Location Allocation of Corn Stover Pretreatment Facilities in South Korea Under an Agent-Based Simulation Framework

Abstract

This research proposes a novel location allocation framework that utilizes agent-based simulation for the efficient production of corn stover-based bioethanol, which requires dedicated pretreatment facilities for the feedstock. The framework comprises two main modules: (1) a Pretreatment Facility Module that assesses the performance of the corn stover-based bioethanol supply chain based on the interactions among three types of agents, namely order agent, pretreatment agent, and transport agent, and (2) an Optimization Module designed to determine the optimal supply chain configuration by selecting the most suitable number and locations for pretreatment facilities to achieve the lowest total operational cost. The framework is implemented in a case study for South Korea, which aims to raise the bioethanol blending ratio from 4% in 2025 to 8% by 2030. Experimental results reveal that, within the bioethanol supply chain comprising eight farms and four refineries, a 1% increase in bioethanol blending ratio leads to an increase in the demand for approximately 2229 kL of ethanol (10,225 tons of corn stover), and the proposed framework enables to identify the optimal location of pretreatment facilities in the subject supply chain according to the change in ethanol demand.

1. Introduction

Biofuels are renewable energy sources derived from living organisms, including plants and animals [1], and are frequently utilized in transportation fuels such as motor fuel, marine oil, and jet fuel [2]. Bioethanol, a type of biofuel, is considered environmentally friendly, as it can reduce carbon dioxide emissions by 13% compared to conventional fossil fuels [3]. Demand for biofuels (e.g., ethanol and biodiesel) is expected to increase from 164.3 billion liters in 2022 to 188.7 billion liters in 2025 due to their application in ground and air transport [4]. Specifically, in the U.S., the Renewable Fuel Standard (RFS) established under the Energy Policy Act of 2005 (EPACT) was amended in 2007 to require the production of 845.28 billion liters of biofuel by 2025 [5,6]. Furthermore, the Brazilian government introduced its Fuel of the Future policy in 2024, which modifies the required blending ratios for biofuels and fossil fuels. The new policy raises the bioethanol blend ratio from 22% to 27% and the biodiesel blend from 14% to 20% by 2030 [7]. India’s national policy on biofuels, revised in 2022 by the Ministry of Petroleum and Natural Gas, sets a target of 20% bioethanol content in gasoline blends by 2025–2026, including production of ethanol through the use of crop residues [8,9].

Aligned with the global trend of increasing bioethanol production, South Korea has incorporated bioenergy into its first national basic plan on carbon neutrality and green growth, aiming to reduce GHG emissions to 436.6 million carbon dioxide equivalent (CO₂e) by 2030, which represents a 40% reduction compared to 2018 GHG emissions [10]. Bioenergy constitutes approximately 27% of total renewable energy production and makes up more than 25% of total renewable energy generation in the country. Notably, blending biodiesel has been mandated since 2015; the blending ratio began at 2.5% in 2015, was increased to 3.5% in July 2021, and is projected to reach 5% by 2030 [11]. South Korea’s Renewable Fuel Standard (RFS), established in 2015, requires 4% bioethanol blending in gasoline by 2025, and this target will increase to 8% by 2030, doubling the 2025 target [12].

Nonetheless, to achieve significant GHG emission reductions through bioethanol production and consumption, selecting appropriate feedstock is vital. For instance, clearing rainforests for palm oil plantations to produce biofuels may actually raise GHG emissions [13]. There is also a need to assess first-generation bioethanol and consider new raw materials to mitigate issues related to food security threats. The majority of first-generation bioethanol is produced in the U.S. and Brazil, accounting for 74% of global supply, with corn comprising 64% of this production [14]. While corn is the second most produced crop worldwide, utilizing corn byproducts such as corn stover can help stabilize the corn grain supply, which has been impacted by corn-based bioethanol production, by generating additional ethanol [15]. The use of corn stover is projected to boost global bioethanol output by over 60% [15].

Corn stover is considered one of the optimal feedstocks for second-generation bioethanol production; however, its utilization presents challenges related to lengthy pretreatment processes and the requirement for supplementary infrastructure. As a form of second-generation biomass, corn stover possesses a complex crystalline structure and, unlike first-generation bioethanol, necessitates pretreatment before its conversion into bioethanol [16]. Khan et al. [17] report that physical pretreatment of corn stover requires a minimum of 1.5 h and can extend up to 96 h, while biological pretreatment lasts from at least 1.5 h to a maximum of 30 days. Furthermore, specialized equipment is essential to support the pretreatment process, such as milling machines for physical methods and reactors for biological and chemical treatments [18,19]. Given these temporal and spatial limitations, dedicated pretreatment facilities are being established for crop by-product processing. Notably, in Europe, the Renewable Energy Group (REG), recognized as one of North America’s largest biofuel companies, commissioned a feedstock pretreatment facility in Germany in 2023 [20]. This facility significantly enhances the processing of challenging raw materials and contributes substantially to the biofuel supply chain. In addition, Valmet has implemented pretreatment facilities in Germany, Romania, and Denmark, each with the capacity to process over 0.25 million tons of biomass per year. This infrastructure addresses the increasing demand for biofuels in Europe [21].

South Korea has an advantage in installing corn stover pretreatment facilities. First, in South Korea, 70% of the country is mountains, so the area of farmland is limited [22]. Therefore, small farms are distributed across multiple locations, and this geographical characteristic can be addressed by installing separate pretreatment facilities, which are more cost-effective to operate than operating multiple pretreatment facilities on individual farms. In addition, South Korea has high access to the transport and logistics infrastructure [23]. Therefore, when pretreatment facilities are centrally installed, travel time and cost can be reduced. In other words, in countries with limited farmland area, like South Korea, installing pretreatment facilities can improve the efficiency of bioethanol production.

This study introduces an innovative location allocation framework utilizing agent-based simulation (ABS) to improve the production efficiency of corn stover-based bioethanol. In contrast to supply chain optimization methods based on mathematical programming models (such as linear programming and non-linear programming), ABS adopts a bottom-up methodology suited to making realistic performance forecasts for complex systems like supply chains, since agents with unique characteristics make autonomous decisions based on situational factors [24]. The framework comprises two main modules: (1) the Pretreatment Facility Module, which assesses the performance of the corn stover-based bioethanol supply chain, and (2) the Optimization Module, which determines the optimal number and placement of corn stover pretreatment facilities. This study models the interactions among three types of agents: (1) Order agent, (2) Pretreatment agent, and (3) Transport agent. Specifically, the pretreatment agent processes corn stover by evaluating physical, chemical, physicochemical, and biological pretreatment methods. The objective is to represent the bioethanol supply chain network using corn yield data from South Korea and minimize supply chain costs through optimal siting of pretreatment facilities by applying the mixed integer linear programming (MILP) technique within the proposed simulation-based optimization framework. The framework offers several key contributions. First, the methodology identifies the optimal location of pretreatment facilities capable of performing both physical and chemical pretreatment of biomass prior to transportation to a refinery. Placing pretreatment facilities optimally between corn farms and refineries supports the minimization of total operating costs for the bioethanol supply chain. Second, transportation costs are investigated using ABS, accounting for the variation in corn stover quantity resulting from different pretreatment technologies, and these insights are integrated into supply chain optimization. Finally, by examining the optimal siting of pretreatment facilities under diverse bio-policy scenarios in South Korea, this approach enables analysis of the impacts on both facility location and operating cost due to policy changes, supporting more effective practical management of the bioethanol supply chain.

The remainder of the paper is structured as follows. Section 2 outlines the process of pretreating corn stover for bioethanol production and presents various location allocation methods for the bioethanol supply chain. Section 3 details the agent-based simulation framework used to select pretreatment facilities that minimize the total cost within the bioethanol supply chain. Section 4 examines the optimal placement of corn stover pretreatment facilities in response to current and prospective bioethanol policies in South Korea. Finally, Section 5 summarizes the research findings and concludes the study.

2. Background

2.1. Corn Stover Pretreating Methods for Bioethanol Production

As mentioned in Section 1, corn stover is recognized as a primary raw material for second-generation bioethanol due to its lignocellulosic structure, which consists of 38–40% cellulose ${\left(C_6{H}_{10}{O}_5\right)}_n$, 28% hemicellulose ${\left(C_5{H}_8{O}_4\right)}_m$, and 7–21% lignin [25]. For bioethanol production, access to cellulose and hemicellulose within the lignocellulosic matrix is required, but this matrix is encapsulated by a protective layer called lignin [26]. As illustrated in Figure 1, the conversion of corn stover to bioethanol necessitates a pretreatment stage to separate lignin from lignocellulose.

Pretreatment methods are divided into four main categories: (1) the physical pretreatment method, (2) the chemical pretreatment method, (3) the physicochemical pretreatment method, and (4) the biological pretreatment method [27]. First, the physical pretreatment method disrupts the structure of corn stover through the application of high temperature and pressure, without the use of chemicals or microorganisms [17]. The predominant techniques include milling, ultrasonic, and hydrothermal processes; milling requires a pretreatment duration of 48 h, while ultrasonic and hydrothermal methods require up to 1.5 h and 96 h, respectively [17]. These methods increase the exposed surface area and reduce particle size, which improves accessibility; however, due to limited standalone efficiency, they are generally implemented as auxiliary steps either preceding or following other pretreatment methods [28]. Second, the chemical pretreatment method disrupts the crystalline structure of corn stover and removes lignin by applying chemicals such as oxidizers, alkaline agents, acids, and bases [29]. Acid, alkaline, and solvent-based methods are the primary approaches [30]. The average durations are approximately 55 min for acid pretreatment, several hours for alkaline pretreatment, and about 1.5 h for solvent pretreatment [31]. Chemical pretreatment methods typically have a production cost of USD 1.28 per liter of bioethanol [32], and they facilitate accelerated hydrolysis for effective biofuel production [33]. However, these approaches result in a relatively high environmental burden, including the largest Carbon dioxide (CO₂) emissions of 385 kg per kilogram of sugar produced [27], and also require specialized pretreatment reactors [34]. Third, the physicochemical pretreatment method involves integration of physical variables (such as temperature and pressure) with chemical agents [35], encompassing techniques like steam explosion, liquid hot water, and ammonia-based processes [36]. The required treatment durations are shorter compared to other categories, with steam explosion, liquid hot water, and ammonia-based procedures taking 5, 15, and 45 min, respectively [31]. These techniques enable rapid pretreatment, and among them, the ammonia-based method achieves the highest bioethanol conversion, producing up to 216 kg of bioethanol per ton of corn stover [37]. Despite these advantages, similar to other chemical routes, they emit CO₂ and necessitate the use of reactors, resulting in the need for additional equipment [34]. Finally, biological pretreatment is more sustainable, as it achieves lignin removal using microorganisms instead of chemicals [33], with brown fungi, white fungi, and soft-rot fungi being frequently utilized [17]. Nevertheless, depending on the selected microorganism, the duration can extend to 30 days, and it incurs the highest processing cost at USD 4.82 per liter of biofuel [32]. Furthermore, only 155 kg of bioethanol can be produced from one ton of corn stover [37], indicating that the degradation efficiency is low and presents challenges for industrial application [38].

Table 1. Comparison of corn stover pretreatment methods based on their characteristics [17,27,30,31,32,34,36,37].

Method	Physical	Chemical	Physicochemical	Biological
Techniques	Milling, Ultrasonic, Hydrothermal	Acid, Alkaline, Solvent	Steam explosion, Liquid hot water, Ammonia-based processing	Brown fungi, White fungi, Soft-rot fungi
CO2 Emission (kg/kg of sugar)	-	385	14.30	-
Cost (USD/liter)	-	1.28	2.11	4.82
Bioethanol Yield (kg/ton)	-	149–195	178–216	155
Time Consumption	90 min–96 h	20 min–90 min	5 min–60 min	90 min–30 days
Additional Equipment	Milling machine	Reactors	Reactors	-

outlines the characteristics of each pretreatment method with respect to CO₂ emissions, cost, yield, time consumption, and the need for additional equipment.

As each pretreatment method exhibits distinct characteristics, selecting and applying an appropriate method is essential. This study focuses on a physicochemical pretreatment method. Compared to chemical pretreatment, the physicochemical approach offers the benefit of lower CO₂ emissions and enables the rapid production of substantial quantities of bioethanol. Since locating pretreatment facilities in optimal locations can reduce the total operational cost of the bioethanol supply chain, this study proposes a framework for selecting optimal locations for pretreatment facilities.

2.2. Location-Allocation Problem and Solving Methods in a Bioethanol Supply Chain

Table 2. Literature review on location allocation in the bioethanol supply chain.

Reference	Decision Variables	Methodologies	Case Study
Ng and Maravelias (2017) [39]	Regional depot location, Biorefinery location	MILP	United States
Nazari et al. (2015) [40]	Facility location, Area allocation	MILP	Australia
Lin et al. (2020) [41]	Biofuel plant location	MOLP	Taiwan
Ranisau et al. (2017) [42]	Biorefinery location, Biomass and biofuel transportation volume	MILP	Canada
Habibi et al. (2018) [43]	Distribution center location, Distribution center quantity, Inventory policy optimization for distribution centers	GA, SA, FA	Iran
Costa et al. (2017) [44]	Siting of biodiesel manufacturing plants	Process Simulation, MILP	Colombia
Kim et al. (2018) [45]	Siting of biomass storage centers, Biomass transportation volume	ABS	United States
Singh et al. (2014) [46]	Siting of biorefineries, Biorefinery capacity optimization	ABS, GA	United States

highlights previous research on the facility location problem within bioethanol supply chains, detailing decision variables addressed, analytic methodologies employed, and the geographical areas covered by case studies. Prior studies have mainly emphasized determining parameters such as facility siting, quantity, and capacity, as well as transport volumes. Additionally, facility types considered encompassed biorefineries, biomass storage sites, and distribution centers.

Numerous problem-solving techniques have been employed for these location allocation problems, such as mathematical programming (e.g., mixed integer linear programming (MILP), multi-objective linear programming (MOLP)), simulation models, and metaheuristic algorithms, including the genetic algorithm (GA), simulated annealing (SA), and the firefly algorithm (FA). Given the interrelation of supply chain components with factors like feedstock costs, sites of bioethanol production, production expenses, and logistics costs, supply chain design is critical for effective control of bioethanol production costs [47]. As presented in Table 2, mathematical programming is the most widely adopted approach for addressing the location allocation problem. This technique involves formulating a mathematical model to represent a real-world scenario, followed by optimizing (maximizing or minimizing) an objective function to obtain the best possible solution [48]. For example, Ng and Maravelias [39] developed a MILP model to address the allocation of regional warehouse and biorefinery locations, aiming to minimize the total annual cost of the biofuel supply chain in South Wisconsin, USA. Their approach introduces linearization techniques and approximates transportation distances and logistics flows to enhance computational efficiency, but the approximated model has the disadvantage of not being able to consider the exact location and transportation distance. Similarly, Ranisau et al. [42] used MILP to select the optimal biorefinery location in Kanata, Ontario, Canada, where they found that 14.8 billion liters of biofuel could be produced at an operating cost of USD 42.822 billion.

Despite its widespread use, mathematical programming faces the issue that greater supply chain complexity leads to increased modeling complexity because of the growing number of constraints that must be addressed [49]. Thus, mathematical programming models may not adequately capture the intricate interactions present in the bioethanol supply chain [45]. To resolve this issue, simulation-based optimization has been applied. Simulation provides a more realistic solution than mathematical programming by modeling the structure of complex supply chains and the resources (human resources, equipment, machinery, etc.) actually used [50]. In particular, ABS is advantageous in finding more realistic solutions because it uses agents with autonomous decision-making capabilities according to the dynamic supply chain situation [51]. Moreover, ABS enables the detailed representation of system characteristics in scenarios where it is challenging to account for all factors at once, and it can be integrated with geographic information system (GIS) technology to create realistic supply chain models using actual location and road data [52]. In fact, Kim et al. [45] introduced a two-stage simulation approach, combining Agricultural Land Management Alternatives with Numerical Assessment Criteria (ALMANAC) and ABS, to address the location allocation problem of biomass storage facilities between farms and biorefineries in the southern Great Plains, USA. This approach attempted to predict corn yields considering climate and accurately predict the cost of feedstock transportation using GIS. Singh et al. [46] developed a method to model dynamic corn price fluctuations using ABS and applied a genetic algorithm to determine optimal facility location and capacity, aiming to maximize profits for a biorefinery in Illinois, USA. However, both studies have limitations in that they only applied ABS to first-generation biofuels.

Therefore, this study proposes a method for selecting the optimal locations of corn stover pretreatment facilities using ABS. As the bioethanol supply chain network is complex, three types of agents are modeled to provide a more realistic solution compared to the mathematical modeling. In particular, a complex ABS modeling framework that integrates GIS and MILP is proposed. Unlike previous studies, this model integrates factors such as distance and the cost of the supply chain, enabling more precise simulation results that can be applied to supply chain decision making.

3. Materials and Methods

This study develops a supply chain design for bioethanol production and determines the allocation of pretreatment facilities that minimizes total costs using agent-based simulation. Figure 2 presents the framework used for identifying the optimal pretreatment facility location. The framework is structured into two components: the pretreatment facility module and the optimization module. The pretreatment facility module comprises the order database, transport database, and pretreatment database. The order database provides information to estimate corn stover demand. The transport database contains GIS data and unit transportation costs for calculating transportation expenses (see Section 3.1 for additional details). The pretreatment database includes the costs associated with the corn stover pretreatment process. Each process step is modeled as a distinct agent to represent operations within the pretreatment facility according to the respective database information. The order agent modifies corn stover demand based on gasoline production volumes and the bioethanol blending ratio. Subsequently, corn stover is transported from the farm to the pretreatment plant by the transport agent. The pretreatment agent processes the corn stover received from the farm and evaluates both capital and operating costs of the pretreatment facility. This involves selecting the pretreatment method, preparing the required materials, and conducting the pretreatment of corn stover. Pretreated corn stover is then delivered to the refinery by the transportation agent (see Section 3.2 for further explanation). The optimization module utilizes a mathematical model to determine the optimal pretreatment facility location that minimizes the total cost of the bioethanol supply chain, drawing on inputs from the pretreatment facility module. The mathematical model is detailed in Section 3.3. Implementation is performed using AnyLogic^® University 8.7.7 simulation software, Chicago, IL, USA. Additionally, explanations of the symbols in the formulas appearing in the subsections are provided in Appendix A.

The experiment applies the ABS model to a South Korean scenario to understand how the location of pretreatment facilities is optimally selected, and evaluates the cost and environmental impact. In particular, South Korean national bioethanol policy mandates the blending of gasoline with bioethanol, with the blending ratio set at 4% by 2025 and 8% by 2030, respectively (see Section 4.1 for more details). Through the experiment, the difference in the supply chain network according to the policies for each period.

3.1. Data Collection

To determine the location for the pretreatment facility in bioethanol production, data were collected through a case study in South Korea. Figure 3 illustrates the locations of bioethanol supply chain facilities considered in this analysis.

Eight corn farms in South Korea serve as corn stover suppliers for this study.

Table 3. Information about selected farms in South Korea.

Site	Latitude (°N)	Longitude (°W)	Corn Quantity (ton)	Corn Stover Quantity (ton)
Gyeonggi	37.27	127.44	11,636	11,636
Gangwon	37.49	127.98	30,334	30,334
Chungbuk	36.79	127.58	20,346	20,346
Chungnam	36.89	126.65	2142	2142
Jeonbuk	35.8	126.88	2783	2783
Jeonnam	34.8	126.7	11,935	11,935
Gyeongbuk	36.8	128.62	4060	4060
Gyeongnam	35.32	128.26	6138	6138

lists the locations and production quantities for each of the eight farms. It represents the actual production quantities in specific regions of South Korea where corn was harvested in 2023 [53]. Among the regions analyzed, Gangwon has the highest corn production, while Chungnam has the lowest. Because corn stover is generated in a 1:1 ratio with corn harvests [54], the corn stover quantity matches the corn production. In 2023, South Korea’s total corn stover production reached 89,374 tons.

There are five refining facilities in South Korea, distributed across four regions: Ulsan, Yeosu, Seosan, and Incheon [55,56,57].

Table 4. Information about refineries in South Korea.

Location	Latitude (°N)	Longitude (°W)	Capacity (kL)
Ulsan	35.51	129.35	241,680
Yeosu	34.85	127.71	116,070
Seosan	37.00	126.40	109,710
Incheon	37.51	126.66	43,725

provides details regarding the locations and capacities of these refineries. Since Ulsan hosts two distinct refining companies, their combined capacity is treated as the overall capacity for Ulsan. Ulsan holds the highest refining capacity, whereas Incheon refinery possesses the lowest, with a capacity nearly six times less than that of Ulsan. These refineries produce gasoline suitable for blending with bioethanol, facilitating its use as a transportation fuel. In 2023, 56% of the gasoline produced at South Korean refineries was exported, and 44% was used for domestic consumption [58,59].

Pretreatment facility candidates have been identified across 8 regions throughout South Korea. For analytical purposes, the land area for a single pretreatment facility is set at 3251 m², supporting a bioethanol production capacity of up to 167,000 tons per year [20]. In this analysis, each pretreatment facility occupies 379.29 m², which is sufficient for the pretreatment of corn stover to produce 19,484 tons of bioethanol annually.

Table 5. Information about storage cost and labor cost in South Korea [60,61].

Location	Land Cost (USD/m2/Year)	Labor Cost (USD/Year)
Gyeonggi	186.03	193,231
Gangwon	41.16	31,620
Chungbuk	50.52	32,790
Chungnam	46.6	41,514
Jeonbuk	31.04	31,771
Jeonnam	23.41	41,812
Gyeongbuk	35.17	55,884
Gyeongnam	50.31	54,864

presents the land and labor costs associated with the 8 candidate pretreatment facility sites. Gyeonggi, due to its proximity to Seoul, exhibits the highest land and labor costs among the regions. Jeonnam has the lowest land cost, while Gangwon records the lowest labor cost.

Trucks are utilized to transport corn stover from farms to pretreatment facilities and subsequently from these facilities to refineries.

Table 6. Distance matrix between the farms and the pretreatment facility candidates (unit: km).

Farm	Potential Pretreatment Facility Sites
	Gyeonggi	Gangwon	Chungbuk	Chungnam	Jeonbuk	Jeonnam	Gyeongbuk	Gyeongnam
Gyeonggi	47	127	88	131	204	342	161	302
Gangwon	111	60	146	198	267	400	151	333
Chungbuk	101	184	22	103	145	285	113	264
Chungnam	83	191	99	33	165	267	239	340
Jeonbuk	199	307	138	120	22	133	268	253
Jeonnam	314	443	254	249	141	33	349	248
Gyeongbuk	184	182	145	251	279	388	37	226
Gyeongnam	340	385	248	302	186	223	197	51

shows the distances from farms to each pretreatment facility candidate. Located centrally, the Chungbuk candidate offers the shortest average transport distance from farms at 152.125 km.

Table 7. Distance matrix between the pretreatment facility candidates and the refineries (unit: km).

Refinery	Potential Pretreatment Facility Sites
	Gyeonggi	Gangwon	Chungbuk	Chungnam	Jeonbuk	Jeonnam	Gyeongbuk	Gyeongnam
Ulsan	352	385	257	352	313	337	199	104
Yeosu	321	450	261	269	152	140	288	131
Seosan	108	221	127	63	183	285	261	367
Incheon	60	142	165	133	241	365	253	396

provides data on the distances between pretreatment facility candidates and refineries. The minimum distance observed is between the Gyeonggi pretreatment facility candidate and the Incheon refinery, calculated at 60 km. By contrast, the maximum distance is between the Gangwon pretreatment facility candidate and the Yeosu refinery, measuring 450 km. The truck capacity adopted in this study is 5 tons, and the transportation cost is 0.81/km [62].

3.2. Pretreatment Facility Module

3.2.1. Order Agent

The order agent estimates the demand for bioethanol based on gasoline production quantities and the bioethanol blending ratio, subsequently requesting the corresponding amount of corn stover from farms. Figure 4 presents the state diagram of the order agent, which comprises three states: (1) set domestic demand, (2) set pretreated corn stover demand, and (3) order corn stover. Initially, to determine the required bioethanol for blending with automotive gasoline, domestic gasoline production and the established blending ratio are defined. Next, the sum of gasoline production ($Q_r^{gasoline}$) for each refinery is multiplied by the domestic gasoline production ratio (${RT}_{domestic}$) to calculate total domestic gasoline production using Equation (1). The overall corn stover demand is subsequently determined using ${RT}_{cornstover}$, which factors in both the blending and conversion ratios.

$$Q_{cornstover}=\sum _{r\in R}{Q}_r^{gasoline}\times {RT}_{domestic}\times {RT}_{cornstover}$$

(1)

An order is submitted for the calculated corn stover requirement, after which the corn stover is transported from the farm to the pretreatment facility by the transport agent. The pretreatment facility then requests corn stover supply from all farms. Through these operations, the order agent adjusts the biomass demand in response to different blending ratios and coordinates with the farms.

3.2.2. Transport Agent

The transport agent is responsible for transferring corn stover among the farm, pretreatment facilities, and refinery. As shown in Figure 5, two distinct transaction types are represented. When the order agent submits a corn stover request, the truck transports corn stover from the farm to the pretreatment facilities. In another transaction, pretreated corn stover is collected from the pretreatment agent and delivered to the refinery. The truck routes are computed via the GIS module in AnyLogic^®, supporting precise calculation of transport distances. Equation (2) in this model computes transportation cost based on both the transportation distance (i.e., $D_{fp}$, and $D_{pr}$) and transportation quantity (i.e., $Q_{fp}^{transport}$, and $Q_{pr}^{transport}$).

$${TC}_{transport}=\left(D_{fp}\times Q_{fp}^{transport}+D_{pr}\times Q_{pr}^{transport}\right)\times C_{transport}$$

(2)

3.2.3. Pretreatment Agent

The pretreatment agent operates through 7 defined states: (1) Facility initial state, (2) Set biomass, (3) Drying, (4) Disrupt cell wall, (5) Remove lignin, (6) Storage, and (7) Send to refinery. Figure 6 depicts the state diagram for the pretreatment agent. Once a region is selected for the pretreatment plant installation, the total amount of corn stover ($Q_p^{cornstover}$) received at the facility is allocated equally among the chosen pretreatment plants via Equation (3).

$$Q_p^{cornstover}=\sum _{f\in F}{Q}_f/\sum _{p\in P}{x}_p$$

(3)

In this study, because the capacity of the pretreatment plants installed in each candidate region is identical, each region is allocated an equal share of corn stover. Subsequently, the capital cost is determined using Equation (4), with regional variation in land cost contributing to the total capital cost.

$${TC}_p^{capital}=Q_p^{cornstover}\times \left(C_{Building}+C_{Equipment}+C_p^{land}\right)$$

(4)

The transported corn stover undergoes a pretreatment process, beginning in the drying state and progressing to the lignin removal state. The initial step is drying, which is essential for maintaining the quality of corn stover, as a moisture content above 20% can enhance microbial activity and cause decomposition [63]. Following the drying stage, the crystalline structure of the corn stover is disrupted by breaking down the cell wall and removing lignin. The operating cost for this process comprises both the pretreatment and labor costs, which are calculated proportionally to the quantity of corn stover using Equation (5). Labor costs were specifically set to account for regional variations.

$${TC}_p^{operation}=Q_p^{cornstover}\times \left(C_p^{labor}+C_{pretreatment}\right)$$

(5)

During the pretreatment process, the amount of corn stover decreases as impurities, including hydrophobic components, are eliminated [64]. This pretreatment process is completed within approximately 60 min.

Table 8. Pretreated corn stover quantity and bioethanol yield by pretreatment technology (edited from Zhao et al. [65]).

Pretreatment Process Design	Types of Pretreatment Methods	Corn Stover Quantity (kg)	Pretreated Corn Stover (kg)	Bioethanol Yield (kg)
Acid	Chemical	100	$59.7\pm$ 14.8	$14.5\pm$ 3.6
Alkaline	Chemical	100	$66.0\pm$ 15.7	$20.1\pm$ 4.6
Solvent based	Chemical	100	$58.7\pm$ 20.1	$19.2\pm$ 3.5
Steam explosion	Physicochemical	100	$65.6\pm$ 16.2	$16.6\pm$ 2.9
Liquid hot water	Physicochemical	100	$61.9\pm$ 12.9	$16.6\pm$ 4.1
Ammonia based	Physicochemical	100	$74.9\pm$ 17.2	$21.8\pm$ 4.7
Fungi	Biological	100	$67.7\pm$ 22.6	$11.0\pm$ 2.5
Combined	Physical, Chemical	100	$72.2\pm$ 24.6	$18.0\pm$ 5.1

presents the amounts of pretreated corn stover and corresponding bioethanol yields achieved by each pretreatment technology per 100 kg of corn stover. Among these, the ammonia-based pretreatment technology yields the highest amount of treated corn stover, producing 74.9 kg per 100 kg of initial feedstock. This technique also delivers the maximum bioethanol output, approximately 21.8 kg. In contrast, the solvent-based pretreatment technology results in the lowest yield after processing, producing about 58.7 kg.

After the completion of pretreatment, the corn stover is stored and subsequently transported to the refinery. The overall cost is ultimately determined by Equation (6). Calculated values and respective costs associated with these stages are recorded in the relevant variables.

$${TC}_p^T={TC}_p^{capital}+{TC}_p^{operation}$$

(6)

3.3. Optimization Module

The optimization module determines the best location for the pretreatment facility through a mathematical model. Equation (7) establishes the objective function, aiming to minimize the combined transport cost across the bioethanol supply chain and operating cost of the pretreatment facility.

$$\begin{array}{l}\mathrm{M}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{m}\mathrm{i}\mathrm{z}\mathrm{e}\mathrm{z}={\displaystyle \sum _{f\in F}}{\displaystyle \sum _{p\in P}}(Q_{fp}\times D_{fp}\times x_p)\times C_{transport}\\ +{\displaystyle \sum _{p\in P}}{\displaystyle \sum _{r\in R}}\left(Q_{pr}\times D_{pr}\times x_p\right)\times C_{transport}\\ +{\displaystyle \sum _{p\in P}}{x}_p\times {TC}_p\end{array}$$

(7)

Equations (8) through (13) are the constraints that support minimization of the objective function. Equation (8) ensures that the total amount of corn stover transported from each farm to pretreatment facilities does not surpass the production capacity of each farm. Equation (9) guarantees that the total yield of produced bioethanol matches the total capacity of all refineries. Equation (10) establishes that the amount of corn stover processed at pretreatment facilities is equal to the amount transported to the refinery. Equation (11) requires that the quantity of corn stover delivered to each pretreatment facility is distributed equally. Equation (12) enforces the non-negativity constraint on transportation quantities. Equation (13) specifies the binary variable associated with the pretreatment facilities.

$$\sum _{p\in P}{Q}_{fp}\times x_p\le Q_f,for\forall f\in F$$

(8)

$$\sum _{p\in P}{Q}_{pr}\times x_p\times {RT}_r=Q_r^{capcity},for\forall r\in R$$

(9)

$$\sum _{f\in F}{Q}_{fp}\times {RT}_p=\sum _{r\in R}{Q}_{pr},for\forall p\in P$$

(10)

$$\sum _{f\in F}{Q}_{fp}=\sum _{f\in F}{Q}_f/\sum _{p\in P}{x}_p,for\forall p\in P$$

(11)

$$Q_{fp},Q_{pr}\ge 0,for\forall f\in F,\forall r\in R,and\forall p\in P$$

(12)

$$x_p\in \left\{0,1\right\},for\forall p\in P$$

(13)

Therefore, the optimal locations for pretreatment facilities that minimize the aforementioned objective function while satisfying all constraints are selected. Subsequently, the quantities of corn stover allocated to the selected pretreatment facilities and refineries are determined.

4. Experiments

4.1. Scenario

The proposed ABS framework is applied to South Korean scenarios to identify the optimal location of a pretreatment facility that minimizes the total cost within the bioethanol supply chain. South Korea has implemented a national policy mandating the use of a specified percentage of bioethanol to reduce dependency on petroleum-based fuels, known as the Renewable Fuel Standard (RFS) [66]. Established in 2015, the policy updates its targets every three years. By 2025, the RFS in South Korea requires a 4% bioethanol blending with gasoline. In 2030, this target increases to 8%, doubling the 2025 goal [12]. Accordingly, bioethanol demand is calculated as 4% and 8% of domestically produced gasoline across four refineries in South Korea. Thus, two scenarios are considered in this study: (1) a 4% blending scenario and (2) an 8% blending scenario. The 4% blending scenario refers to blending 4% bioethanol with gasoline, while the 8% blending scenario refers to blending 8% bioethanol with gasoline. Additionally, ammonia-based pretreatment technology is chosen due to its effectiveness in processing corn stover with high cellulose content [38].

4.2. Validation

Model validation is performed by comparing results from the simulation model with the observed weights of pretreated corn stover in four different experiments. In a 2010 US experiment, 10 g of corn stover was pretreated with 100 mL of 15% ammonia [67]. In 2005, an experiment conducted in China pretreated 20 g of corn stover with 200 mL of 10% ammonia [68]. In another experiment, 100 g of corn stover from NREL, Golden, CO, USA, was pretreated with 500 mL of 30% ammonia [69]. Furthermore, 1000 g of corn stover harvested in China in 2012 was pretreated with 700 mL of approximately 143% ammonia [70]. The experimental conditions included temperatures set between 60 °C and 130 °C, and durations ranging from 10 min to 24 h. Figure 7 shows a comparison between the simulated and observed values for 10 g, 20 g, 100 g, and 1000 g of corn stover. A paired t-test is conducted to evaluate the consistency between simulated and observed results (see

Table 9. The result of paired t-test between simulated value and observed value.

	Simulated		Observed		t-Statistics	p-Value
	Mean	Standard Deviation	Mean	Standard Deviation
Paired t-test	211.59	311.37	198.13	291.28	1.1364	0.3383

). Since the p-value is 0.3383, exceeding the significance level of 0.05, it indicates there is no statistically significant difference between the two datasets. Consequently, the proposed simulation represents the selected pretreatment technology’s operations with acceptable accuracy. A paired t-test is calculated using the standard statistical formula.

4.3. Results

The number and location of pretreatment facilities are determined to minimize the total cost of the bioethanol supply chain while satisfying the refinery’s demand. Alternatives are considered for the installation of between one and eight pretreatment facilities, ranging from one pretreatment facility (centralized operation) to a dedicated pretreatment facility installed near eight farms supplying corn stover (decentralized operation). OptQuest in AnyLogic, a metaheuristic-based optimization engine utilizing Tabu search and Scatter search [71], is used, and the results are described in

Table 10. Annual cost (million USD/year) of the bioethanol supply chain (4% and 8% blending scenario).

Category		4% Blending Scenario				8% Blending Scenario
		Number of Pretreatment Facilities				Number of Pretreatment Facilities
		1	2	4	8	1	2	4	8
Capital cost (million USD)	Land cost	0.02	0.03	0.06	0.18	0.02	0.03	0.06	0.18
	Building expenditure	1.34	2.67	5.34	10.69	1.34	2.67	5.34	10.69
	Equipment expenditure	0.96	1.93	3.86	7.72	0.96	1.93	3.86	7.72
Operating cost (million USD)	Labor expenses	1.37	1.35	1.69	2.53	2.75	2.70	2.89	5.07
	Transport cost	1.53	1.59	1.24	1.16	3.77	3.86	3.26	2.73
	Pretreatment expenditure	1.08	1.08	1.08	1.08	2.16	2.16	2.16	2.16
Total cost (million USD/year)		6.30	8.65	13.27	23.36	11.00	13.35	17.57	28.55

. In both the 4% and 8% blending scenarios, the total cost is significantly reduced as the number of pretreatment facilities is reduced. The total cost can be reduced by 73.03% from USD 23.36 million to USD 6.30 million in the 4% blending scenario, and by 61.47% from USD 28.55 million to USD 11.00 million in the 8% blending scenario. The most significant factor in the total cost reduction is the reduction in capital costs related to the installation and management of pretreatment facilities, which can lead to a capital cost reduction of 87.52% in both the 4% and 8% blending scenarios. However, increased transportation costs are required to transport corn stover to the single pretreatment facility, with transportation costs increasing by 31.89% in the 4% blending scenario and by 38.10% in the 8% blending scenario. Notice that the difference in transportation cost between the two scenarios is the difference in corn stover transportation volume (40,902 tons in the 4% blending scenario and 81,803 tons in the 8% blending scenario). This result implies the importance of selecting the location of the pretreatment facilities and establishing an appropriate transportation policy for efficient operation of the supply chain.

Figure 8 shows the optimal locations of the pretreatment facilities selected in the 4% blending scenario and the 8% blending scenario. As described in Table 10, in both scenarios, the optimal solution is to install and operate a single pretreatment facility in Chungbuk in South Korea (CPF). These results emphasize that operating a single centralized pretreatment facility is more cost efficient than establishing multiple distributed facilities [72]. Corn stover is transported from farms to the CPF, after which the pretreated biomass is delivered to refineries. The lines demonstrate transportation flows between facilities, with line thickness directly reflecting transport volumes (see

Table 11. Transportation quantity (ton/year) of the bioethanol supply chain (4% and 8% blending scenario).

Farm	4% Blending Scenario				8% Blending Scenario
	Number of Pretreatment Facilities				Number of Pretreatment Facilities
	1	2	4	8	1	2	4	8
Gangwon (ton)	-	20,451	10,225	5113	30,334	30,334	27,124	22,763
Chungbuk (ton)	20,346	20,346	10,225	9136	20,346	20,346	20,345	20,346
Gyeonggi (ton)	11,636	105	-	5113	11,636	11,636	11,636	11,636
Gyeongnam (ton)	-	-	6138	5113	6138	6138	5733	6138
Jeonnam (ton)	-	-	10,254	7442	4364	4364	11,935	11,935
Gyeongbuk (ton)	3995	-	4060	4060	4060	4060	105	4060
Jeonbuk (ton)	2783	-	-	2783	2783	2783	2783	2783
Chungnam (ton)	2142	-	-	2142	2142	2142	2142	2142
Total transport quantity (ton)	40,902	40,902	40,902	40,902	81,803	81,803	81,803	81,803
Total distance (km)	529,596	336,848	482,918	425,188	1,943,694	1,502,609	1,483,663	1,340,422
Total transport cost (million USD)	0.43	0.27	0.39	0.34	1.57	1.22	1.20	1.09

for more detail). In Figure 8a, since the 4% blending scenario requires 40,902 tons of corn stover (46% of the total corn stover production quantity from eight candidate farms), only five farms with low transportation costs are selected for corn stover production. The optimal transportation quantities are determined by considering the corn stover production constraints of the eight farms, as depicted in Table 3, and their values are described in more detail in Table 11. In Figure 8b, Unlike the 4% blending scenario, all farms are utilized as suppliers of corn stover. The total travel distance between the farms and the CPF is 1,943,694 km, which is about 3.67 times longer than in the 4% blending scenario (see Table 11).

As shown in Figure 8, the selected pretreatment facility is located closer to the farms than the refineries. In the 4% blending scenario, the average distance between farms and the selected pretreatment facility is 96.8 km, and the average distance between refineries and the selected pretreatment facility is 202.5 km. This shows that corn stover availability is an important factor in selecting the location of the corn stover pretreatment facilities [73]. Furthermore, allocating pretreatment facilities close to the farms can contribute to reducing CO₂ emissions. As the weight of corn stover is reduced through the pretreatment processes, the yield transported from farms to pretreatment facilities is heavier than the yield transported from pretreatment facilities to the refineries. This leads to an increase in the number of trucks transported from the farms to pretreatment facilities. Therefore, installing pretreatment facilities near farms mitigates CO₂ emissions by reducing the total transport distance.

Table 11 describes the quantities of corn stover transported from farms to the pretreatment facilities. In the 4% blending scenario, a single largest volume of corn stover is transported from the Chungbuk farm (50%), which is closest to the CPF, followed by the Gyeonggi farm (28%), which is the next nearest. The Gyeongbuk farm (10%), Jeonbuk farm (7%), and Chungnam farm (5%) are next in order. Although the Chungnam farm is relatively near the CPF at a distance of 99 km, its corn stover production volume is limited (see Table 3). In contrast, the Gangwon farm, despite having the highest production volume of corn stover, is approximately seven times farther from the CPF than the Chungbuk farm, which results in no transport of corn stover from Gangwon. With increased demand (81,803 tons of corn stover) for bioethanol under the 8% blending scenario, a greater amount of corn stover is required compared to the 4% blending scenario. Since the Gangwon farm produces the largest volume of corn stover (37%), it supplies the greatest amount to the CPF. All corn stover from the Chungbuk (25%), Gyeonggi (14%), Gyeongnam (8%), Jeonnam (5%), Gyeongbuk (5%), Jeonbuk (3%), and Chungnam (3%) farms is transported to the CPF. This result reflects the correlation between transportation distance and transportation cost and the constraints on the production of corn stover that can be produced on each farm.

Table 12. Average transport quantity between selected pretreatment facility and refineries (4% and 8% blending scenario).

Number of Pretreatment Facilities		Refineries				Total Transport Quantity (ton)	Average Distance (km/Refinery)	Total Distance (km)	Total Transport Cost (Million USD)
		Ulsan (ton)	Yeosu (ton)	Seosan (ton)	Incheon (ton)
4% blending scenario	1	14,484	6956	6575	2620	30,635	340,368	1,361,471	1.10
	2	7242	3478	3288	1310	30,635	407,516	1,630,063	1.32
	4	3621	1739	1644	655	30,635	263,246	1,052,984	0.85
	8	1811	870	822	328	30,635	254,154	1,016,614	0.82
8% blending scenario	1	28,968	13,912	13,150	5241	61,271	680,629	2,722,516	2.20
	2	14,484	6956	6575	2621	61,271	814,770	3,259,081	2.64
	4	7242	3478	3288	1310	61,271	635,577	2,542,306	2.06
	8	3621	1739	1644	655	61,271	508,207	2,032,826	1.65

describes the average transport quantity from the CPF to four refineries under 4% and 8% blending scenarios. In both scenarios, the total demand for pretreated corn stover is not related to the change in the number of pretreatment facilities, so the average transport volume decreases as the number of pretreatment facilities increases. In addition, as the number of pretreatment facilities increases from 1 to 8, the total transport distance and total transport cost tend to decrease. When comparing the transport quantity by refinery in the 4% blending scenario, since Ulsan refinery has the largest refining capacity, the total quantity of pretreated corn stover transported there is about 14,484 tons (47%), which is the highest among the refineries. Yeosu and Seosan refineries receive about 6956 tons (23%) and 6575 tons (21%) of pretreated corn stover from the CPF, respectively. Incheon refinery, which has the smallest capacity, receives less than 3000 tons of pretreated corn stover (9%). For the 8% blending scenario, a total of 61,271 tons of pretreated corn stover is delivered to the refineries, with 47% transported to the Ulsan refinery, the largest recipient. Yeosu and Seosan refineries receive comparable quantities, representing 23% and 21% of the total transported, respectively. The smallest amount, approximately 5240 tons, is delivered to the Incheon refinery. The total transport cost of pretreated corn stover is relatively high because the transport distance from the CPF to the refineries is 1.40 to 2.57 times longer than the transport distance from the farm to the CPF.

4.4. Discussion

The usage of bioethanol has the effect of reducing CO₂ emissions. A total of 8997 kL of gasoline, which is 4% of the annual fuel consumption in South Korea, generates greenhouse gas of about 19,592 tCO₂. The heating value of gasoline is 30.1 MJ per liter, and corn stover-based bioethanol is 21.1 MJ [74]. The heating value of the corn stover-based bioethanol is about 70% of the heating value of gasoline. Also, the carbon emission factor of corn stover-based bioethanol is lower than that of gasoline’s carbon emission factor, which is 10.9 gC/MJ and 19.7 gC/MJ, respectively [75]. According to these values, the amount of greenhouse gas generated by the 8997 kL of corn stover-based bioethanol is 7594 tCO₂. This is about 39% of the greenhouse gas emissions from gasoline. Therefore, with the case of the 4% of the annual fuel consumption in South Korea, it is able to reduce greenhouse gas by about 12,000 tCO₂. It corresponds to approximately 0.09% of the 2025 reduction target of 14 million tCO₂ greenhouse gas for the transportation sector under the South Korean 2030 National Greenhouse Gas Reduction Target [76]. From a numerical point of view, it seems difficult to achieve the reduction goals, but as energy conversion technology advances, methods to further reduce greenhouse gas emissions continue to be developed. In particular, Galven et al. [77] showed that recovering CO₂ generated during the production of bioethanol and reusing it in other industries can significantly reduce greenhouse gas emissions. Therefore, in order to achieve the national greenhouse gas reduction goal, more subsidies should be provided to the bioethanol supply chain, and policies should be established so that more people can use bioethanol.

5. Conclusions

This study introduces a simulation framework for optimizing the location allocation of corn stover pretreatment facilities, aimed at enhancing the design of the corn stover-based bioethanol supply chain. The allocation of facilities is addressed through an agent-based simulation model, which structures the supply chain. The actions of transport agents and pretreatment agents are modeled to reflect their interactions within the supply chain. An optimization module is incorporated to minimize the total supply chain cost, determining both the optimal number and locations of pretreatment facilities. Experimental findings indicate that the installation of one pretreatment facility is optimal for the scenario of South Korea. When operating with one pretreatment facility, the total supply chain cost is USD 6,303,942/year for a 4% blending scenario and USD 11,004,892/year for an 8% blending scenario. Since the demand for ethanol increases by approximately 2229 KL (10,225 tons of corn stover) as the bioethanol blending ratio increases by 1%, the proposed framework selects the optimal location of pretreatment facilities according to the ethanol demand change. Consequently, the simulation methodology proposed in this study provides a practical decision-support tool for engineers and researchers seeking to minimize supply chain costs and maximize financial benefits.

Despite its contributions, this study presents certain limitations. In the South Korean context, farms are widely distributed, but limited data on corn production at the individual farm level led to a simplification of the supply chain network, using a representative corn-producing region for each province. Furthermore, the analysis relies on corn production quantities and gasoline demand data from 2023. Variations in future corn production or gasoline demand could significantly alter the supply chain networks compared to those identified here.

Future work will extend the application of the proposed simulation model to global supply chain networks that incorporate multiple types of agricultural residues. Additionally, the model could support bioethanol production that aligns with environmental policy objectives. The integration of a broader set of pretreatment technologies at the facilities, as well as the evaluation of alternative transportation modes beyond the exclusive use of trucks, may further enhance supply chain optimization. Furthermore, installing these facilities in rural areas can bring regional development, such as employment increase. Therefore, it is possible to expand consideration of the social impact of the supply chain network.