact dimensions of the logistics center location has become particularly significant. Therefore, the following section will systematically review research advancements regarding the environmental impacts of the logistics center location, establishing a theoretical foundation for constructing optimization models that integrate economic efficiency with environmental sustainability.

2.4. Comprehensive Review of Environmental Impact of Research on Logistics Center Location

After establishing the computational complexity dimensions of logistics center location problems, we now examine another critical aspect: environmental impact considerations in location decision making. While traditional optimization approaches have predominantly focused on economic efficiency metrics, the increasing prominence of sustainability imperatives necessitates a comprehensive analysis of environmental factors in location decisions. This section provides a systematic review of environmental impact research related to the logistics center location, establishing the theoretical foundation for our integrated optimization approach.

As sustainable development principles increasingly permeate logistics planning paradigms, the environmental dimension of the logistics center location has emerged as a focal point of scholarly inquiry. This section systematically reviews research advancements regarding the environmental impacts of logistics center location to provide a theoretical foundation for optimization model construction. Figure 1 presents a conceptual framework of environmental factors in logistics center location decisions, illustrating the inter-relationships among environmental elements and their interactions with economic efficiency and regional integration. The following sections will systematically analyze theoretical developments across various environmental impact dimensions based on this framework.

2.4.1. Direct Environmental Impacts of Logistics Distribution Centers

Land-Use and Ecosystem Impacts: Logistics center construction significantly influences land-use patterns. Zhang et al., through spatial analysis methods, quantified the impact of logistics distribution centers on land transformation, revealing significant differences in ecological footprints between urban areas and suburban greenfield development [50]. Their empirical analysis demonstrates that land-use changes induced by logistics center development in peri-urban areas can diminish regional biodiversity by 15–25% within affected ecosystems. Li et al. demonstrated that the impact of the logistics center location on habitat integrity should be evaluated through landscape connectivity indicators, which correlate significantly with the traffic flow and built-up area expansion [51].

Waste Management and Circular Economy Pathways: Waste management efficiency in logistics distribution centers constitutes a critical indicator for environmental impact assessment. Ajayi et al. applied material flow analysis to identify the primary components of logistics center waste streams and their environmental contributions [52]. Research indicates that large logistics distribution centers generate 2500–4000 tons of solid waste annually, with approximately 65% recoverable through optimized management practices. Gholizadeh et al. established a life-cycle waste assessment model that incorporated construction and operational phase waste into a unified analytical framework, addressing deficiencies in traditional assessment methodologies [53]. A comparative analysis by Tornese and Pazour indicated that regional disparities in the waste management infrastructure capacity significantly affect the environmental performance of logistics distribution centers and should serve as a key constraint in location decisions [54]. These studies not only provide robust methodological frameworks for integrating waste management factors into optimization models, but also establish quantitative benchmarks for sustainable logistics planning.

2.4.2. Indirect Emissions and Life-Cycle Assessment

Life-cycle assessment methodologies are increasingly applied in logistics center environmental impact research. Mohamed Abdul Ghani et al. constructed a carbon footprint model for logistics distribution centers that quantified full life-cycle emissions from facility construction, operation, and maintenance, providing a scientific basis for emission reduction strategies at different stages [55]. Studies indicate that operational phase emissions typically account for 65–80% of total emissions, though high-intensity emissions during the construction phase remain significant. Zhang et al. quantified indirect environmental burdens of energy consumption, water resource utilization, and auxiliary infrastructure through an input–output life-cycle assessment [56]. Comparative research by Aronsson and Huge Brodin confirmed that comprehensive assessments considering indirect emissions are crucial for logistics network optimization in both developed and underdeveloped regions [57]. These life-cycle research methodologies directly influenced the carbon emission factor construction in our model, particularly the comprehensive consideration of emissions across the supplier–logistics center–customer chain.

2.4.3. Environmental Benefits and Cost Trade-Off Analysis

Logistics center location decisions involve complex trade-offs between environmental benefits and economic costs. Tadaros et al. developed a multi-criteria decision model that integrated environmental indicators such as the carbon emission intensity, energy efficiency, and land use efficiency with traditional economic indicators [58]. Empirical research by Salçuk and Şahin demonstrated that optimized configurations can achieve synergistic improvements in environmental benefits and economic efficiency, challenging traditional environment–economy trade-off hypotheses [59]. Progress has also been made in monetary valuation methods for environmental benefits, with Yin proposing a sensitivity analysis framework based on carbon pricing that provides scientific evidence for policy formulation [60]. These studies provide theoretical support for balancing economic efficiency and environmental sustainability objectives in our model.

2.4.4. Environmental Considerations Specific to the Hohhot–Baotou–Ordos–Ulanqab Urban Agglomeration

The Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, situated at the junction of the Inner Mongolian Plateau and the Loess Plateau, possesses unique ecological characteristics. This region experiences an arid climate with annual precipitation of only 300–400 mm, high ecosystem vulnerability, and significant desertification risk. According to data from the Inner Mongolia Ecological Environment Department, this region’s ecological sensitivity exceeds the national average by approximately 28%, making logistics infrastructure development particularly impactful on local grassland ecosystems.

Concurrently, as a major energy export base, carbon emissions from coal mining and transportation account for 43.7% of the regional total emissions. Logistics network configuration significantly influences regional carbon reduction targets. Based on data from the Inner Mongolia Autonomous Region Development and Reform Commission, optimized logistics network layouts can reduce regional transportation carbon emissions by 15–20%, which constitutes a primary rationale for emphasizing carbon emission costs in our model.

Regional water scarcity represents another critical environmental factor in logistics center location decisions. Large logistics distribution centers consume 200–300 cubic meters of water daily, while strategic location selection can reduce water consumption by 30%. These region-specific environmental factors directly influenced the preliminary screening process for logistics center candidate sites and the environmental cost parameter configuration in our study.

Our comprehensive systematic review of environmental impact research on the logistics center location reveals several critical lacunae in the extant literature, underscoring the significance of our contribution: (1) environmental factors are predominantly employed as post hoc assessment indicators rather than integrated decision variables within optimization frameworks; (2) region-specific ecological sensitivity considerations—particularly for ecologically vulnerable areas—remain severely underexplored; (3) comprehensive life-cycle perspectives are largely absent, with most studies focusing exclusively on operational-phase impacts while neglecting construction and maintenance phases; and (4) water resource management, despite being a critical environmental constraint in many regions, has been virtually overlooked in logistics network optimization research.

This study addresses these theoretical and practical gaps by developing a methodological framework that synergistically optimizes environmental factors alongside economic objectives, calibrating parameters specifically for ecologically sensitive regions, and establishing an integrated assessment framework. Our empirical investigation of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration demonstrates how this approach can enhance ecological resilience while maintaining economic efficiency, providing decision support for sustainable logistics planning in comparable regions.

The literature review in this section establishes the theoretical foundation for subsequently constructing logistics center location optimization models that integrate environmental impact factors. In our model construction, we will focus specifically on the following environmental factors: (1) carbon emission costs, integrating the carbon footprint across the supplier–logistics center–customer chain; (2) land resource utilization, reducing land consumption through location optimization; and (3) water resource management, considering the impact of the logistics center location on the regional water-carrying capacity. These environmental factors are specifically quantified as carbon emission costs, environmental sensitivity constraints, and water resource limitations in Section 4’s model construction, and are optimized through multi-objective trade-offs in the GA implementation in Section 5, achieving synergistic optimization of economic efficiency and environmental sustainability.

3. Problem Description

3.1. Actual Conditions

The Hohhot–Baotou–Ordos–Ulanqab urban agglomeration has achieved substantial progress in integrated development, emerging as a pivotal regional node. However, logistics development has predominantly focused on infrastructure-centric initiatives rather than system optimization. Recent data indicate that logistics expenditure accounts for approximately 8.6% of the region’s GDP—significantly higher than the national average of 6.2%—suggesting considerable systemic inefficiencies (Inner Mongolia Statistical Bureau, 2023).

Current logistics practices reflect a fragmented approach characterized by independent municipal decision making with minimal inter-jurisdictional coordination. This results in resource duplication, the suboptimal utilization of regional advantages, reduced network efficiency, and elevated costs. Environmental impacts are particularly concerning, with regional logistics carbon emissions having increased by 14.3% over the past five years (Inner Mongolia Ecological Environment Department, 2023).

The region’s unique ecological characteristics—situated at the junction of the Inner Mongolian Plateau and Loess Plateau, with annual precipitation of merely 300–400 mm and high ecosystem vulnerability—require logistics planning that balance the operational efficiency with environmental preservation. This environmental sensitivity significantly exceeds the national average by approximately 28%, making logistics infrastructure development particularly impactful on local grassland ecosystems.

The logistics distribution center location problem addressed in this paper involves selecting optimal distribution centers from multiple candidate locations to serve various demand points with different supply sources, while minimizing the total systemic cost and adhering to environmental constraints. Figure 2 illustrates the conceptual structure of the regional logistics distribution center network.

Specifically, key challenges include

• Determining the alternative locations based on geographic positioning, transportation accessibility, market demand, and environmental sensitivity;
• Selecting optimal locations that balance economic efficiency metrics with environmental sustainability considerations.

Following a comprehensive evaluation of geographical, economic, and infrastructural factors, nine potential locations were identified (Figure 3): the four major cities (Hohhot, Baotou, Ordos, and Ulanqab) representing economic centers with a substantial logistics demand, plus five strategically positioned counties (Zhungeer, Helingeer, Dalate, Tumotezuo, and Zhuozi) selected for their geographic advantages, transportation connectivity, and cost efficiency.

3.2. Specific Problems and Bottlenecks in the Current Logistics System

The current logistics system of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration exhibits several critical inefficiencies that necessitate strategic optimization. A comprehensive analysis of key performance indicators reveals significant deviations from industry benchmarks, highlighting the urgent need for systematic improvement (Figure 4).

Figure 4 illustrates the substantial performance gaps across multiple dimensions. Notably, the logistics cost to the GDP ratio (8.6%) exceeds the national average by 38.7%, while the road transport dominance (84.5%) indicates a severe modal imbalance compared to more developed regions. These inefficiencies are structured around three interconnected challenge domains.

Transportation Network Inefficiencies and Infrastructure Imbalance: The regional logistics network is characterized by significant redundancy, with an average freight transportation distance of 264.5 km—42.8% longer than optimized routing would require. This inefficiency is compounded by an alarmingly high empty-running rate of 38.2%, substantially exceeding the national average of 31.5% and generating unnecessary economic costs of approximately 3.87 billion CNY annually. Furthermore, the modal distribution is markedly imbalanced, with road transport accounting for 84.5% of the total freight volume, while multimodal transport comprises merely 12.3%—significantly below the 36.7% achieved in comparable urban agglomerations. The logistics infrastructure landscape exhibits pronounced disparity, characterized by functional homogeneity (76.2% functional overlap) and uneven service coverage (93.5% in urban centers versus 71.2% in surrounding counties).

Governance Fragmentation and Information Asymmetry: The absence of integrated coordination mechanisms has resulted in administrative fragmentation, with a logistics policy consistency index of only 0.63 (on a scale of 1.0). This governance inefficiency is exacerbated by limited information sharing (23.7% compared to 62.5% in the Pearl River Delta region) and pronounced standardization deficiencies (unitization rate of 43.2%, 11.8 percentage points below the national average). The consequent information asymmetry between the logistics supply and demand extends the cargo-vehicle matching time to 2.7 h, significantly higher than the 0.8 h observed in more developed logistics regions.

Environmental Impact and Resource Utilization Challenges: The environmental footprint of current logistics operations is particularly concerning given the region’s ecological sensitivity. The carbon emission intensity stands at 0.38 kg CO₂/ton-kilometer, 31% above the national average, while the water consumption efficiency remains suboptimal at 0.43 cubic meters per ton—72% higher than in advanced logistics regions. The land utilization efficiency is similarly compromised, with an average of 78.5 square meters per 10,000 tons of throughput, exceeding optimized levels by 23.8%. The region’s energy consumption intensity of 0.087 tons of standard coal equivalent per 10,000 CNY is 15.3% higher than the national advanced level, further intensifying environmental pressure in this ecologically vulnerable region.

These quantitative indicators underscore the systemic inefficiencies within the current logistics network, highlighting the critical importance of optimizing the regional logistics distribution center layout. The subsequent modeling approach systematically addresses these interdependent challenges by balancing economic efficiency objectives with environmental sustainability imperatives.

4. Model Building

Having identified the candidate locations for logistics distribution centers, we now address the second key issue: how to select the optimal combination of locations from these alternatives to serve the Hohhot–Baotou–Ordos–Ulanqab region. To solve this optimization problem, we develop a mathematical model that integrates economic efficiency and environmental sustainability considerations.

4.1. Model Assumptions and Formulation

4.1.1. Modified Basic Assumptions

Transportation Dynamics: We have supplanted the traditional linear transportation cost assumption with a more sophisticated nonlinear function that incorporates distance decay effects, traffic congestion factors, and temporal variations:

$$T{C}_{ij}=c·{d_{ij}}^{\alpha }·{q_{ij}}^{\beta }·f\left(t\right)$$

where $d_{ij}$ denotes distance, $q_{ij}$ represents volume, $t$ represents time period, and $\alpha , \beta , c$ are empirically calibrated parameters. The function $f(t)$ captures cyclical variations in the transportation efficiency based on diurnal or seasonal factors. This formulation more accurately reflects the sub-linear relationship between distance and transportation costs observed in regional logistics operations, wherein per-kilometer costs typically diminish over extended distances.

Capacity Constraints: Conventional models typically posit static capacity assumptions, whereas operational logistics distribution centers exhibit dynamic capacity utilization patterns. We have implemented a dynamic capacity adjustment model: $C=C_o·{\theta }_s·{\eta }_j$, $C_o$ is the base capacity, ${\theta }_s$ is the seasonal adjustment factor, and ${\eta }_j$ is the utilization efficiency. This formulation accommodates seasonal demand fluctuations characteristic of the Hohhot–Baotou–Ordos–Ulanqab region, where winter conditions significantly affect logistics operations.

Rather than employing fixed processing coefficients, we have modeled these as functions of facility conditions:

$$PC=P{C}_0·{(1+\delta ·u_j-0.8)}^2)$$

where $u_j$ denotes the utilization rate and $P{C}_0,\delta$ are system parameters. This captures the operational reality that the processing efficiency varies with facility utilization, typically declining at both extremes of utilization rates.

Multi-modal Transportation: We have incorporated the option for different transportation modalities based on distance thresholds, volume requirements, and time sensitivity. Specifically:

• Highway transportation is prioritized for distances under 300 km;
• Railway transportation becomes economically viable for distances ≥ 300 km and volumes > 50 tons per batch;
• Air transportation is considered for time-sensitive, high-value cargo (>50,000 CNY/ton).

Environmental Constraints: given the ecological sensitivity of the Hohhot–Baotou–Ordos–Ulanqab region, we have explicitly incorporated environmental impact factors into the model, including:

• Carbon emissions from transportation and operations;
• Land use impacts based on ecological sensitivity indices;
• Water resource consumption limitations reflecting regional water scarcity.

These modified assumptions significantly enhance the model’s capacity to capture the complex operational, economic, and environmental dynamics of the regional logistics system, thereby addressing the oversimplification concerns identified in previous research.

4.1.2. Cost Functions

In our enhanced model, we have reformulated several cost functions to more accurately represent the complex dynamics of regional logistics operations.

Transportation Cost Function

Traditional logistics cost models typically adopt a linear expression, $TC=u·d·q$, where $u$ represents the transportation cost per unit distance per unit weight, $d$ denotes distance, and $q$ indicates the cargo volume. While this linear model offers simplicity, it fails to capture several key phenomena observed in actual logistics operations:

• Economies of scale in long-distance transportation, where the unit distance cost decreases as distance increases;
• Marginal cost variations for different cargo volumes, with larger shipments typically enjoying lower unit costs;
• Effects of traffic congestion and temporal factors, where peak-period transportation costs significantly exceed off-peak costs.

To address these limitations, we introduce a more realistic nonlinear transportation cost function:

$$T{C}_{ij}=c·{d_{ij}}^{\alpha }·{q_{ij}}^{\beta }·f\left(t\right)$$

where:

• $c$ represents the basic freight rate coefficient;
• ${d_{ij}}^{\alpha }$ captures the distance effect, with $\alpha$ < 1 reflecting economies of scale in long-distance transportation;
• ${q_{ij}}^{\beta }$ expresses the volume effect, with $\beta$ < 1 reflecting cost benefits of batch transportation;
• $f\left(t\right)$ is a time-dependent function capturing traffic congestion and peak-period effects.

To determine precise values for these parameters, we applied logarithmic linear regression to analyze 1432 transportation records from the Hohhot–Baotou–Ordos–Ulanqab region:

$$ln\left(TC\right)=ln\left(c\right)+\alpha ·ln\left(d\right)+\beta ·ln\left(q\right)+ln\left(f\left(t\right)\right)$$

Regression analysis revealed $\alpha$ = 0.85 (confidence interval [0.82, 0.87], p < 0.001), indicating that a 100% increase in distance results in only an approximately 85% increase in cost, confirming significant economies of scale in long-distance transportation. Parameter $\beta$ = 0.93 (confidence interval [0.91, 0.95], p < 0.001) indicates slight but statistically significant cost savings with the increased cargo batch size. The model’s overall fit ($R^2$= 0.91) validates the nonlinear model’s strong explanatory power for actual transportation cost variations.

The time function $f\left(t\right)$ was determined as $f\left(t\right)=1+0.25·sin{\left(\pi ·\left(t-6\right)/12\right)}^2$, precisely capturing traffic congestion effects during morning (7–9 AM) and evening (5–7 PM) peak periods, when costs increase by up to 25%. This time function is based on an analysis of 24 h traffic flow monitoring data from the Hohhot and Baotou urban areas, reflecting the actual impact of urban traffic flow fluctuations on logistics operational costs.

Carbon Emission Function

Quantifying carbon emissions from logistics operations requires a comprehensive approach that accounts for multiple contributing factors. Traditional emissions models often employ simplistic linear relationships that fail to capture the complexities of real-world logistics operations. To address these limitations, we developed a multi-factor carbon emission function:

$$C{E}_{ij}=(e_0+e_1·q_{ij}+e_2·t_{ij})·d_{ij}$$

where:

• $e_0$ represents the basic emission factor (kg CO₂/ton-km);
• $e_1$ denotes the load-specific emission adjustment (kg CO₂/ton²-km);
• $e_2$ captures the time-dependent emission factor (kg CO₂/hour-km);
• $q_{ij}$ is the cargo volume from i to j (tons);
• $t_{ij}$ represents the service time;
• $d_{ij}$ indicates the transportation distance (km).

This formulation integrates three critical emission components:

• Base emissions ($e_0·d_{ij}$): capture the fundamental relationship between distance and emissions, reflecting fuel consumption under standard operating conditions;
• Load-dependent emissions ($e_1·q_{ij}·d_{ij}$): Account for the additional fuel consumption and resultant emissions caused by an increased vehicle load. This component reflects the physics of transportation, where a greater payload requires more energy to overcome inertia and resistance forces;
• Time-dependent emissions ($e_2·t_{ij}·d_{ij}$): incorporates emissions associated with service time, including idling during loading/unloading operations and speed variations in congested conditions.

The multiplicative relationship with distance ($d_{ij}$) reflects the cumulative nature of emissions across the transportation network. Through rigorous calibration using regional emission data from the Inner Mongolia Development and Reform Commission, we determined precise parameter values:

• $e_0$ = 0.062 kg CO₂/ton-km;
• $e_1$ = 0.0015 kg CO₂/ton²-km;
• $e_2$ = 0.023 kg CO₂/hour-km.

These parameters were validated through a comparative analysis between model predictions and actual measured emissions data from regional logistics operations, with validation results showing an average deviation of less than 8%, confirming the model’s accuracy for environmental impact assessment applications.

Operating Cost Function

To accurately capture the operational cost dynamics of logistics distribution centers, we developed a comprehensive cost function integrating three key components:

$$O{C}_j=a·{u_j}^2+b·m_j+g·e_j$$

where:

• $u$ is the utilization rate;
• $m_j$ represents the maintenance level;
• $e_j$ signifies energy consumption;
• $a,b,g$ are calibrated coefficients.

This formulation is built upon both theoretical principles and empirical observations from regional logistics operations. Each component represents a distinct cost driver.

The utilization cost term ($a·{u_j}^2$): This quadratic term reflects the nonlinear relationship between facility utilization and operational costs. We adopted a quadratic rather than linear form based on observations that costs accelerate as utilization approaches capacity limits. This nonlinearity stems from:

• Increased congestion and waiting times at high utilization levels;
• Accelerated equipment wear requiring more frequent maintenance;
• Higher labor costs including overtime payments during peak periods;
• Decreased operational efficiency due to space constraints.

The maintenance cost term ($b·m_j$): This linear component captures the direct relationship between maintenance expenditure and system performance. It encompasses routine maintenance, equipment repairs, facility upkeep, and technical personnel costs. The maintenance level $m_j$ represents a standardized index of maintenance intensity.

The energy cost term ($g·e_j$): This component accounts for energy-related expenditures proportional to energy consumption $e_j$, including electricity for lighting, climate control, material handling equipment, and information systems. Energy costs represent a significant portion of operational expenses in the Hohhot–Baotou–Ordos–Ulanqab region, particularly given seasonal temperature extremes requiring substantial heating and cooling. The empirical validation of this three-component model using operational data from five major logistics distribution centers in the region demonstrated superior explanatory power ($R^2$ = 0.89) compared to alternative linear formulations ($R^2$ = 0.72). The specific calibration process and coefficient values are detailed in Section 4.3.

4.1.3. Environmental Impact Integration

To address the ecological sensitivity of the Hohhot–Baotou–Ordos–Ulanqab region, we have incorporated two key environmental dimensions into our optimization framework:

Land-Use Impact:

$$L{I}_j=l_j·A_j·S_j$$

This formulation integrates three critical variables:

• $l_j$ represents the land impact coefficient, varying by land type: grassland (1.4), farmland (1.2), urban land (0.9);
• $A_j$ denotes the area occupied by the logistics center (m²);
• $S_j$ indicates the location-specific ecological sensitivity index.

The differentiated design of land impact coefficients ($l_j$) is based on the ecological vulnerability assessment, with grassland ecosystems assigned the highest coefficient (1.4), reflecting their unique ecological value in the Inner Mongolia Autonomous Region and greater restoration difficulty. The lower coefficient for urban land (0.9) reflects its already significantly altered ecological function and relatively smaller marginal ecological impact.

The ecological sensitivity index ($S_j$) is constructed by integrating multiple indicators, including:

• Biodiversity significance (based on species richness and rarity);
• Soil erosion risk (based on slope and soil characteristics);
• Desertification sensitivity (based on vegetation coverage and wind erosion risk);
• Ecosystem service value (based on the ecosystem service function assessment).

These indicators are quantified through GIS spatial analysis and regional ecological environmental assessment reports, ensuring the adequate protection of ecologically sensitive areas in the optimization process.

Water Resource Considerations:

$$W{C}_j=w_j·q_j$$

where:

• $w_j$ represents the water consumption coefficient for logistics center j (m³/ton);
• $q_j$ denotes the total cargo flow through logistics center j (tons).

This linear relationship reflects the proportional connection between the logistics throughput and water consumption, encompassing:

• Water used in goods processing and packaging operations;
• Facility maintenance and cleaning requirements;
• Staff usage and auxiliary services;
• Climate control systems in warehouse environments.

The water consumption coefficient ($w_j$) varies from 0.25–0.35 m³/ton depending on the facility scale, technological efficiency, and operational characteristics. This variation was determined through the detailed analysis of water consumption patterns across existing logistics distribution centers in the region, with larger facilities typically exhibiting higher water-use efficiency through economies of scale and advanced water recycling technologies.

The regional water resource limitation ($WL$) establishes a critical environmental constraint within our optimization framework. Based on comprehensive hydrological assessments from the Inner Mongolia Water Resources Department, this upper bound is set at 3.8 million m³/year, representing the sustainable withdrawal limit that prevents long-term aquifer depletion and maintains ecological water requirements.

By incorporating this water resource constraint, our model ensures that logistics network optimization respects fundamental environmental boundaries, particularly crucial in this water-stressed region where sustainable water management represents a primary environmental priority alongside carbon emissions reduction.

4.2. Mathematical Model Modifications

4.2.1. Updated Model Symbol Table

Table 1. Indices and decision variables.

Symbol	Description	Range	Type
Indices
$i$	Index of suppliers	$i\in \left\{1,2,\dots ,S\right\}$	Index
$j$	Index of logistics distribution centers	$j\in \left\{1,2,\dots ,J\right\}$	Index
$k$	Index of customers	$k\in \left\{1,2,\dots ,K\right\}$	Index
Variables
$X_{ij}$	Transportation volume from supplier i to logistics center j	$X_{ij}\in R^{+},\forall i\in S,\forall j\in J$	Variable
$Y_{jk}$	Transportation volume from logistics center j to customer k	$Y_{jk}\in R^{+},\forall j\in J,\forall k\in K$	Variable
$Z_j$	Binary variable (1 if logistics center j is selected, 0 otherwise)	$Z_j\in \left\{0,1\right\},\forall j\in J$	Variable

and

Table 2. Model parameters.

Symbol	Description	Range
$S$	Number of suppliers	$S\in Z^{+}$
$J$	Number of logistics distribution centers	$J\in Z^{+}$
$K$	Number of users	$K\in Z^{+}$
$S_i$	Total supply capacity of supplier i	$S_i\in R^{+},\forall i\in S$
$D_k$	Demand of customer k	$D_k\in R^{+},\forall k\in K$
$C_0$	Base capacity of logistics center j	$C_0\in R^{+}$
$P$	Maximum number of logistics distribution centers to be established	$P\in Z^{+},P\le J$
$f_j$	Fixed cost of logistics center j	$f_j\in R^{+},\forall j\in J$
$d_{ij}$	Distance from supplier i to logistics center j	$d_{ij}\in R^{+},\forall i\in S,\forall j\in J$
$d_{jk}$	Distance from logistics center j to customer k	$d_{jk}\in R^{+},\forall j\in J,\forall k\in K$
$c$	Basic freight rate coefficient	$c\in R^{+}$
$\alpha$	Distance effect parameter	$0<\alpha \le 1$
$\beta$	Volume effect parameter	$0<\beta \le 1$
$t$	Time period	$t\in \left[0,24\right]$
${\theta }_s$	Seasonal adjustment factor	$0.8\le {\theta }_s\le 1.2$
${\eta }_j$	Facility utilization efficiency of center j	$0.7\le {\eta }_j\le 0.95,\forall j\in J$
$P{C}_0$	Base processing cost coefficient of center j	$P{C}_0\in R^{+}$
$u_j$	Utilization rate of logistics center j	$0\le u_{j }\le 1,\forall j\in J$
$m_j$	Maintenance level of logistics center j	$1\le m_j\le 10,\forall j\in J$
$e_j$	Energy consumption of logistics center j	$e_j\in R^{+}$
$a,b,g$	Operating cost function parameters	$a,b,g\in R^{+}$
$e_0, e_1, e_2$	Carbon emission function parameters	$e_0, e_1, e_2\in R^{+}$
$p_c$	Unit price of carbon emissions	$p_c\in R^{+}$
$l_j$	Land impact coefficient of location j	$l_j\in R^{+},\forall j\in J$
$A_j$	Area of logistics center j	$A_j\in R^{+},\forall j\in J$
$S_j$	Ecological sensitivity index of location j	$0.8\le S_j\le 1.5,\forall j\in J$
$w_j$	Water consumption coefficient of center j	$w_j\in R^{+},\forall j\in J$
$WL$	Regional water resource limit	$WL\in R^{+}$
$LL$	Regional land impact limit	$LL\in R^{+}$

$Z^{+}$ represents positive integers, and $R^{+}$ represents positive real numbers.

present the mathematical notation used in our model. Table 1 defines the indices and decision variables that form the core structure of the optimization problem. Table 2 details the model parameters, which are predetermined input values that define the problem’s economic and environmental characteristics.

4.2.2. Objective Function

The integrated objective function minimizes the total systemic cost while balancing economic efficiency with environmental sustainability:

$$Min Z=TC+FC+CC+OC+LI+WC$$

(1)

where:

$$TC={\displaystyle \sum {\displaystyle \sum [\left(c·{d_{ij}}^{\alpha }·{q_{ij}}^{\beta }·f\left(t\right)\right)·x_{ij}]+{\displaystyle \sum {\displaystyle \sum [\left(c·{d_{jk}}^{\alpha }·{q_{jk}}^{\beta }·f\left(t\right)\right)·y_{jk}]}}}}$$

(2)

This transportation cost function incorporates the nonlinear relationships between distance, volume, and temporal factors. The sub-linear distance exponent ($\alpha$ = 0.85) reflects economies of scale in transportation over longer distances, which is particularly relevant in the geographically dispersed Hohhot–Baotou–Ordos–Ulanqab region. The volume exponent ($\beta$ = 0.93) captures slight economies of scale in batch transportation, while the temporal function $f\left(t\right)$ accounts for congestion effects during peak periods.

$$OC={\displaystyle \sum [a·{u_j}^{\sigma }+b·m_j+g·e_j]·Z_j}$$

(3)

The operating cost function incorporates three key operational dimensions: utilization-related costs (governed by the utilization rate $u_j$), maintenance costs (determined by the maintenance level $m_j$), and energy consumption costs (based on energy usage $e_j$). This formulation captures the operational reality that costs increase nonlinearly with utilization due to congestion effects, overtime labor requirements, and equipment depreciation.

$$FC={\displaystyle \sum f_j·Z_j}$$

(4)

The fixed cost component encompasses land acquisition, facility construction, and essential equipment investments. In the context of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, these costs vary significantly between urban centers and rural counties, with land prices in Hohhot City approximately 2.3 times higher than in peripheral counties.

$$CE=p_c·\left({\displaystyle \sum {\displaystyle \sum [(e_0+e_1·q_{ij}+e_2·t_{ij})·d_{ij}·x_{ij}]}}+{\displaystyle \sum {\displaystyle \sum [(e_0+e_1·q_{jk}+e_2·t_{jk})·d_{jk}·y_{jk}]}}\right)$$

(5)

The carbon emission cost function integrates three emission components: basic distance-related emissions ($e_0$), load-dependent emissions ($e_1$), and time-dependent emissions ($e_2$). This comprehensive formulation captures both the direct emissions from fuel consumption and indirect emissions associated with congestion and idling, providing a more accurate assessment of the environmental impact of transportation operations.

$$LI={\displaystyle \sum l_j·A_j·S_j·Z_j}$$

(6)

The land impact component quantifies the environmental impact of land utilization based on the land impact coefficient ($l_j$), the area of the logistics center ($A_j$), and the ecological sensitivity of the location ($S_j$). This term is particularly important in the Hohhot–Baotou–Ordos–Ulanqab region, where grassland ecosystem preservation represents a critical environmental priority.

$$WC={\displaystyle \sum [w_j·\left({\displaystyle \sum X_{ij}}\right)·Z_j]}$$

(7)

The water consumption component accounts for water resource utilization based on the water consumption coefficient ($w_j$) and the total cargo flow through each logistics center. This constraint is especially relevant given the significant water scarcity challenges in this arid region.

4.2.3. Constraints

Supply Constraints:

$${\displaystyle \sum X_{ij}\le S_i},\forall i\in S$$

(8)

This constraint ensures that the total supply from each supplier to all logistics distribution centers does not exceed the supplier’s maximum supply capacity. This is particularly important for suppliers in Ordos City, where the coal production capacity exhibits significant seasonal fluctuations.

Demand Constraints:

$${\displaystyle \sum Y_{jk}\ge D_k, \forall k\in K}$$

(9)

This constraint guarantees that customer demand is fully satisfied. The inequality permits potential oversupply, which may be necessary during peak seasonal periods or to accommodate emergency reserves, particularly for essential commodities in remote areas of the urban agglomeration.

Flow Balance:

$${{\displaystyle \sum X}}_{ij}={\displaystyle \sum Y_{jk}}, \forall j\in J$$

(10)

This constraint maintains the conservation of flow at each logistics center, ensuring that the total inflow equals the total outflow. This prevents inventory accumulation or depletion over time, which is essential for sustainable operational efficiency.

Dynamic Capacity:

$${{\displaystyle \sum X_{ij}\le C_0·{\theta }_s·{\eta }_j·Z}}_j,\forall j\in J$$

(11)

The dynamic capacity constraint incorporates seasonal adjustments (${\theta }_s$) and utilization efficiency factors (${\eta }_j$), allowing for more realistic capacity modeling. This is especially relevant in the Hohhot–Baotou–Ordos–Ulanqab region, where winter conditions can significantly affect logistics operations.

Environmental Constraints:

$${\displaystyle \sum w_j·\left({\displaystyle \sum x_{ij}}\right)·Z_j}\le WL$$

(12)

$${\displaystyle \sum l_j·A_j·S_j·Z_j}\le LL$$

(13)

These constraints ensure that the logistics network’s environmental impact remains within acceptable limits for water consumption ($WL$) and land use ($LL$), reflecting the region’s ecological sensitivity and resource limitations.

Installation Constraints:

$${\displaystyle \sum Z_j}\le P$$

(14)

This constraint limits the total number of logistics distribution centers to be established, balancing service coverage with investment efficiency. For the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, a maximum of five centers was determined to be optimal based on a preliminary analysis.

Non-negativity Constraints:

$$X_{ij},Y_{jk}\ge 0,Z_i\in \left\{0,1\right\}$$

(15)

These fundamental constraints ensure the logical consistency of the model, with non-negative flow variables and binary location decisions.

The integrated mathematical formulation balances economic efficiency objectives with environmental sustainability imperatives, providing a comprehensive framework for sustainable logistics network optimization in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

These constraints work in concert to create an effective decision framework balancing economic viability with environmental sustainability. Supply, demand, and flow balance constraints ensure operational functionality; dynamic capacity constraints incorporate seasonal variations; environmental constraints address regional ecological sensitivity; while installation constraints optimize resource allocation. The interactions among these constraints—such as when environmental requirements influence center distribution—highlight the value of GAs in navigating this complex solution space while avoiding local optima.

4.3. Parameter Calibration and Extension

4.3.1. Nonlinear Function Parameter Calibration

To ensure the robust applicability of the model in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, we calibrated the nonlinear function parameters through the following systematic methodology:

• Distance Effect Parameters ($\alpha$): Through a comprehensive analysis of 1432 transportation records within the region collected from 2020 to 2022, we employed logarithmic linear regression to estimate the relationship between transportation cost and distance. The calibration yielded $\alpha$ = 0.85 ($R^2$= 0.91), indicating a sub-linear relationship between distance and cost—a finding consistent with economies of scale in transportation over longer distances. This dataset encompassed diverse operational scenarios across seasonal variations, vehicle typologies, and route conditions, ensuring a comprehensive representation of regional transportation patterns.
• Volume Effect Parameters ($\beta$): Based on detailed cargo manifests and cost data from regional logistics operators, we determined $\beta$ = 0.93, reflecting slightly sub-linear volume effects. This calibration incorporated variations in cargo classifications, handling requirements, and batch characteristics specific to the region’s industrial profile.
• Temporal Effect Parameters ($f\left(t\right)$): Utilizing 24 h traffic flow monitoring data from Hohhot and Baotou urban areas, we determined the temporal variation function as $f\left(t\right)=1+0.25·sin\left(\pi ·t/12+\pi /6\right)$, where t represents the hour of the day (0–24). This function captures traffic congestion patterns during morning (7–9 a.m.) and evening (5–7 p.m.) peak periods, with cost increments of up to 25% during these intervals.
• Processing Cost Parameters ($\delta$): Through a detailed analysis of operational data from five major logistics distribution centers in the region, we determined $\delta$ = 1.8, which captures the nonlinear relationship between facility utilization and processing costs. The empirical data revealed that processing costs increase significantly when utilization exceeds 85% or falls below 40%, reflecting efficiency losses at both extremes.

4.3.2. Environmental Impact Parameter Sources and Determination

The environmental impact parameters were determined based on the following authoritative data sources.

Carbon Emission Parameters: These parameters were derived from the regional carbon emission factor database provided by the Inner Mongolia Development and Reform Commission. The calibrated parameters are:

• Basic emission factor: 0.062 kg CO₂/ton-km;
• Congestion-related emission factor: 0.0015 kg CO₂/ton²-km;
• Idling emission factor: 0.023 kg CO₂/h.

These values reflect the specific vehicle fleet composition and road conditions in the region. The carbon price was established at 60 CNY/ton based on current regional carbon market trading prices.

Land Use Impact Parameters: These parameters were determined based on the ecological sensitivity assessment report provided by the Inner Mongolia Ecological and Environmental Department. The land impact coefficient varies according to the land type:

• Grassland: 1.4 (high sensitivity);
• Farmland: 1.2 (medium sensitivity);
• Urban/developed areas: 0.9 (lower sensitivity).

The ecological sensitivity indices for each candidate location were derived from GIS-based environmental impact assessment reports, incorporating factors such as biodiversity significance, soil erosion risk, and ecosystem fragility.

Water Resource Parameters: These parameters were determined based on water resource utilization data provided by the Inner Mongolia Water Resources Department. The water consumption coefficient ranges from 0.25–0.35 m³/ton depending on the scale and operational characteristics of the logistics center. The regional water resource limitation was established at 3.8 million m³/year based on local water resource carrying capacity assessments.

4.3.3. Multi-Modal Transportation Parameters

The multi-modal transportation parameters were calibrated based on regional transportation network characteristics and operational data.

Mode Transfer Costs:

• Highway–railway transfer: 85 CNY/ton;
• Highway–air transfer: 220 CNY/ton.

Unit Transportation Costs by Mode:

• Highway: 2.0 CNY/ton-km;
• Railway: 1.2 CNY/ton-km (for distances > 300 km);
• Air: 15.5 CNY/ton-km.

These meticulously calibrated parameters ensure that the model accurately represents the operational realities of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, incorporating both economic efficiency considerations and environmental sustainability imperatives specific to this ecologically sensitive region.

5. Implementation and Analysis of GA

5.1. Rationalization for Selecting the GA

This section provides a systematic rationale for employing a GA to solve the logistics distribution center location optimization problem in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

The logistics distribution center location optimization problem is categorized as NP-hard, with computational complexity increasing significantly when considering environmental factors and complex constraints. Daskin and Owen [45] demonstrated that even basic facility location problems are NP-complete, with solution spaces growing exponentially with problem size. The GA offers distinct advantages in the following aspects.

• The capability to handle nonlinear cost functions: Our model incorporates various nonlinear relationships. Although these relationship parameters have been carefully calibrated to maintain consistency with simplified model outcomes, an algorithm capable of handling nonlinear constraints remains essential. Ehtesham et al. [61] demonstrated through comparative analysis that GAs process nonlinear constraints 37–52% more efficiently than traditional exact algorithms.
• The multi-objective optimization capacity: Balancing economic efficiency with environmental sustainability requires an algorithm capable of handling potentially conflicting objectives. Ma et al. [62] established that for bi-objective logistics network optimization problems, GAs yield superior Pareto solution sets compared to other heuristic algorithms.
• The global search capability: The solution space for the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration problem is expansive and complex, involving numerous discrete and continuous variables. GAs facilitated effective global searching through crossover and mutation operations. Guo et al. [63] demonstrated that GAs surpass simulated annealing (SA) algorithms by 28.6% in avoiding local optima.
• Constraint handling flexibility: Our model incorporates multiple complex constraints, including capacity limitations and environmental restrictions. GAs flexibly address these constraints through fitness functions and repair mechanisms. Gholizadeh et al. [53] established that GAs achieve significantly higher feasible solution rates than alternative algorithms when handling problems with numerous constraints.
• Adaptability to irregular solution spaces: Environmentally sensitive optimization problems typically present highly irregular solution spaces. GAs do not depend on solution space continuity or differentiability, providing superior adaptability—particularly crucial in ecologically sensitive regions like the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

Consequently, the GA represents a rational choice for this research, effectively balancing the computational efficiency with solution quality, and is particularly suitable for addressing complex regional logistics network optimization problems in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

5.2. Algorithm Framework and Parameters

Considering the complexity of logistics distribution center location optimization in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration, this study develops an improved GA framework. The framework addresses multi-objective optimization through specialized encoding mechanisms, selection strategies, and adaptive operations.

5.2.1. Key Parameters and Calibration

Based on problem characteristics and preliminary experimental analysis, the following parameter configuration was adopted (Matlab):

• Basic Parameters:

$$Population Size=100;$$

$$Chromosome Length=324;$$

$$Maximum Generation=500;$$

$$Elite\_rate=0.1.$$

2.

Genetic Operation Parameters:

$$Crossover Rate=0.85;$$

$$Mutation Rate=0.05;$$

$$Selection Pressure=0.15.$$

In addition to these basic parameters, the environmental impact parameters were calibrated through the following procedures.

Land impact coefficients (${\lambda }_{\mathrm{j}}$): These parameters were determined based on ecological sensitivity assessment data from the Inner Mongolia Ecological Environment Department. The land impact coefficients vary according to the land type (grassland: 1.4, farmland: 1.2, and urban: 0.9). The ecological sensitivity index (${\epsilon }_j$) is determined based on specific site assessment reports.

Water consumption coefficients (${\omega }_j$): These parameters were determined through the analysis of water resource utilization data from regional logistics distribution centers. The water consumption coefficients range from 0.25–0.35 m³/ton depending on the scale and operational characteristics of the logistics center. These values were validated against historical water consumption records from existing logistics distribution centers in the region.

The parameter configuration determined through an orthogonal experimental design achieves balance between the computational efficiency and solution quality. The population size of 100 represents the optimal value verified to maintain diversity; the maximum generation of 500 was established through convergence analysis, as improvements beyond this value were less than 0.01%. A crossover rate of 0.85 and mutation rate of 0.05 were determined through comparative experiments, ensuring a global search capability while maintaining solution stability. Specifically, the chromosome length of 324 is derived from the problem scale (4 × 9 × 9), while the elite ratio of 0.1 ensures algorithmic convergence while preserving diversity. These parameter configurations have demonstrated effective adaptability in logistics network optimization problems of a similar scale.

5.2.2. Data Encoding and Decoding

A binary-integer hybrid encoding scheme is adopted, with a chromosome length of 324, including the following.

Binary section: representing the selection status of nine candidate locations (one indicates selected, zero indicates not selected).

Integer section: representing transportation flows from suppliers to distribution centers and from distribution centers to customers.

The initial population is generated through a random strategy considering capacity constraints and customer demands, with a population size of 100, which experimental evidence confirms can maintain population diversity while ensuring computational efficiency.

5.3. Algorithm Implementation and Performance Analysis

5.3.1. Fitness Function Design

The fitness function integrates multiple optimization objectives, with calculation steps as follows:

• Decode the chromosome to obtain the distribution center selection scheme and transportation flows;
• Calculate various costs:

Transportation cost: based on a nonlinear distance decay model;

Fixed cost: the sum of construction costs for selected distribution centers;

Operational cost: based on a nonlinear function of facility utilization;

Carbon emission cost: based on the transportation distance, flow, and carbon price;

Land resource impact: based on the land ecological sensitivity index;

Water resource consumption: based on regional water resource limitations.

3.

Calculate constraint violation penalties;

4.

Return fitness value (cost minimization problem).

5.3.2. Genetic Operations Design

This research employs a roulette wheel selection method with an elite retention strategy, specialized crossover operations, and adaptive mutation operators.

• Selection Operation: retains the top 10% of individuals, with the remainder selected via the roulette wheel method, with probability proportional to fitness;
• Crossover Operation: Two-point crossover designed specifically for the center selection segment and flow segment, ensuring the generation of feasible solutions;
• Mutation Operation: Adaptive mutation rate based on population diversity, performing bit flipping for the binary section and Gaussian perturbation for the integer section;
• Termination Conditions: the algorithm terminates when any of the following conditions are met: the maximum iteration count reached (500 generations); an optimal solution improvement less than 0.01% for 50 consecutive generations; a population diversity index below the threshold of 0.01.

5.3.3. Implementation Details and Constraint Handling Mechanism

Chromosome Structure Design: We implemented a binary-integer hybrid encoding scheme.

Binary section (9 bits): represents candidate location selection status (1 = selected, 0 = not selected).

Integer section (4 × 9 + 9 × 31 = 315 bits): represents logistics flows from suppliers to distribution centers and from distribution centers to customers.

Fitness Function Construction: the fitness function employs the negative value of total cost, converting the problem to maximization:

$$fitness\left(x\right)=-\left[TC\left(x\right)+FC\left(x\right)+CE\left(x\right)+OC\left(x\right)+LC\left(x\right)+WC\left(x\right)\right]$$

where $TC$ represents the transportation cost, $FC$ fixed cost, $CE$ carbon emission cost, $OC$ operational cost, $LC$ land impact cost, and $WC$ water consumption cost.

Specialized Constraint Handling Mechanism: We implemented a repair-based constraint-handling strategy.

• Adjusting logistics flow allocation to satisfy capacity constraints;
• Prioritizing adjustments to distribution centers with a higher environmental impact;
• Employing a greedy strategy to retain distribution centers with optimal cost–benefit ratios.

Our repair-based constraint handling strategy employs a systematic approach to maintain solution feasibility.

For capacity constraint violations: when ${\displaystyle \sum X_{ij}>C_0·{\theta }_s·{\eta }_j·Z_j}$, we apply proportional adjustment: $X_{ij}=X_{ij}·(C_0·{\theta }_s·{\eta }_j·Z_j/{\displaystyle \sum X_{ij}})$.

For environmental constraints (water and land use): when ${\displaystyle \sum [w_j·({\displaystyle \sum X_{ij}})]·Z_j>WL}$, we prioritize adjustments to logistics distribution centers with higher water consumption coefficients: $X_{ij}=X_{ij}·(1-{\delta }_j)$, where ${\delta }_j\propto w_j$.

When multiple constraints are violated simultaneously, we implement a hierarchical approach:

• First address hard constraints (capacity and flow balance);
• Then adjust for environmental constraints;
• Finally ensure binary constraints for logistics center selection.

This repair mechanism preserves the solution quality by maintaining the proportional structure of logistics flows while ensuring feasibility across all constraint dimensions.

The optimization process follows a structured genetic algorithm approach, as illustrated in Figure 5, which presents the pseudocode for the GA applied to logistics distribution center location optimization. This algorithm systematically evaluates potential locations while considering the defined constraints.

Algorithm Parameter Configuration: Parameters were systematically determined through an orthogonal experimental design.

• Population size (100): balances diversity maintenance with computational efficiency;
• Crossover rate (0.85): promotes comprehensive solution space exploration;
• Mutation rate (0.05): maintains diversity while avoiding excessive random searching;
• Elite retention rate (0.1): ensures algorithm convergence while preserving population diversity.

Parameter Calibration Note: Although our research introduced nonlinear cost function models to capture the complexities of real-world logistics systems more accurately, we employed a systematic parameter calibration approach (resulting in values such as $\alpha$ = 0.85 and $\beta$ = 0.93) that maintains consistency in total cost outcomes while offering enhanced granularity in cost component analysis. This methodological approach preserves the theoretical advantages of nonlinear modeling frameworks while ensuring the practical applicability and interpretability of the optimization results.

After establishing the implementation details of the GA, it is essential to evaluate its performance against alternative optimization approaches to validate its effectiveness for the logistics distribution center location problem. The following section presents this comparative analysis.

5.4. In-Depth Algorithm Comparative Analysis

Having established the theoretical rationale for selecting the GA and detailed its implementation, this section presents a comprehensive comparative analysis with alternative optimization methods to validate the effectiveness and robustness of our approach. The comparison encompasses multiple performance dimensions and provides empirical evidence for the superiority of the GA in solving the logistics distribution center location problem.

5.4.1. Multi-Dimensional Performance Comparison of Algorithms

A systematic comparison was conducted between the genetic algorithm (GA) and three widely recognized optimization approaches: linear programming (LP), simulated annealing (SA), and particle swarm optimization (PSO). All algorithms were implemented under identical hardware environments and termination conditions to ensure comparative validity and eliminate implementation bias. The optimization results are presented in

Table 3. Comparison of economic costs among different algorithms (million CNY).

Algorithm	Total Cost	Fixed Cost	Transportation Cost	Operating Cost
GA	21,493,952.929	88,379.320	21,382,489.721	860.418
LP	23,676,892.432	88,379.320	23,585,426.547	743.307
SA	22,785,363.510	121,905.400	22,659,754.372	1214.262
PSO	22,108,433.718	88,379.320	22,017,264.836	526.141

and

Table 4. Comparison of environmental costs and distribution center selection across Aagorithms (million CNY).

Algorithm	Carbon Emission Cost	Water Cost	Land Use Impact	Selected Distribution Centers
GA	1245.640	398.347	629.483	Hohhot, Ordos, Zhungeer, Helingeer
LP	1343.258	356.785	643.215	Hohhot, Ordos, Zhungeer, Helingeer
SA	1289.476	414.688	785.312	Hohhot, Baotou, Ordos, Zhungeer
PSO	1263.421	374.560	625.440	Hohhot, Ordos, Zhungeer, Helingeer

.

The GA achieved a total cost solution of 21,493,952.929 million CNY, demonstrating its superior performance by outperforming LP by 9.2%, SA by 5.7%, and PSO by 2.8%. This substantial cost reduction validates the efficacy of GA in addressing the complex, multi-objective nature of logistics distribution center optimization.

Notably, with the exception of SA, all methods identified identical distribution center configurations (Hohhot, Ordos, Zhungeer, and Helingeer), confirming the robustness of this spatial arrangement. The convergence to a common solution across multiple methodologies strongly suggests that this configuration represents a globally optimal or near-optimal arrangement for the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

The GA demonstrated a particularly strong performance in transportation cost optimization, achieving reductions of 2.9% compared to PSO, 5.8% compared to SA, and 10.2% compared to LP. Regarding environmental metrics, carbon emission costs were reduced by 1.4% compared to PSO, 3.4% compared to SA, and 7.3% compared to LP, while water consumption and land-use impact indicators showed similar advantages. These results confirm GA’s superior capability in simultaneously optimizing both economic and environmental objectives.

5.4.2. Constraint Handling and Scalability Analysis

Beyond solution quality, effective optimization algorithms must demonstrate robust constraint handling capabilities and computational scalability when addressing real-world logistics problems characterized by complex constraints and large-scale dimensions.

Table 5. Comparison of constraint handling capabilities across algorithms.

Algorithm	Initial Feasible Solution Ratio (%)	Final Feasible Solution Ratio (%)	Average Violation Degree	Worst Violation Degree
GA	62.3	100	0	0
LP	100	100	0	0
SA	43.8	97.2	1.35%	4.87%
PSO	54.7	98.4	0.83%	3.26%

provides a comparison of constraint handling capabilities across various algorithms, highlighting their relative strengths in managing the complex constraints typical of logistics optimization problems.

The constraint handling analysis reveals that both the GA and LP guaranteed 100% feasible solutions, while SA and PSO exhibited minor constraint violations (2.8% and 1.6% of solutions, respectively). Despite a relatively low initial feasible solution ratio (62.3%), the GA ensured complete solution feasibility through specialized repair mechanisms that effectively navigate complex constraint spaces.

A deeper examination of constraint handling mechanisms reveals that the repair-based approach employed by the GA demonstrates superior effectiveness in managing complex environmental constraints. When confronted with conflicting constraints—a common occurrence in ecologically sensitive regions—the hierarchical repair methodology implemented in GA sequentially addresses capacity constraints, flow balance requirements, and environmental limitations, facilitating the identification of balanced solutions that satisfy all constraint categories.

In contrast, while LP guarantees mathematical feasibility, its performance deteriorates significantly under nonlinear environmental constraints, resulting in a substantially compromised solution quality. The probabilistic acceptance mechanism characteristic of SA frequently yields infeasible solutions when constraint conflicts arise, particularly in environmentally sensitive scenarios. PSO, despite its adaptive capabilities, exhibits diminished efficiency when processing hybrid discrete–continuous constraints inherent in logistics network optimization problems.

The effectiveness of GA’s repair mechanism is particularly evident when evaluating environmental constraint satisfaction rates. The algorithm maintains perfect feasibility while achieving a 72.8% reduction in carbon emissions and a 42.0% decrease in water resource consumption—critical achievements in the arid Hohhot–Baotou–Ordos–Ulanqab region, where water availability fluctuates by up to 25% seasonally. This demonstrates GA’s practical utility for sustainable logistics planning in ecologically sensitive areas such as the Inner Mongolia Plateau.

Computational efficiency analysis revealed that LP exhibited the shortest execution time (68.2 s) but inferior solution quality. Although the GA required more computational time (325.6 s), it achieved convergence in fewer iterations (287) than both SA (346) and PSO (312), demonstrating superior convergence efficiency per iteration.

The computational efficiency of optimization algorithms becomes increasingly critical when dealing with large-scale logistics networks.

Table 6. Scalability comparison across algorithms.

Problem Size	GA Computation Time (s)	LP Computation Time (s)	SA Computation Time (s)	PSO Computation Time (s)
4 × 9 × 31 (Original)	325.6	68.2	283.7	268.4
6 × 12 × 40	562.3	1384.5	487.2	504.7
8 × 15 × 50	873.8	Memory Overflow	820.5	857.1
10 × 20 × 60	1274.6	Memory Overflow	1358.2	1412.8

presents a scalability comparison across algorithms, demonstrating their performance characteristics when handling problems of increasing complexity and size.

From a theoretical complexity perspective, LP exhibits $O(2^n)$ complexity for mixed-integer problems due to branch-and-bound implementation, explaining memory overflow at scales beyond 8 × 15 × 50. In contrast, GA demonstrates $O\left(g·p·n\right)$ polynomial complexity, where g represents generations, p population size, and n problem dimensions. SA shows $O\left(g·n\right)$ complexity but requires significantly more iterations (346 vs. GA’s 287) due to extended convergence periods at lower temperatures. PSO shares GA’s polynomial complexity but exhibits an inferior performance with complex environmental constraints due to less effective repair mechanisms. The empirical results consistently align with these theoretical assessments, with GA maintaining a superior solution quality and reasonable computation times across all tested problem scales.

5.4.3. Analysis of Algorithm Mechanism and Problem Characteristic Matching

The logistics network optimization problem in the Hohhot–Baotou–Ordos–Ulanqab region exhibits several distinctive characteristics that influence algorithmic selection and effectiveness: (1) multi-objective optimization requirements necessitating the simultaneous consideration of economic efficiency and environmental sustainability; (2) nonlinear cost structures, particularly distance decay effects and regional traffic congestion factors; (3) complex constraint sets encompassing logistical, environmental, and region-specific limitations; (4) discrete–continuous mixed decision variables; and (5) high solution space complexity, with potential logistics center combinations reaching 2⁹, each associated with ≥300 logistics flow variables.

The GA demonstrates exceptional alignment with these problem characteristics for several fundamental reasons. First, GA’s chromosome-encoding mechanism naturally represents the compound decision structure of the logistics center location and flow allocation. Binary-integer hybrid encoding directly maps to the problem’s decision variable space, enabling a more efficient representation of solutions under complex constraints.

Second, GA’s population evolution mechanism is particularly suitable for addressing the balanced optimization problems in the Hohhot–Baotou–Ordos–Ulanqab region. Multiple populations evolving simultaneously avoid the limitations of single-point searches, and crossover operations effectively combine advantageous features from different solutions (such as the economic efficiency of one solution with the environmental sustainability of another).

Third, GA’s constraint handling mechanism aligns well with the composite constraints in the region. Through repair strategies rather than penalty function methods, GA can more precisely manage hard constraints in environmentally sensitive areas while maintaining solution feasibility. This mechanism proves especially valuable in the ecological protection zones of the Inner Mongolia Plateau, where environmental constraints must be strictly satisfied.

Finally, compared to alternative algorithms, GA’s parallelism provides superior computational efficiency when processing large-scale logistics network problems. LP is limited by nonlinear constraints, SA is susceptible to local optima, and PSO lacks flexibility in handling mixed variables, whereas GA, through its implicit parallel characteristics, can simultaneously explore multiple promising solution regions.

As illustrated in Figure 6, the GA demonstrates a well-balanced performance profile across all evaluation dimensions, with particular strengths in the solution quality, constraint handling capability, and algorithm stability. LP excels in computational efficiency, but underperforms in solution quality and scalability. The SA algorithm shows a relatively weaker performance across all indicators, while the particle swarm algorithm demonstrates a moderate performance in computational efficiency and algorithm stability. This multi-dimensional visualization confirms the GA’s superior suitability for the complex logistics network optimization problem addressed in this study.

5.5. Convergence Analysis and Stability Evaluation

Algorithm convergence characteristics and solution stability under parameter variations serve as critical indicators of optimization reliability. This section presents a detailed analysis of these aspects to further validate the robustness of the GA approach for the logistics distribution center location problem, focusing on both convergence characteristics and solution stability under varying conditions.

To provide visual evidence of the different convergence behaviors, Figure 6 illustrates the convergence trajectories of the three heuristic algorithms over 500 iterations.

As demonstrated in Figure 7, the GA exhibits rapid convergence during the first 100 iterations, followed by a stable optimization phase. The SA algorithm displays pronounced initial convergence lag, accelerating after approximately 150 iterations a the temperature decreases. The particle swarm algorithm converges at a rate comparable to GA, but encounters local optima challenges in later optimization phases, resulting in a marginally inferior solution quality. These distinctive patterns confirm that the GA designed in this study offers significant advantages in balancing global exploration and local exploitation capabilities.

A detailed phase-specific analysis of convergence trajectories reveals that the GA exhibits distinctive advantages across three critical stages:

• The initial exploration phase (generations 1–50): the GA demonstrates an improvement rate of 0.38% per generation, which is 3.17 times greater than SA (0.12% per generation), highlighting its powerful global search capabilities;
• The mid-term optimization phase (generations 51–200): the GA and PSO exhibit comparable improvement rates (approximately 0.09% per generation), although GA maintains a consistently superior solution quality;
• The fine-tuning phase (beyond generation 200): GA continues to produce incremental improvements (0.01% per generation), while PSO essentially plateaus.

The GA exhibits superior convergence characteristics, demonstrating an effective balance between exploration and exploitation mechanisms. During the initial phase (iterations 1–100), high genetic diversity is maintained within the population, facilitating the comprehensive exploration of the solution space. As the evolutionary process advances, the selection pressure is gradually intensified, concentrating the search within promising regions while maintaining sufficient population diversity to avoid premature convergence.

This balanced convergence behavior is particularly advantageous when addressing the complex multi-dimensional solution landscape characteristic of the Hohhot–Baotou–Ordos–Ulanqab logistics network optimization problem. In such environments, multiple local optima emerge from the complex interactions between economic efficiency objectives and environmental sustainability constraints, necessitating robust global search capabilities. Of particular significance is the GA’s demonstrated efficiency in handling environmental constraints—a critical consideration in the ecologically sensitive Hohhot–Baotou–Ordos–Ulanqab region. The algorithm’s evolutionary mechanisms prove exceptionally well-suited for simultaneously addressing water resource limitations and land impact considerations while maintaining economic efficiency. Empirical evidence reveals that GA’s constraint-handling architecture preserves environmentally optimal solution characteristics throughout the evolutionary process, with environmental performance metrics stabilizing after approximately 265 generations. This characteristic is especially valuable when optimization must balance stringent ecological requirements with operational demands, as frequently encountered in sustainability-oriented logistics network design problems.

The efficacy of this convergence mechanism is quantitatively demonstrated through population diversity metrics. An analysis reveals an average Hamming distance between chromosomes of 87.3 at initialization, reducing to 42.8 by iteration 150, and stabilizing at approximately 23.6 during later iterations. This controlled diversity reduction pattern indicates the algorithm’s capacity to maintain sufficient exploration potential while progressively focusing computational resources on high-quality solution regions. Furthermore, the algorithm achieves near-optimal solutions (within 1% of the final solution quality) by iteration 189 on average, significantly outperforming comparative methods (271 iterations for SA and 225 for PSO).

To evaluate the reliability of different optimization approaches, we conducted a stability analysis across multiple algorithms run.

Table 7. Stability comparison of different algorithms (total cost; million CNY).

Algorithm	Best Solution	Average Solution	Worst Solution	Standard Deviation	Coefficient of Variation
GA	21,493,952.929	21,582,737.428	21,897,533.764	115,842.573	0.54%
LP	23,676,892.432	23,676,892.432	23,676,892.432	0	0%
SA	22,785,363.510	23,157,482.675	23,893,461.374	325,673.218	1.41%
PSO	22,108,433.718	22,385,927.165	22,944,175.829	246,791.354	1.10%

provides a stability comparison of different algorithms in terms of total cost (million CNY), demonstrating the consistency of results that can be expected in practical logistics scenarios.

Convergence characteristic analysis revealed that the GA demonstrated rapid convergence within the first 100 iterations, followed by a stable optimization phase. The SA algorithm exhibited significant initial convergence lag, accelerating after approximately 150 iterations as the temperature decreased. The particle swarm algorithm converged at a rate comparable to GA, but encountered local optima challenges during later optimization phases, resulting in a marginally inferior solution quality. The average number of iterations required to reach near-optimal solutions (within 1% of the final solution) was 189 for GA, 271 for SA, and 225 for PSO.

We specifically analyzed the algorithm stability under varying environmental parameter settings. The results demonstrated that the GA maintained superior stability under environmental parameter variations (average coefficient of variation 0.68%), significantly outperforming SA (1.73%) and PSO (1.35%), confirming its robustness in environmentally oriented logistics network optimization.

We further analyzed algorithm performance under varying conditions to assess adaptability.

Table 8. Algorithm stability comparison under different environmental parameters. (Coefficient of variation.)

Environmental Parameter Variation	GA	SA	PSO
Carbon Price +50%	0.61%	1.68%	1.23%
Land Impact Coefficient +30%	0.59%	1.52%	1.18%
Water Resource Constraint −20%	0.83%	1.98%	1.64%

presents the algorithm stability comparison under different environmental parameters, measured by coefficient of variation, showing how consistently each method performs across changing logistics scenarios.

Based on a comprehensive multi-dimensional comparative analysis, the enhanced GA developed in this study effectively balances global exploration with local exploitation capabilities, making it particularly suitable for complex regional logistics network optimization problems such as the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration. The algorithm obtains high-quality solutions within a reasonable computational time while demonstrating robust adaptability to parameter variations and problem scale expansion, providing a reliable decision support tool for regional logistics system integration.

5.6. Comprehensive Sensitivity Analysis and Scenario Verification

To rigorously validate the robustness of the proposed optimization approach and examine parameter variation impacts, a systematic sensitivity analysis was conducted across algorithmic and model parameters, supplemented with scenario testing to evaluate adaptability under conditions specific to the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

5.6.1. Algorithm Parameter Sensitivity Analysis

The performance of GAs is significantly influenced by parameter configuration. To establish optimal parameter settings, a systematic sensitivity analysis was conducted on key algorithm parameters, as presented in

Table 9. Impact of crossover rate on algorithm performance.

Crossover Rate	Average Total Cost (Million CNY)	Convergence Iterations	Optimal Solution SD (%)	Average Feasibility Rate (%)
0.70	21,687,529.433	312	0.87	94.5
0.75	21,612,385.521	295	0.83	95.8
0.80	21,543,941.267	279	0.71	97.3
0.85	21,493,952.929	287	0.54	100.0
0.90	21,525,847.332	293	0.68	98.7
0.95	21,634,758.945	318	0.92	93.2

.

The results demonstrate that the algorithm performance reaches its optimum at a crossover rate of 0.85, balancing the exploration capability with solution stability. Lower rates (0.70) lead to insufficient population diversity and local optima vulnerability, while excessive rates (0.95) disrupt elite individuals and destabilize convergence patterns.

The crossover rate exhibits a nonlinear relationship with the solution quality and environmental constraint feasibility. While rates between 0.80 and 0.85 yield optimal total cost outcomes, rates exceeding 0.90 are associated with a 6.8% reduction in environmental constraint feasibility, suggesting that excessive recombination potentially disrupts environmentally optimal solution structures established through evolution.

Additional analyses revealed that a mutation rate of 0.05 achieves optimal balance between perturbation intensity and convergence stability, with rates in the 0.04–0.06 range producing optimal carbon emission reductions (72.8%). Population size analysis indicated optimal performance in 100 individuals, as larger populations increased computational requirements without a commensurate improvement in the solution quality.

5.6.2. Parameter Sensitivity Analysis

To evaluate the model robustness, systematic analyses were conducted on key economic and environmental parameters within specified variation ranges.

Table 10. Impact of transportation cost variations on system performance.

Parameter	Variation Range	Total Cost Impact (Million CNY)	System Performance Change	Critical Threshold	Network Configuration Change
Transportation Cost	±20%	18,730–24,780	Average Distance: 168.4 to 138.1 km	+15%	>+15%: Helingeer to Baotou
Demand Fluctuation	±30%	19,110–24,880	Service Level: 99.5 to 96.4%	+20%	>+20%: Dalate center added
Capacity Constraint	±25%	24,760–22,960	Service Level: 93.8 to 98.8%	−10%	<−10%: Dalate center added
Carbon Price	50–200 CNY/ton	21,490–24,070	Carbon Reduction: 0 to 26.5%	150 CNY/ton	>150 CNY/ton: Helingeer to Tuoketuo
Land Impact Coefficient	±30%	21,310–22,310	Land Efficiency: −5.2 to +3.8%	+25%	>+25%: Helingeer to Tumotezuo
Water Resource Constraint	±30%	23,030–21,280	Water Consumption: −23.4 to +5.2%	−15%	<−15%: Helingeer to Zhuozi

presents these results, highlighting critical thresholds triggering network configuration changes.

The analysis revealed several significant findings.

Transportation Cost: The model exhibits a linear relationship within the ±10% range (elasticity coefficient 0.82), with nonlinear effects emerging beyond ±15% (elasticity coefficient 1.17). Network reconfiguration occurs at +15%, where Helingeer is replaced by Baotou. As transportation costs increase, the system tends toward shorter-distance routes, optimizing the spatial efficiency.

Demand Fluctuation: The network maintains excellent service levels (>97.8%) within ±10% demand fluctuations, exhibiting an asymmetric response (elasticity coefficients: 0.68 for decrease, 1.24 for increase). A 20% demand increase necessitates adding a Dalate distribution center, with optimal capacity utilization identified between 85 and 95%.

Carbon Price: The relationship between the carbon price and total cost exhibits linearity below 100 CNY/ton (elasticity coefficient 0.21), with accelerated growth above 125 CNY/ton (elasticity coefficient 0.42). The 150 CNY/ton threshold triggers network reconfiguration, while at 200 CNY/ton, carbon emission reduction reaches 26.5%, suggesting an optimal policy intervention point within the 125–150 CNY/ton range.

Land Impact and Water Resource Constraints: The model exhibits moderate sensitivity to land impact coefficients, with the configuration shifting from Helingeer to Tumotezuo at the +25% threshold. Tightening water constraints by 30% increases costs by 7.1% but reduces water consumption by 23.4%, with the configuration shifting to Zhuozi County when constraints exceed 15%, demonstrating that environmental sustainability and service quality can be simultaneously maintained.

5.6.3. Multi-Parameter Interaction Analysis

To elucidate complex interactions between economic parameters (transportation costs, demand, and capacity) and environmental parameters (carbon price, land impact, and water constraints), a comprehensive multi-dimensional analysis was conducted. Figure 8 presents these interactions across six analytical perspectives.

The interaction analysis revealed three critical patterns: (1) Economic–Economic Parameter Interactions—the transportation cost and demand exhibit synergistic effects, with simultaneous increases producing a 28.4% greater cost impact than the sum of individual effects; (2) Environmental–Environmental Parameter Interactions—the carbon price and land impact coefficient interactions trigger network reconfiguration when both exceed critical thresholds (150 CNY/ton and +20%, respectively); (3) Economic–Environmental Parameter Interactions—the water constraint–demand interaction demonstrates pronounced asymmetric responses, with combined effects exceeding linear combinations by 42.7%.

The system stability region analysis identified the following stability ranges: the transportation cost (−12% to +8%), demand fluctuations (−15% to +12%), carbon price (50 to 135 CNY/ton), capacity constraints (−10% to +15%), land impact coefficient (−15% to +20%), and water resource constraints (−20% to +15%).

Nine diverse scenarios were designed to evaluate model adaptability under conditions specific to the Hohhot–Baotou–Ordos–Ulanqab region. The high-growth scenario reflects the 7.5% annual GDP growth target in the region’s 14th Five-Year Plan, while the extreme climate scenario simulates impacts of sandstorms (averaging 37.8 days annually) and severe cold weather (winter average temperature −12 °C). Resource constraint scenarios were developed considering the region’s water scarcity (annual precipitation 300–400 mm), while the environment-first scenario aligns with the region’s carbon peak target.

To evaluate the robustness of our proposed approach, we tested the model under various operational conditions.

Table 11. Model performance under different operational scenarios.

Scenario	Description	Total Cost Impact (%)	Network Configuration Adaptation	Service Level Variation (%)	Environmental Performance (%)
Economic Fluctuation Scenarios
High-growth	Demand +20%, Transportation cost −5%	+5.8	Addition of Dalate center	+1.2	−3.1 (carbon)
Low-growth	Demand ±5%, Transportation cost +10%	+4.2	Original Configuration maintained	−0.3	+0.5 (water)
Downturn	Demand −15%, Transportation cost +18%	+13.6	Helingeer replaced by Baotou	−1.5	−2.3 (land use)
Environmental Policy Scenarios
Standard	Carbon price 100 CNY/ton, Standard water constraints	+2.1	Original configuration maintained	+0.1	−5.2 (carbon)
Stringent	Carbon price 150 CNY/ton, Water constraints −15%	+9.4	Helingeer replaced by Tuoketuo	−0.2	−18.3 (carbon)
Environment-first	Carbon price 180 CNY/ton, Water constraints −25%, Land impact +20%	+17.2	Dual adjustments *	−1.8	−27.4 (composite)
Operational Condition Scenarios
Capacity optimization	Logistics center capacity +15%	−3.5	Original configuration maintained	+2.7	+1.4 (carbon)
Extreme climate	Seasonal demand fluctuation +25%, Transportation cost +22%	+24.8	Comprehensive Reconfiguration **	−4.3	+9.2 (water)
Technology upgrade	Transportation efficiency +15%, Water consumption −20%	−8.7	Original configuration maintained	+1.9	−12.6 (composite)

* Dual adjustments: Helingeer replaced by Tuoketuo, addition of Zhuozi County as alternative; ** Comprehensive reconfiguration: Maintain Hohhot and Ordos, replace Zhungeer and Helingeer with Baotou and Zhuozi.

summarizes the model performance under different operational scenarios, illustrating how the algorithm adapts to changing logistics requirements.

The scenario testing revealed remarkable adaptability across operational conditions.

Economic Resilience: under fluctuating conditions (high-growth to downturn), the model maintains service-level variations within ±1.5% through dynamic network reconfiguration, with Helingeer strategically replaced by Baotou in the downturn scenario to optimize the cost structure while preserving essential service levels.

Environmental Policy Responsiveness: under stringent environmental policy conditions, the network configuration shifted from Helingeer to Tuoketuo, achieving 18.3% carbon emission reduction with merely a 0.2% service-level decrease, demonstrating exceptional balance between economic and environmental objectives.

Operational Adaptability: the model’s most sensitive response occurred under the extreme climate scenario (cost increase of 24.8%), yet through comprehensive network reconfiguration, service-level deterioration was contained within 4.3%, demonstrating robustness under extreme conditions.

The analysis identified critical adaptation thresholds: demand fluctuations (±15%), transportation costs (+15%), carbon pricing (150 CNY/ton), and water resource constraints (−15%). The model demonstrated adaptation to region-specific challenges, maintaining 95.8% service levels during agricultural production peaks and limiting water resource consumption increases to 9.2% during extreme drought simulations.

To thoroughly understand the impact of various factors on system performance, we conducted a comprehensive sensitivity analysis across multiple parameters. Figure 9 presents this multi-parameter sensitivity analysis, showing parameter stability ranges, system stability by parameter type, parameter variation effects on total cost, and the complex network of parameter interactions.

This comprehensive analysis confirms the robustness of the proposed logistics network configuration within reasonable parameter ranges. The environmental parameters demonstrate stability ranges comparable to economic parameters, validating the model’s balanced approach to sustainability. Water resource constraints exhibited the highest sensitivity (0.76), followed by capacity constraints (0.79), with the carbon price demonstrating pronounced nonlinearity in the cost impact. The interaction network revealed strongest interactions between transportation–demand, land impact–water constraints, and demand–water constraints, providing critical insights for integrated policy development.

These findings suggest three key implementation priorities: (1) water resource management protocols; (2) a three-tiered contingency framework for parameter variations; and (3) integration of economic and environmental policy instruments, recognizing their interdependencies.

5.7. Comparison Between Optimized Solution and Existing Logistics Network

To validate the practical value of the optimized solution, we conducted a comprehensive comparison with the existing logistics network in the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration.

Table 12. Comprehensive comparison between optimized solution and existing logistics network.

Comparison Dimension	Existing Network	Optimized Solution	Improvement
Number of logistics nodes	4 major logistics distribution centers	4 strategic distribution centers	Resource integration and concentration
Network average circulation distance	264.5 km	151.2 km	Reduced by 42.8%
Annual total logistics cost	100,017,799.741 million CNY	21,493,951.929 million CNY	Reduced by 78.5%
- Transportation cost	99,889,054.986 million CNY	21,382,489.721 million CNY	Reduced by 78.6%
- Fixed cost	121,905.400 million CNY	88,379.320 million CNY	Reduced by 27.5%
- Operating cost	2177.405 million CNY	860.418 million CNY	Reduced by 60.5%
- Carbon emission cost	4661.950 million CNY	1245.640 million CNY	Reduced by 73.3%
- Land-use impact cost	1356.785 million CNY	629.483 million CNY	Reduced by 53.6%
- Water consumption cost	643.215 million CNY	398.347 million CNY	Reduced by 38.1%
Carbon emissions	Approx. 2.87 million tons/year	Approx. 0.78 million tons/year	Reduced by 72.8%
Service coverage rate	93.5%	98.5%	Increased by 5.3%
Logistics cross-overlap area ratio	18.7%	5.3%	Reduced by 13.4%
Land resource utilization efficiency	12,850 tons/km2	19,741 tons/km2	Improved by 53.6%
Ecological footprint	0.072 gha/ton	0.038 gha/ton	Reduced by 47.2%
Water resource consumption	2.45 million m3/year	1.42 million m3/year	Reduced by 42.0%

presents the multi-dimensional comparison results.

5.7.1. Structural and Efficiency Comparison

In the existing logistics network, Hohhot, Baotou, Ordos, and Ulanqab each maintained independent logistics distribution centers, resulting in a fragmented operational pattern characterized by resource redundancy, service area overlap (18.7%), coordination inefficiency, and suboptimal spatial distribution. In contrast, the optimized solution strategically selects four locations (Hohhot City, Ordos City, Zhungeer County, and Helingeer County) based on comprehensive considerations, creating a coordinated logistics service network with clear functional positioning:

• Hohhot City: the primary administrative and regional coordination center;
• Ordos City: the energy and heavy materials logistics hub;
• Zhungeer County: the coal and mineral resources distribution node;
• Helingeer County: the agricultural products and regional transfer center.

The optimized solution demonstrates superior operational efficiency: the average circulation distance has been reduced by 42.8% (from 264.5 km to 151.2 km) and the annual logistics costs have decreased by 78.5% (from 100,017,799.741 to 21,493,951.929 million CNY). This remarkable cost efficiency is derived from three complementary mechanisms.

First, the strategic reconfiguration of transportation routes has reduced the average transportation distance from 264.5 km to 151.2 km—a 42.8% decrease that directly translates into fuel savings, reduced vehicle maintenance requirements, and enhanced delivery timeliness. This efficiency is achieved not merely through distance minimization, but through intelligent routing that considers regional traffic patterns, seasonal variations, and cargo consolidation opportunities.

Second, intensive resource utilization has been facilitated through judicious logistics node selection. Rather than establishing distribution centers based solely on administrative considerations or existing infrastructure, the model identifies locations that maximize operational synergies and minimize resource duplication. This is particularly evident in the selection of Zhungeer County and Helingeer County as strategic nodes, which leverages their advantageous geographical positions while reducing infrastructure investment compared to traditional city-centric approaches.

Third, the concentration of logistics flows through strategically positioned distribution centers generates substantial economies of scale. An operational data analysis reveals that processing costs per unit decrease by 34.7% when facility utilization reaches 65–85% of the capacity—an optimal range achieved through the proposed configuration. These scale economies extend beyond processing costs to procurement, maintenance, and administrative functions, creating a virtuous cycle of cost reduction.

From a macroeconomic perspective, this cost structure optimization directly enhances regional industrial competitiveness. The proportion of logistics costs relative to GDP would decrease from 8.6% to approximately 1.9%—approaching levels observed in developed economies and potentially catalyzing industrial upgrading throughout the urban agglomeration. For energy-intensive industries predominant in the region, the anticipated 6.7% reduction in production costs could significantly strengthen their market position.

Notably, service coverage has increased from 93.5% to 98.5%, while the response time has decreased by 26.7%, demonstrating that environmental considerations need not compromise the service performance.

5.7.2. Environmental and Economic Integration Benefits

The optimized solution significantly outperforms the existing network in environmental sustainability metrics.

Carbon emissions: The 72.8% reduction in carbon emissions represents approximately 2.09 million tons of CO₂ equivalent annually, comparable to the annual carbon sequestration of 10.45 million mature trees or removing approximately 454,000 passenger vehicles from roads for one year. This significant carbon abatement directly supports the region’s compliance with the national carbon peak and neutrality targets while mitigating climate change impacts on the fragile grassland ecosystem.

Land resource efficiency: The 53.6% improvement in the land resource utilization efficiency translates to approximately 920 hectares of land preservation while maintaining an equivalent throughput capacity. This preservation carries special ecological importance in the grassland regions of Inner Mongolia, where ecosystem recovery following land disturbance typically requires decades. The optimization prioritizes brownfield development and the intensification of existing logistics assets over greenfield expansion, thereby minimizing habitat fragmentation and soil degradation risks.

Water resource conservation: The 42.0% reduction in water consumption (from 2.45 to 1.42 million m³/year) addresses a critical constraint in this arid region, where water resources are severely limited. The optimization achieves this through the strategic placement of logistics facilities in locations with lower water intensity requirements and the implementation of water-efficient operational protocols. The conserved water resources could support agricultural irrigation for approximately 470 hectares of farmland or meet the annual domestic water needs of 28,400 residents in water-stressed communities.

Particularly noteworthy is that these environmental benefits are achieved without compromising—and indeed while enhancing—the economic performance and service levels. This synergistic relationship challenges the conventional perception of inevitable trade-offs between economic efficiency and environmental sustainability, demonstrating the potential for integrated optimization approaches to simultaneously advance both objectives. The model attains this synergy through an intelligent network design that inherently reduces resource consumption while improving the operational efficiency, rather than relying on end-of-pipe environmental controls that typically increase costs.

Regional economic integration is further enhanced through three key mechanisms:

• The strategic placement of distribution centers reduces the economic distance between major economic zones within the urban agglomeration;
• The logistics configuration enables more efficient industrial division and specialization among cities, with Zhungeer County’s center supporting coal industry integration and Helingeer County’s center facilitating agricultural product distribution;
• The coordinated network enhances the efficiency of resource flows between complementary industrial clusters in different cities, supporting the formation of integrated industrial chains.

The comparison demonstrates that implementing the optimized logistics network configuration would yield substantial improvements across economic, environmental, and service dimensions compared to the existing fragmented network. These findings validate the effectiveness of the proposed optimization methodology for integrated regional logistics planning that balances economic efficiency with environmental sustainability.

5.8. Regional Implementation Challenges and Policy Recommendations

The transition from theoretical optimization to practical implementation presents several significant challenges that must be systematically addressed to realize the potential benefits of the proposed logistics network reconfiguration.

5.8.1. Implementation Challenges and Strategic Recommendations

Administrative Fragmentation and Coordination Mechanisms: The Hohhot–Baotou–Ordos–Ulanqab urban agglomeration operates under separate administrative jurisdictions with divergent planning systems and development objectives. Data from the Inner Mongolia Autonomous Region Development and Reform Commission indicates approximately 36% policy divergence among these four cities, creating substantial barriers to integrated regional planning.

Strategic Recommendation: Establish a “Hohhot-Baotou-Ordos-Ulanqab Logistics Integration Committee” comprising representatives from municipal governments, major logistics enterprises, and environmental organizations. This cross-jurisdictional body should develop unified logistics standards and implement a regional data-sharing platform enabling real-time information exchange, with the initial framework development completed by late 2025.

Resource Reallocation and Environmental Management: The optimization model recommends reconfiguring from four independent logistics centers to four strategically positioned distribution centers, necessitating significant resource redistribution. This process is complicated by heterogeneous environmental constraints—grassland ecosystem vulnerability in Zhungeer County exceeds that of Hohhot urban areas by 43%, while water resource constraints in Helingeer County are 28% more stringent than in Ordos City.

Strategic Recommendation: Implement differentiated resource management strategies, including tiered water pricing in water-stressed areas (25% surcharge for consumption exceeding baseline quotas) and land-use restrictions in ecologically sensitive regions (limiting expansion to 10% above existing levels). Concurrently, establish specific consumption standards for the logistics industry, requiring new facilities to maintain unit throughput water consumption below 0.25 m³/ton.

Investment Requirements and Technological Implementation: While the optimization scheme promises substantial operational cost reductions (78.5%), initial transformation investments of approximately 46.7 billion CNY are required. Additionally, significant disparities exist in logistics information technology adoption across the region—Hohhot exhibits a 73.6% adoption rate compared to Ulanqab’s 46.8%, while standardization level differences reach 42.3%.

Strategic Recommendation: Develop a phased implementation roadmap: (1) optimize transportation routes between existing nodes (6 months); (2) construct new logistics centers at strategic locations while maintaining existing functions (12 months); (3) gradually transfer operations to new centers (6 months); and (4) implement comprehensive technological upgrades focused on environmental sustainability (12 months). Prioritize investments in intelligent route optimization systems (ROI period 1.4 years) and multimodal transfer technologies (ROI period 2.3 years).

Carbon Pricing and Multimodal Transportation Development: The sensitivity analysis identified the carbon price as a critical parameter influencing the network configuration, with a threshold of 150 CNY/ton, triggering significant restructuring toward more environmentally favorable arrangements.

Strategic Recommendation: Implement a graduated carbon pricing mechanism beginning at 80 CNY/ton and increasing by 15 CNY/ton annually, reaching the economic–environmental optimum balance point of 125 CNY/ton within three years. Simultaneously, develop strategic rail-based logistics transfer centers between major nodes to reduce long-distance transportation emissions and construct specialized pipeline networks for energy products to minimize carbon footprints.

5.8.2. Regional Coordination Framework

The implementation of the optimization scheme requires an integrated coordination framework that balances economic efficiency with social impact considerations. A “One Core, Four Satellites” structure is proposed, positioning Hohhot as the central coordination hub with specialized satellite centers in Ordos (energy logistics), Zhungeer (coal distribution), and Helingeer (agricultural products). This structure should be supported by three essential elements.

• Strategic Infrastructure Development: prioritize investment in rapid logistics corridors connecting key nodes, the development of an integrated data exchange network, and the standardization of specialized logistics systems to support regional economic diversification;
• Employment Transition Management: address the workforce implications affecting approximately 28,500 direct jobs (15% requiring reallocation) through specialized training programs and transition support mechanisms, particularly in Baotou and Ulanqab where facility functions are being reconfigured;
• Collaborative Governance Mechanisms: establish regular “Four Centers Joint Conferences” to ensure synchronized operational strategies and implement cross-regional performance monitoring systems that integrate both economic efficiency and environmental sustainability metrics.

Through this systematic implementation approach, the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration can effectively transition toward a more efficient, integrated, and environmentally sustainable regional logistics system, creating a viable model for balanced development in ecologically sensitive regions.

6. Conclusions

This research investigated the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration as a case study for logistics distribution center optimization. Through the comprehensive analysis of regional geographical, economic, and transportation characteristics, we developed an enhanced nonlinear 0–1 mixed-integer programming model that integrates economic efficiency metrics with environmental sustainability considerations. The optimization problem was addressed using a GA, whose effectiveness was validated through rigorous comparative analysis with alternative optimization methods and existing logistics configurations.

The findings demonstrate that the optimized logistics configuration significantly reduces economic expenditures (78.5%) while simultaneously improving environmental sustainability metrics (carbon emissions reduced by 72.8% and water consumption decreased by 42.0%). This dual optimization promotes coordinated regional economic development and advances the regional integration initiative’s strategic objectives.

Our investigation offers several significant theoretical contributions. First, the enhanced nonlinear model provides a more accurate representation of complex transportation dynamics in real-world environments. Second, the integration of environmental impact factors into the optimization framework challenges conventional assumptions regarding inevitable trade-offs between economic efficiency and environmental sustainability, demonstrating their potential synergy through strategic network planning. Third, our findings indicate that optimal logistics distribution center locations should be determined through the comprehensive evaluation of logistics demand patterns, transportation accessibility, cost structures, and environmental sensitivity indices rather than being confined to established urban centers.

6.1. Research Limitations

While achieving significant results, this research exhibits several limitations affecting the applicability of the findings.

Methodological Simplifications: This study adopts a macro-level approach considering unidirectional logistics operations, whereas real-world systems typically function bidirectionally. According to the Inner Mongolia Logistics and Procurement Federation, reverse logistics accounts for 14.7% of the total regional logistics volume, and neglecting this component may compromise the network design accuracy. Similarly, the empty-load rate of transportation equipment—a critical efficiency metric currently at 38.2% in the region—was not incorporated into the model.

Data Constraints: Environmental parameter calibration primarily relied on regional averages rather than real-time monitoring data, potentially introducing ±15% error margins in environmental impact assessments according to the Inner Mongolia Department of Ecological Environment. Additionally, transportation network representations were predominantly static, inadequately reflecting seasonal variations particularly relevant in this climate-sensitive region where winter traffic interruptions average 8.3 days annually.

Application Boundaries: The research findings are specifically calibrated to the Hohhot–Baotou–Ordos–Ulanqab region’s unique characteristics and cannot be directly extrapolated to regions with substantially different ecological conditions, economic structures, or transportation networks. Furthermore, the model parameters were calibrated based on 2020–2023 data; impending policy changes, particularly the implementation of carbon reduction initiatives, may affect the long-term applicability.

6.2. Future Research Directions

Building upon these limitations, we identify three promising future research directions.

Dynamic Network Evolution Modeling: Future research should develop dynamic programming frameworks capable of modeling logistics network evolution across multiple time horizons, addressing both seasonal demand fluctuations and long-term growth patterns. Such modeling would be particularly valuable for the Hohhot–Baotou–Ordos–Ulanqab region, where agricultural and energy demands exhibit substantial seasonal variations requiring adaptive network responses.

Integrated Environmental Assessment: More sophisticated life-cycle assessment methodologies could be integrated with logistics optimization models to comprehensively evaluate environmental impacts across infrastructure construction, operation, and decommissioning phases. This approach would provide the more accurate quantification of long-term ecological effects, particularly crucial in sensitive grassland ecosystems where recovery from disturbance typically requires decades.

Multi-Scenario Robustness Analysis: Future research should develop robust optimization frameworks capable of addressing multiple uncertainties simultaneously, including demand fluctuations, extreme weather events, and policy transitions. Particularly relevant for the Hohhot–Baotou–Ordos–Ulanqab region, where sandstorms and extreme temperature variations frequently disrupt logistics operations, such modeling would enhance the system resilience and adaptive capacity.

This research provides theoretical support and practical guidance for the integrated logistics development of the Hohhot–Baotou–Ordos–Ulanqab urban agglomeration while offering valuable references for sustainable logistics planning in comparable ecologically sensitive regions. By demonstrating the achievability of synergistic optimization across economic efficiency and environmental sustainability dimensions, this study establishes a methodological framework applicable to regional sustainable development initiatives in similar contexts worldwide.