Agent-Based Simulation Modeling for Sustainable Chemical Production

and Resource Management

Afshin Poorkhanalikoudehi

, Thorsten Wack

2 a

, Sebastian Kliesow

2 b

, Martin Distelhoff

and Goerge Deerberg

1,2

FernUniversit

at in Hagen, Universit

atsstraße 47, 58097 Hagen, Germany

Fraunhofer-Institut f

ur Umwelt-, Sicherheits- und Energietechnik UMSICHT

Osterfelder Straße 3, 46047 Oberhausen, Germany

Keywords:

Sustainable Chemical Production, Resource Optimization, Energy Efﬁciency, Pareto Optimization, Resource

Availability-Based Selection, Industrial Network, Multi-Objective Optimization.

Abstract:

This study investigates the optimization of resource allocation and energy efﬁciency within a sustainable chem-

ical production network using three distinct methods: Resource Availability-Based Selection, Pareto-based

Selection, and Pareto Optimization. Each method was analyzed based on its ability to manage energy con-

sumption, production efﬁciency, and resource utilization across multiple iterations. The Resource Availability-

Based Selection method prioritized available resources in storage, while the Pareto-based Selection introduced

input price considerations. Pareto Optimization, the most advanced approach, balanced production efﬁciency

and cost-effectiveness, resulting in the highest overall performance. Findings demonstrate that multi-objective

optimization, particularly Pareto Optimization, enhances operational efﬁciency and sustainability. The study’s

implications suggest adopting advanced optimization strategies to achieve energy efﬁciency and sustainabil-

ity goals in the chemical industry. Additionally, recommendations for future research include incorporating

real-time market dynamics, logistical factors, and renewable energy sources into the model to further enhance

decision-making.

1 INTRODUCTION

In recent years, the chemical industry has faced in-

creasing pressure to adopt more sustainable practices

due to environmental and economic considerations.

Sustainable chemical production (SCP) minimizes

environmental impact while optimizing resource uti-

lization, ensuring long-term viability (Haleem et al.,

2023; Mishra et al., 2023). Traditional evaluation

methods often fail to capture the complexity of in-

teractions in production systems. To address this,

the purpose of this study is to develop and analyze

an agent-based modeling and simulation (ABMS)

framework for sustainable chemical production and

resource management. By simulating interactions

between production facilities, markets, and resource

providers, the study aims to optimize production pro-

cesses, reduce waste, and enhance sustainability.

https://orcid.org/0009-0000-3839-4429

https://orcid.org/0009-0008-4509-0413

2 LITERATURE REVIEW

Agent-Based Modeling and Simulation (ABMS) has

emerged as a powerful tool for analyzing complex

systems, particularly in resource allocation and sup-

ply chain management. Its ability to model au-

tonomous agents and their interactions provides valu-

able insights into system dynamics and performance.

In supply chain networks, ABMS has been inte-

grated with the Supply Chain Operations Reference

(SCOR) model to enhance the modeling of distributed

supply chain systems (Long, 2014). This integra-

tion allows for a more comprehensive analysis of sup-

ply chain processes and performance metrics. Addi-

tionally, ABMS has been combined with reinforce-

ment learning to optimize stochastic supply chains,

particularly in managing supplier disruptions (Aghaie

and Hajian Heidary, 2019). This simulation-based

optimization approach has proven effective in han-

dling uncertainties within supply chains. Further-

more, ABMS has been applied to model Liqueﬁed

Poorkhanalikoudehi, A., Wack, T., Kliesow, S., Distelhoff, M. and Deerberg, G.

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management.

DOI: 10.5220/0013339600003970

In Proceedings of the 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2025), pages 147-158

ISBN: 978-989-758-759-7; ISSN: 2184-2841

147

Natural Gas (LNG) import terminals, demonstrating

its effectiveness in supply chain management for en-

ergy sectors (Venkataramanan and Srinivasan, 2024).

These studies highlight the capability of ABMS to

evaluate real-world scenarios and disruptions, rein-

forcing its value in complex supply chain manage-

ment.

In chemical production, ABMS enables the explo-

ration of various scenarios, including the effects of

different resource management strategies on produc-

tion efﬁciency and environmental sustainability (Helo

and Rouzafzoon, 2023; Zhou et al., 2024). This capa-

bility is particularly valuable in the context of sustain-

able development, where ABMS facilitates the bal-

ancing of economic, environmental, and social ob-

jectives. Additionally, ABMS supports the integra-

tion of real-time data and adaptive strategies, making

it a powerful tool for managing dynamic and com-

plex industrial environments (Ionescu et al., 2024).

Recent studies have demonstrated its effectiveness in

optimizing production processes, enhancing resource

allocation, and improving decision-making under un-

certainty (Zhu et al., 2023).

These studies collectively demonstrate the versa-

tility and efﬁcacy of ABMS in addressing various

challenges in resource allocation and supply chain

management. By capturing the behaviors and interac-

tions of individual agents, ABMS facilitates a deeper

understanding of complex systems, leading to more

informed and effective decision-making.

3 CASE STUDY: SCP MODEL

The sustainable chemical production (SCP) model

is a comprehensive agent-based simulation frame-

work designed to optimize and manage chemical pro-

duction and resource allocation within an industrial

ecosystem. The model is composed of various inter-

connected components and agents, each representing

a speciﬁc function within the system. The primary

agents in the SCP model include facilities (such as re-

actors, storage units, and treatment plants), markets,

and suppliers. Each facility agent is characterized by

its input and output materials, storage capacities, op-

erational costs, and production scales (Figure 1). The

model simulates the dynamic interactions between

these agents, focusing on their decision-making pro-

cesses related to purchasing raw materials, producing

goods, and selling outputs in response to market con-

ditions.

Figure 1 illustrates the structural components and

interactions within the SCP model, emphasizing re-

source ﬂows between key agents such as facilities

Figure 1: Resource Flow and Decision-Making in the SCP

Model. The diagram illustrates the exchange of raw mate-

rials, intermediate products, and ﬁnal outputs among facili-

ties, markets, and suppliers, highlighting key interactions in

production, purchasing, and sales.

(e.g., methanol plants, electrolyzers), markets, and

suppliers. Arrows represent the movement of raw

materials (e.g., CO

, H

, coal), intermediate prod-

ucts (e.g., methanol, treated water), and ﬁnal out-

puts. The model integrates dynamic decision-making

for purchasing, production, and sales while ensuring

that storage levels remain within capacity limits. By

capturing resource exchanges alongside market inter-

actions, this representation underscores the model’s

capability to optimize resource allocation, production

efﬁciency, and sustainability in chemical production.

Remarks on Markets:

• Prices derive from time series (selling prices typi-

cally differ from purchase prices). Each product’s

price (e.g., for CO

) is dynamically derived from

time series data sourced directly from external

ﬁles, allowing for realistic variations over time.

• Demand is inﬂuenced by factors that are repre-

sented through time series data.

Remarks on Input Facility:

• Demand is decided from own production, use of

storage, and market purchases.

• Demand is inﬂuenced by factors represented in a

time series.

• Long-term constraints can necessitate purchases

(e.g., purchase contracts).

• Market prices are inﬂuenced by factors that can be

represented through time series data (which often

differ from selling prices).

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

148

Remarks on Output Facility:

• Production is limited by facility capacity.

• Selling price is calculated as Totex (Capex +

Opex), with Opex inﬂuenced by input prices.

• Long-term constraints may necessitate production

despite unfavorable market prices (e.g., supply

contracts).

• Market prices are inﬂuenced by factors that can be

represented through time series data (which often

differ from purchase prices).

In industrial ﬁnancial modeling, the selling price

of a product is often calculated based on Totex, which

consists of Capex (Capital Expenditure) and Opex

(Operational Expenditure). Capex represents the in-

vestment in long-term assets, such as infrastructure,

while Opex covers the ongoing costs of production,

including input prices. Opex is particularly sensitive

to ﬂuctuations in market prices for raw materials and

energy, making it a dynamic component. By combin-

ing Capex and Opex, Totex reﬂects the total cost of

ownership, guiding pricing strategies to ensure prof-

itability and sustainability. This comprehensive ap-

proach aids in ﬁnancial planning and resource man-

agement.

Facilities within the SCP model, such as methanol

plants and electrolyzers, play crucial roles in trans-

forming raw inputs into valuable chemical prod-

ucts. For example, the methanol facility converts car-

bon dioxide (CO

) and hydrogen (H

) into methanol

(CH

OH) and water (H

O), while the electrolyzer fa-

cility produces hydrogen from water using electric-

ity (Figure 2). These facilities operate under con-

straints such as production times, capacities, and eco-

nomic factors like operating expenses and market

prices. The model uses advanced decision-making al-

gorithms, including Pareto-based selection and Pareto

Optimization, to optimize the facilities’ operations,

ensuring efﬁcient resource use and maximizing eco-

nomic returns. This holistic approach allows the SCP

model to provide insights into the sustainability and

economic viability of chemical production processes,

making it a valuable tool for industrial resource man-

agement.

Figure 2 provides a detailed view of various facil-

ities within the SCP model, including methanol facil-

ities, electrolyzer, power plant, steel mill, and water

treatment. It showcases the input and output relation-

ships for these facilities, such as the coal market in-

put and power and CO

market outputs for the power

plant. This illustration highlights how the model is

generalized to other facilities like ethanol, urea, and

biomass, offering a comprehensive depiction of the

interconnected industrial ecosystem.

Figure 2: Real-world Representation of Facility Interac-

tions.

The SCP model is chosen for its comprehensive

approach to simulating and optimizing chemical pro-

duction and resource management. Its ability to in-

corporate various facilities such as methanol plants,

electrolyzers, power plants, and water treatment units

allows for a detailed analysis of the interactions be-

tween different production processes and resource

ﬂows. By simulating these interactions, the SCP

model helps identify inefﬁciencies and opportunities

for optimization, making it a powerful tool for en-

hancing sustainability in chemical production.

In the broader context of sustainable chemical

production, the SCP model addresses critical chal-

lenges such as resource utilization, energy efﬁciency,

and environmental impact. By integrating real-time

data and adaptive decision-making strategies, the

model provides insights into how different production

strategies affect overall sustainability. This holistic

perspective is essential for developing practices that

balance economic, environmental, and social objec-

tives, aligning with the goals of sustainable develop-

ment. The SCP model’s ability to simulate complex

industrial ecosystems and optimize resource manage-

ment makes it a valuable asset in the pursuit of sus-

tainable chemical production.

4 METHODOLOGY

4.1 Agent-Based Simulation Modeling

Agent-Based Modeling and Simulation (ABMS) is a

computational framework that enables the represen-

tation of autonomous agents, each with distinct at-

tributes and decision-making capabilities, to simulate

interactions within a system. In the context of sus-

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management

149

tainable chemical production, ABMS provides a de-

tailed and dynamic representation of industrial facil-

ities, markets, and resource ﬂows. This approach is

particularly valuable for capturing the complexity of

interconnected industrial ecosystems, where multiple

facilities, such as methanol plants, electrolyzers, and

power plants (Figure 3), interact through material and

energy exchanges. By simulating these interactions,

ABMS allows for an in-depth analysis of production

processes, resource allocation, and market dynamics.

This capability is essential for evaluating the impact

of different decision-making strategies on resource

utilization, energy efﬁciency, and overall environmen-

tal sustainability, making ABMS a powerful tool for

optimizing industrial operations in real-world scenar-

ios.

Figure 3: Geospatial Distribution of Facilities.

Figure 3 presents the geospatial distribution of

facilities introduced in Section 3, illustrating their

spatial relationships within the industrial ecosystem.

Each facility is represented by a colored circle, main-

taining consistency with the color scheme in Figure 2,

to depict its location. This spatial representation en-

ables the analysis of logistical constraints, transporta-

tion costs, and regional resource availability, which

inﬂuence decision-making within the simulation. The

geospatial component is implemented using MESA-

GEO’s GeoSpace() function in Python, allowing fa-

cilities to interact dynamically based on their loca-

tions. By incorporating spatial constraints, the model

ensures a more realistic representation of industrial

operations, considering factors such as material trans-

portation and facility proximity in resource allocation

strategies.

During each simulation run, market and facility

agents are activated in a partially randomized order,

consistent with standard agent-based modeling ap-

proaches. Market agents, such as the hydrogen mar-

ket, operate passively without initiating actions or

making independent decisions. Instead, they serve

as intermediaries that regulate economic interactions

within the simulation. Their primary functions in-

clude calculating new prices at each time step based

on historical time series data, generating statistical in-

sights at the market level rather than for individual fa-

cilities, and managing market offers, effectively act-

ing as a centralized commodity exchange. By main-

taining an updated list of market offers, these agents

provide a structured platform for facilities to engage

in transactions, ensuring a realistic representation of

industrial market dynamics.

Facility agents, such as electrolyzers, actively en-

gage in decision-making and execute key operational

processes. These agents determine their output prod-

uct offers, setting quantities and prices before listing

them on the market, while also incorporating spatial

attributes such as location to account for transporta-

tion costs and delivery times. Additionally, facility

agents procure necessary input materials by selecting

from market offers based on predeﬁned needs, prefer-

ences, and optimization strategies, including Pareto-

based selection. Their role extends beyond market

interactions, as they continuously initiate production

cycles, converting inputs into outputs at rates deﬁned

by industrial proﬁles contributed by project partners.

This dynamic decision-making structure allows the

simulation to capture the complexities of industrial

operations, resource management, and economic be-

havior within a multi-agent system.

The objective of the simulation is to analyze

system-wide developments when each facility agent

independently optimizes its own outcomes while be-

ing inﬂuenced by market conditions. Facility agents

operate autonomously, making decisions based on

production efﬁciency, cost minimization, and re-

source availability. Their interactions are shaped by

market forces, particularly through pricing and supply

ﬂuctuations, reﬂecting real-world industrial dynam-

ics. Additionally, facilities can establish long-term

agreements, fostering strategic partnerships that in-

ﬂuence resource allocation and production efﬁciency.

These interactions may lead to the emergence of pro-

duction clusters or supply chains that naturally de-

velop under given conditions. By identifying such

emergent patterns, the simulation provides valuable

insights into industrial self-organization, supporting

the design of more efﬁcient and resilient production

networks.

In the ABMS framework, reactors simulate the

transformation of raw materials and energy into prod-

ucts across industrial facilities. Their efﬁciency de-

pends on production capacity, technological con-

straints, and market conditions like resource availabil-

ity and pricing. This dynamic interaction enables re-

alistic modeling of industrial operations and resource

optimization. Table 1 outlines reactor conﬁgurations,

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

150

detailing inputs and outputs that drive the simulation.

Market agents play a crucial role in managing

price dynamics, regulating the availability of goods,

and facilitating transactions within the simulation.

The system distinguishes between input and out-

put prices, incorporating market ﬂuctuations into the

decision-making processes of facility agents. This in-

tegration ensures that pricing strategies reﬂect real-

world economic conditions, allowing agents to re-

spond dynamically to changing supply and demand.

By enabling strategic interactions and accounting for

price volatility, the market framework enhances the

adaptability and realism of the model, providing a

more accurate representation of industrial market be-

havior.

ABMS enables the integration of real-time data

and adaptive decision-making, which are essential

for optimizing chemical production processes. By

simulating various scenarios and their long-term ef-

fects, ABMS helps identify strategies that balance

economic viability with environmental and social ob-

jectives. Its ability to model complex systems, cap-

ture dynamic interactions, and provide comprehen-

sive analyses makes it a valuable tool for advancing

sustainable chemical production. This approach sup-

ports the development of optimized production strate-

gies that enhance resource efﬁciency while minimiz-

ing environmental impact, offering insights beyond

those achievable through traditional modeling meth-

ods.

4.2 Model Speciﬁcations

The Sustainable Chemical Production (SCP) model is

designed with speciﬁc agents, behaviors, interactions,

and environmental contexts to simulate and optimize

chemical production processes.

Agent Behavior. Each facility within the model acts

as an autonomous agent with unique properties and

decision-making processes. For example, a methanol

plant agent has attributes (Figure 2) such as input

materials (e.g., CO

and H

), output products (e.g.,

methanol and water), production capacity, and op-

erational costs as stated in Section 3. These agents

follow speciﬁc rules and algorithms to decide on the

purchase of raw materials, the quantity of production,

and the sale of ﬁnished goods.

Interactions. Interactions between agents are gov-

erned by market dynamics and resource ﬂows. Fa-

cilities interact with market agents to procure raw

materials and sell their products. For instance, the

power plant facility buys coal from the market and

sells power and CO

(Figures 1 and 2). These trans-

actions are inﬂuenced by market prices, availabil-

ity of resources, and contractual obligations. The

model also simulates internal interactions where out-

puts from one facility (e.g., hydrogen from the elec-

trolyzer) serve as inputs for another (e.g., methanol

production).

Environment. The model operates within a simu-

lated industrial ecosystem that includes various mar-

kets and environmental factors. Markets are modeled

to provide time-series data on prices and demand, en-

suring realistic economic conditions. Environmen-

tal constraints such as storage capacities, production

limits, and resource availability are deﬁned in the fa-

cilities’ conﬁguration ﬁles and actively enforced dur-

ing simulations to reﬂect real-world limitations. This

ensures agents operate within these constraints, dy-

namically adjusting their decisions based on factors

like available storage space and production capabili-

ties. Additionally, the model’s environment accounts

for long-term constraints such as purchase and sup-

ply contracts, which inﬂuence facility operations and

decision-making processes (Figure 3).

By incorporating detailed agent behaviors, com-

plex interactions, and realistic environmental con-

ditions, the SCP model provides a comprehensive

framework for analyzing and optimizing sustainable

chemical production processes. This enables the

identiﬁcation of strategies that enhance resource efﬁ-

ciency, reduce environmental impact, and ensure eco-

nomic viability.

4.3 Data Collection

The data collection process in this agent-based mod-

eling and simulation (ABMS) involves gathering and

analyzing key variables to understand the dynamics of

sustainable chemical production. The model utilizes

the MESA framework in Python, which allows for the

simulation of complex interactions between various

agents, including facilities and markets. Data collec-

tion is integrated within the simulation process, cap-

turing detailed information at each step of the model’s

execution.

4.3.1 Sources and Methods of Data Collection

Data are collected from several sources, including the

internal states of agents (such as production levels,

energy consumption, and storage capacities) and ex-

ternal market factors (like price ﬂuctuations and de-

mand). The simulation continuously gathers data on

these variables throughout the iterations, enabling a

comprehensive analysis of the system’s behavior over

time. Key metrics such as total energy consumed, to-

tal energy produced, and ﬁnancial transactions (pur-

chases and sales) are tracked using MESA’s built-in

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management

151

Table 1: Summary of Reactor Inputs and Outputs in Various Facilities.

Facility Input Output Reactor Type

Methanol Plant (Sollai et al., 2023) CO

, H

, Power CH

OH, H

O Chemical

Electrolyzer (El-Shaﬁe, 2023) H

O, Power H

, O

Electrolysis

Power Plant (Okunlola et al., 2023) Coal CO

, Power Combustion

Steel Mill (Singh et al., 2022) O

, CO

, Power Metallurgical

Water Treatment (Fadillah et al., 2024) Wastewater, Power, O

O Filtration

DataCollector() functionality. This data is then

stored for further analysis and visualization, allow-

ing for the assessment of performance indicators criti-

cal to sustainable chemical production. The collected

data provides insights into the efﬁciency of different

strategies and helps identify areas for optimization

within the system.

4.4 Algorithm

Following the detailed speciﬁcations of the agents in

Section 4.2, the core computational framework for

the simulation is described using an algorithm that

governs how facilities and markets interact, allocate

resources, and optimize production. The algorithm

serves as the backbone for managing agent behavior,

resource ﬂows, and decision-making within the simu-

lation environment.

The simulation model operates by ﬁrst initializing

all agents, facilities such as methanol plants, power

plants, and water treatment plants, as well as markets

for key commodities like carbon dioxide (CO

), water

O), hydrogen (H

), oxygen (O

), wastewater, and

coal. The agents operate based on their deﬁned roles,

interacting through the market system to purchase or

sell resources. The algorithm for resource optimiza-

tion in the simulation is broken down into the follow-

ing steps:

1. Initialize Agents and Markets. At the begin-

ning of each simulation run, facilities and markets

are initialized based on predeﬁned conﬁgurations

stored in external JSON ﬁles. Each facility in the

SCP model (Sections 3 and 4) is assigned oper-

ational parameters, including input requirements,

production capacity, available storage, and ﬁnan-

cial characteristics. Markets are dynamically in-

ﬂuenced by time series data sourced from exter-

nal CSV ﬁles and account for ﬂuctuating mate-

rial prices, providing realistic price variability for

agents’ decision-making.

2. Resource Availability Check. Facilities assess

their available resources stored onsite, such as

water, CO

, wastewater, and coal, and determine

their current production capabilities based on re-

source availability. This step ensures that the pro-

duction level does not exceed what is sustained by

the available inputs.

3. Market Interaction. Each facility engages with

the market to purchase the required resources if

the current storage is insufﬁcient. Facilities also

sell their output (e.g., methanol, power) to the re-

spective markets. Prices and availability in the

market inﬂuence these transactions, following a

Pareto optimization or resource-based selection

method.

4. Production and Optimization. Once the neces-

sary resources are secured, the facility initiates

production, with output levels adjusted accord-

ing to the “Resource Availability-Based Selec-

tion”, “Pareto-based Selection”, or “Pareto Op-

timization” strategies (outlined in Sections 5.1.1

to 5.1.3). For instance:

• In the Resource Availability-Based Selection,

production is directly tied to the amount of re-

sources available in storage.

• In the Pareto-based Selection and Pareto Op-

timization methods, production decisions are

based on balancing multiple factors like input

costs and resource availability to maximize ef-

ﬁciency.

5. Data Collection and Analysis. After each simu-

lation step, the system collects data (Section 4.3)

on key variables such as total energy consumed,

total energy produced, prices of purchased and

sold resources, and the amount of materials traded

in the market. These values are recorded for

later analysis to measure the performance of each

method over the simulation iterations.

6. Repeat Process. The simulation iterates through

multiple steps, where agents continuously assess

their resource needs, engage in the market, and

adjust production outputs, reﬂecting real-world

industrial processes.

The algorithm implements a multi-objective op-

timization strategy, with a particular focus on the

Pareto-based methods that balance trade-offs be-

tween cost minimization and production maximiza-

tion. This approach to structured decision-making

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

152

enables the simulation to identify optimal strategies

for resource allocation, supporting sustainable chem-

ical production objectives. By detailing this com-

putational framework, the algorithm systematically

models agent interactions and operations within the

SCP model, accurately simulating real-world indus-

trial processes while effectively integrating resource

constraints and market dynamics.

5 RESULTS AND DISCUSSION

This section presents the ﬁndings from simulating

three decision-making methods in a cross-industrial

network for sustainable chemical production. The

simulation, conducted over 100 iterations per method,

evaluates key metrics such as energy consumption,

production efﬁciency, price sales, and purchases

across various facilities, including methanol plants,

electrolyzers, power plants, steel mills, and water

treatment units. Data outputs were systematically

recorded in CSV ﬁles, capturing facility-market in-

teractions. The model incorporates a diverse range

of industrial facilities, differentiated by scale and ca-

pacity, to reﬂect real-world operations. It includes 40

methanol production plants (25 small, 10 medium,

and 5 large), 20 electrolyzers (15 small, 5 medium),

12 medium-sized power plants, 5 large steel mills,

and 10 medium water treatment facilities. These con-

ﬁgurations, informed by industrial data, contribute to

a realistic representation of an interconnected indus-

trial ecosystem. Figure 3 visualizes the spatial distri-

bution of these facilities within the simulation envi-

ronment.

5.1 Simulation Outputs

5.1.1 Resource Availability-Based Selection

The Resource Availability-Based Selection method

optimizes facility operations by adjusting production

output to match the availability of resources in stor-

age. By ensuring that the facility’s inputs are aligned

with current storage levels, this method reduces the

risk of resource overuse or shortages, promoting ef-

ﬁcient and stable production. It focuses on balanc-

ing input availability with production demands, mak-

ing it well-suited for environments where maintaining

continuous operation without disruptions is a prior-

ity. This approach operates independently of external

market factors, such as ﬂuctuations in input prices.

In the method, the production output of a facil-

ity is determined by aligning available resource levels

with the required reactor inputs for production. The

production process is generalized as follows:

out put

= min(

input1

input2

, ...)

Where:

• P

out put

represents the maximum production out-

put, such as methanol, power, or other products

depending on the facility.

• R

input1

, R

input2

, ... denote the available quantities

of required resources (e.g., carbon dioxide, hydro-

gen, water, or coal) in the facility’s storage.

• r

input1

, r

input2

, ... represent the input requirements

for each resource per unit of production.

For example, a methanol facility requires reactor in-

puts like carbon dioxide, hydrogen, and power, while

a power plant primarily relies on coal. This method

identiﬁes the bottleneck resource for any facility and

adjusts the production output accordingly, ensuring

efﬁcient use of available resources without exceeding

operational constraints.

During the steps, this method demonstrated a

steady increase in both energy consumption and pro-

duction, though its overall efﬁciency plateaued when

compared to more advanced optimization techniques.

Energy consumption grew from 5, 299 megawatt-

hours (MWh) at the start to over 508, 000 MWh by

the ﬁnal iteration, while energy production reached

254, 485 MWh (Table 2). Additionally, the total

amount of carbon dioxide purchased consistently ex-

ceeded the amount sold, with the gap widening as the

number of iterations increased (Figure 4).

Figure 4: Comparison of Total CO

Purchase and Sell in

Three Methods.

In conclusion to this section, clarifying the abbre-

viations and measurement units in Tables 2, 3, and 4

is essential to maintain consistency and facilitate clear

interpretation of the data presented. “Total Energy

Consumed” refers to the energy used by the facili-

ties, measured in megawatt-hours (MWh), while “To-

tal Energy Produced” indicates the energy generated

by the facilities, also measured in MWh. The tables

also report water transactions, with “Total H

O Pur-

chase” and “Total H

O Sell” reﬂecting the quantities

of water purchased and sold, measured in liters (L).

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management

153

Table 2: Summary Statistics of Energy, Prices, and Material Transactions (Resource Availability-Based Selection method).

Field Min Max Mean Std Dev

Total Energy Consumed (MWh) 5,299.0 508,280.6 265,912.4 144,395.2

Total Energy Produced (MWh) 2,231.8 254,485.6 127,184.9 73,861.7

Total H

O Purchase (L) 22,473.0 2,183,441.9 1,172,773.8 614,694.2

Total H

O Sell (L) 13,369.2 2,160,537.2 1,088,608.8 625,706.8

Total H

Purchase (m

) 2,696.9 1,388,470.3 687,796.8 412,387.1

Total H

Sell (m

) 7,737.3 2,161,159.4 1,074,404.4 641,651.2

Total O

Purchase (m

) 8,452.8 3,977,244.7 1,891,612.6 1,185,319.0

Total O

Sell (m

) 13,242.7 4,064,658.7 1,980,978.3 1,193,471.9

Total CO

Purchase (m

) 49,811.8 8,642,920.5 4,323,469.0 2,521,670.9

Total CO

Sell (m

) 34,190.0 5,923,805.7 2,977,017.8 1,731,421.6

Total CH

OH Sell (m

) 7,838.0 4,204,812.0 2,029,045.8 1,235,886.6

Total Coal Purchase (kg) 5,655.8 802,230.4 403,925.5 231,973.2

Total Wastewater Purchase (L) 10,494.0 1,883,808.0 956,280.2 543,677.6

Total Price Sales (Euro) 632,019.8 207,936,053.7 103,164,904.6 60,782,915.7

Total Price Purchases (Euro) 2,608,095.5 376,430,049.9 189,182,797.3 109,385,501.7

For gases like hydrogen, oxygen, and carbon diox-

ide, the terms “Total H

Purchase/Sell”, “Total O

Purchase/Sell”, and “Total CO

Purchase/Sell” de-

note the quantities of these gases traded in cubic me-

ters (m

). Similarly, “Total CH

OH Sell” represents

methanol sold in cubic meters (m

), “Total Wastewa-

ter Purchase” speciﬁes the volume of wastewater ac-

quired, measured in liters (L), while “Total Coal Pur-

chase” speciﬁes the amount of coal acquired, mea-

sured in kilograms (kg). “Total Price Sales” repre-

sents the total revenue generated from sales in euros,

and “Total Price Purchases” captures the total expen-

diture on purchasing resources, also in euros. These

clariﬁcations ensure the correct interpretation of the

numerical data in the corresponding tables. To ensure

clarity in interpreting the table data, the ﬁrst line of

each table includes the columns “Field”, “Min” (mini-

mum values), “Max” (maximum values), “Mean” (av-

erage values), and “Std Dev” (standard deviation).

This structure provides a comprehensive view of the

data’s range and variability, supporting detailed anal-

ysis of each ﬁeld’s statistical distribution.

5.1.2 Pareto-Based Selection

In the Pareto-based Selection method, production op-

timization is achieved by considering both resource

availability and market prices. Mathematically, the

production amount P is determined by the available

amounts of each resource R

, their respective input re-

quirements I

, and the market price M

for each re-

source. The potential production limits L

for each

resource i are calculated as:

Next, the bottleneck resource is identiﬁed as the one

with the smallest production limit, which sets the

maximum feasible production amount:

max

= min(L

This maximum production amount is further adjusted

by considering the inﬂuence of market prices. The

ﬁnal production amount P

ad justed

is calculated by tak-

ing the minimum of the production amount and the

price-adjusted resource availability:

ad justed

= min(

Here, R

represents the available resource quantity, I

the input requirement for each unit of production,

and M

the market price for the resource. This opti-

mization ensures that production is both cost-efﬁcient

and resource-efﬁcient, balancing the availability of re-

sources and economic factors.

This approach led to a marked increase in to-

tal energy consumption and production, with energy

output in later stages reaching up to 6, 640 MWh.

Additionally, total sales and purchases were signiﬁ-

cantly higher, reﬂecting heightened economic activ-

ity within the system. By integrating both market dy-

namics and resource constraints, this strategy enabled

more effective resource utilization. Notably, the sim-

ulation showed that energy production was approx-

imately one-quarter of energy consumption, with the

total energy consumed reaching 21, 329 MWh and en-

ergy produced at 6, 640 MWh by the ﬁnal iteration

(Table 3). Furthermore, the quantities bought and

sold for key commodities, such as carbon dioxide, wa-

ter, hydrogen, and oxygen, remained nearly balanced

(Figure 5), indicating stable trading patterns across

multiple iterations. For instance, the total amounts

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

154

of carbon dioxide purchased and sold were closely

aligned, a trend also observe for other commodities,

with minimal variation over repeated cycles.

Figure 5: Commodity Transactions (Water, Hydrogen, Oxy-

gen, and CO

) Across Iterations for Pareto-based Selection.

5.1.3 Pareto Optimization

The Pareto Optimization method employs multi-

objective optimization to balance resource procure-

ment, production, and sales while ensuring Pareto ef-

ﬁciency where no resource is improved at the expense

of another. This approach minimizes procurement

costs, maximizes production efﬁciency, and optimizes

proﬁts while accounting for constraints such as stor-

age capacity, market prices, demand ﬂuctuations, and

resource availability. The production amount P is de-

termined by balancing available resources R

, input

requirements I

and market prices M

. Initially, the

available resources are adjusted for market price dif-

ferences, yielding the price-weighted resource avail-

ability:

′

Next, the ratios ρ

of available resources to their input

requirements are calculated as:

′

The bottleneck resource, which has the smallest ratio,

limits the production amount. Thus, the maximum

production amount P

max

is given by the resource with

the smallest ratio:

max

= min(ρ

This method ensures a Pareto-efﬁcient production,

meaning no single objective (e.g., minimizing cost

or maximizing output) is improved without worsen-

ing another. By factoring in both resource availability

and price adjustments, the method achieves optimal

resource utilization and economic performance.

The algorithm ﬁrst calculates resource availability

relative to market prices, giving priority to more af-

fordable resources. It then determines the maximum

feasible production amount by identifying bottleneck

resources, which are the limiting factors in produc-

tion. The Pareto optimization balances resource al-

location based on these constraints, ensuring that the

selected production plan optimizes cost-effectiveness

and production efﬁciency. This approach is more

sophisticated than basic selection methods as it in-

tegrates market dynamics and capacity limits, ulti-

mately leading to an efﬁcient allocation of resources

across the facility network.

This approach achieved the highest total price

sales and the lowest price purchases, optimizing the

balance between energy consumption and production.

For example, in the ﬁnal stages of the simulation,

total energy consumption surpassed 10, 842 MWh,

while total energy production reached 5, 565 MWh

(Table 4). Notably, this optimization technique led to

a considerable reduction in raw material purchases,

which positively impacted the operational cost and

resource utilization across the network. Similar to

the Pareto-based Selection method, the purchase and

sell in this approach follow a balanced and system-

atic resource allocation process (Figure 6). However,

in terms of energy consumption and production, the

Pareto Optimization method demonstrates a superior

efﬁciency, achieving more optimal energy usage than

other methods (Figure 7).

Figure 6: Commodity Transactions (Water, Hydrogen, Oxy-

gen, and CO

) Across Iterations for Pareto Optimization.

Figure 7: Comparison of Energy Purchase and Sell in

Pareto-based Selection and Pareto Optimization Methods.

Each method displayed distinct trade-offs be-

tween energy efﬁciency, production capacity, and

economic performance, making the Pareto Optimiza-

tion approach the most effective strategy for sus-

tainable production. The results indicate that ad-

vanced decision-making methods can signiﬁcantly re-

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management

155

Table 3: Summary Statistics of Energy, Prices, and Material Transactions (Pareto-based Selection).

Field Min Max Mean Std Dev

Total Energy Consumed (MWh) 11.0 21,329.4 9,078.5 6,612.2

Total Energy Produced (MWh) 4.0 6,640.6 2,829.5 2,055.5

Total H

O Purchase (L) 984.9 2,232,767.8 915,167.9 690,953.1

Total H

O Sell (L) 1,012.1 2,277,045.6 933,788.2 704,521.6

Total H

Purchase (m

) 132.2 807,743.9 312,888.3 250,393.8

Total H

Sell (m

) 268.7 1,149,619.6 430,642.3 356,096.9

Total O

Purchase (m

) 397.0 1,879,904.0 740,961.1 582,796.3

Total O

Sell (m

) 516.3 2,193,816.2 824,005.5 680,273.7

Total CO

Purchase (m

) 2,455.5 4,725,099.6 1,983,431.5 1,463,308.7

Total CO

Sell (m

) 2,407.7 4,663,713.8 1,955,090.6 1,444,190.4

Total CH

OH Sell (m

) 40.9 2,095,473.4 699,146.8 648,522.5

Total Coal Purchase (kg) 267.0 475,919.5 202,648.7 147,454.9

Total Wastewater Purchase (L) 673.2 1,197,079.5 509,335.3 370,882.9

Total Price Sales (Euro) 16,518.9 138,659,035.2 48,956,108.98 42,909,722.9

Total Price Purchases (Euro) 122,871.4 223,658,461.6 95,048,939.4 69,299,005.8

Table 4: Summary Statistics of Energy, Prices, and Material Transactions (Pareto Optimization).

Field Min Max Mean Std Dev

Total Energy Consumed (MWh) 0.0 10,842.7 4,001.4 3,511.0

Total Energy Produced (MWh) 0.0 5,565.6 2,171.4 1,736.1

Total H

O Purchase (L) 846.0 1,888,945.6 801,317.3 594,864.5

Total H

O Sell (L) 882.0 2,042,219.5 861,946.5 639,048.3

Total H

Purchase (m

) 130.3 797,192.7 302,807.7 245,726.9

Total H

Sell (m

) 260.6 1,128,242.0 409,677.3 347,200.5

Total O

Purchase (m

) 343.0 2,194,182.1 807,061.6 678,033.65

Total O

Sell (m

) 446.0 2,337,957.8 823,313.3 721,073.3

Total CO

Purchase (m

) 2,418.9 4,716,858.8 1,979,621.4 1,459,026.9

Total CO

Sell (m

) 2,418.9 4,661,324.5 1,954,031.7 1,441,608.1

Total CH

OH Sell (m

) 0.0 2,078,860.0 688,369.8 644,578.1

Total Coal Purchase (kg) 252.7 474,804.7 202,517.3 147,283.7

Total Wastewater Purchase (L) 540.0 1,092,150.0 474,468.3 342,149.0

Total Price Sales (Euro) 12,347.7 131,024,730.2 45,056,517.7 40,638,311.4

Total Price Purchases (Euro) 114,409.6 221,473,197.2 94,073,347.6 68,913,709.8

duce costs and improve resource utilization in the in-

dustrial networks.

5.2 Analysis

The Resource Availability-Based Selection method

shows a steady rise in energy consumption and pro-

duction over the iterations (Section 5.1.1). However,

its efﬁciency plateaus when compared to more ad-

vanced methods. This behavior is attributed to the

method’s dependence on available resources in stor-

age, which, although ensuring that the facility does

not exceed its input constraints, limits optimization

potential. The method lacks ﬂexibility in handling

dynamic market prices or external factors, making it

less responsive to changes in resource costs. As a re-

sult, energy consumption increases to 508,000 MWh

by the ﬁnal step, while energy production peaks at

254,485 MWh. Additionally, discrepancies in mate-

rial transactions, particularly in CO

purchases and

sales, highlight the limitations of this approach in bal-

ancing resource inﬂows and outﬂows effectively.

The Pareto-based Selection method, designed to

optimize production by considering both resource

availability and market prices, demonstrates signiﬁ-

cant improvements in efﬁciency (Section 5.1.2). This

approach achieves better resource allocation by mini-

mizing costs and maximizing production outputs. For

instance, by factoring in market dynamics, the method

maintains a closer balance between energy consump-

tion and production, with the ﬁnal energy consump-

tion exceeding 21,000 MWh and production reaching

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

156

6,640 MWh. Material transactions, such as water, hy-

drogen, oxygen, and carbon dioxide, show more sta-

bility in terms of purchase and sell quantities, lead-

ing to more balanced operations. The results conﬁrm

that the Pareto-based approach outperforms simpler

resource allocation strategies by effectively leverag-

ing market conditions alongside resource availability.

The Pareto Optimization method further reﬁnes

the multi-objective approach by focusing on maxi-

mizing production efﬁciency while minimizing op-

erational costs (Section 5.1.3). This method consis-

tently produces the most optimal results, with en-

ergy consumption reaching over 10,800 MWh and en-

ergy production nearing 5,565 MWh by the ﬁnal step.

Notably, the method achieves a signiﬁcant reduction

in raw material purchases compared to the previous

approaches, as it strategically prioritizes cheaper re-

sources and ensures that each production decision is

Pareto-optimal. This leads to overall improvements

in network efﬁciency, with the total purchase and

sell amounts for various commodities closely aligned.

The superior performance of Pareto Optimization, in

comparison to the other two methods, demonstrates

its capability in managing resource constraints and

market ﬂuctuations effectively.

In conclusion, while all three methods have their

merits, Pareto-based Selection and Pareto Optimiza-

tion clearly outperform the Resource Availability-

Based approach, particularly in scenarios involving

dynamic markets. Pareto Optimization, in particular,

achieves the most efﬁcient balance between energy

consumption, production, and material transactions,

making it the most effective method for large-scale

industrial applications.

5.3 Implications

The simulation framework developed in this study has

signiﬁcant implications for both sustainable chemical

production and broader industrial applications. By

employing the Resource Availability-Based Selection

method, facilities can align production with available

input resources, minimizing energy consumption and

resource waste while improving overall efﬁciency.

This approach reduces environmental impact, particu-

larly in resource-intensive industries, by ensuring op-

erations remain within input constraints. Similarly,

the Pareto-based Optimization strategy enables facil-

ities to balance multiple objectives, such as minimiz-

ing raw material purchases and maximizing produc-

tion, ultimately lowering operational costs and emis-

sions. These strategies enhance economic viability

while promoting sustainability and regulatory compli-

ance. Beyond chemical production, the framework’s

adaptability extends to industries such as logistics,

energy, and supply chain management. In logistics,

resource-aligned decision-making can optimize trans-

portation routes and inventory control, reducing costs

and inefﬁciencies. In the energy sector, Pareto-based

optimization can balance energy generation and con-

sumption based on resource availability, fuel costs,

and environmental factors, providing a valuable tool

for managing renewable energy ﬂuctuations. Addi-

tionally, in global supply chains, where resource costs

and availability vary regionally, the framework opti-

mizes inventory levels and production schedules, en-

suring efﬁcient and sustainable operations. These

diverse applications highlight the model’s ﬂexibility

in improving resource management and sustainability

across multiple industries.

5.4 Future Research

While this study provides valuable insights into re-

source optimization and energy efﬁciency in sustain-

able chemical production, several limitations high-

light areas for future research. The model simpli-

ﬁes market conditions, assuming stable input prices

and supply chains, which may not reﬂect real-world

volatility. Additionally, it focuses on a limited set of

materials, excluding catalysts, secondary emissions,

and waste management complexities. Spatial assump-

tions, though accounting for geographic distribution,

overlook transportation logistics and regional price

variations critical for large-scale operations. Further-

more, the model does not distinguish between renew-

able and non-renewable energy sources or consider

energy transmission losses, limiting its sustainability

assessment. Economic and regulatory factors, such

as taxes and subsidies, are also omitted, restricting its

policy applicability. Future research should integrate

dynamic market conditions, reﬁne spatial logistics,

incorporate renewable energy considerations, and ex-

pand material scope to enhance the model’s realism

and applicability in industrial sustainability.

6 CONCLUSIONS

This study explored the optimization of resource allo-

cation and energy efﬁciency within sustainable chem-

ical production networks using three distinct meth-

ods: Resource Availability-Based Selection, Pareto-

based Selection, and Pareto Optimization. Each

method demonstrated unique strengths in manag-

ing resource inputs and energy consumption. The

Resource Availability-Based Selection method fo-

cused on aligning production output with available

Agent-Based Simulation Modeling for Sustainable Chemical Production and Resource Management

157

resources in storage, resulting in a steady increase in

energy consumption and production over time. How-

ever, it exhibited limitations in maximizing efﬁciency

compared to more advanced methods. The Pareto-

based Selection method balanced input prices with re-

source availability, leading to more efﬁcient produc-

tion outcomes and higher economic activity. Lastly,

the Pareto Optimization approach, as the most ad-

vanced method, consistently minimized operational

costs while maximizing production efﬁciency, yield-

ing the highest total price sales and demonstrating op-

timal energy use.

The ﬁndings highlight the signiﬁcance of multi-

objective optimization in improving resource man-

agement and production processes in chemical indus-

tries. By leveraging Pareto Optimization, companies

can achieve a more balanced approach to energy use

and resource allocation, ultimately enhancing both

economic and environmental sustainability. However,

the study also acknowledges that real-world complex-

ities, such as market ﬂuctuations and logistical con-

straints indicating areas for further improvement.

Based on the results, it is recommended that

chemical production facilities adopt multi-objective

optimization techniques like Pareto Optimization to

enhance operational efﬁciency and sustainability. Fu-

ture research should focus on incorporating dynamic

market conditions, transportation logistics, and re-

newable energy sources into optimization models.

Additionally, considering regulatory and economic

factors, such as carbon pricing and subsidies, will of-

fer a more comprehensive view of sustainability in

chemical production. By adopting these improve-

ments, industries can better align with global sustain-

ability goals while maintaining economic viability.

ACKNOWLEDGEMENTS

We would like to thank all our partners of the project

Carbon2Chem

for the pleasant and successful in-

terdisciplinary collaboration. Also, we would like to

thank the Federal Ministry of Education and Research

(BMBF) for funding the project Carbon2Chem

(project number 03EK3037D).

REFERENCES

Aghaie, A. and Hajian Heidary, M. (2019). Simulation-

based optimization of a stochastic supply chain con-

sidering supplier disruption: Agent-based model-

ing and reinforcement learning. Scientia Iranica,

26(6):3780–3795.

El-Shaﬁe, M. (2023). Hydrogen production by water elec-

trolysis technologies: A review. Results in Engineer-

ing, page 101426.

Fadillah, G., Alariﬁ, N. T., Suryawan, I. W. K., and Saleh,

T. A. (2024). Advances in designed reactors for water

treatment process: A review highlighting the designs

and performance. Journal of Water Process Engineer-

ing, 63:105417.

Haleem, A., Javaid, M., Singh, R. P., Suman, R., and Qadri,

M. A. (2023). A pervasive study on green manufactur-

ing towards attaining sustainability. Green Technolo-

gies and Sustainability, 1(2):100018.

Helo, P. and Rouzafzoon, J. (2023). An agent-based sim-

ulation and logistics optimization model for manag-

ing uncertain demand in forest supply chains. Supply

Chain Analytics, 4:100042.

Ionescu, S¸., Delcea, C., Chirit¸

a, N., and Nica, I. (2024). Ex-

ploring the use of artiﬁcial intelligence in agent-based

modeling applications: A bibliometric study. Algo-

rithms, 17(1):21.

Long, Q. (2014). Distributed supply chain network mod-

elling and simulation: integration of agent-based dis-

tributed simulation and improved scor model. Inter-

national journal of production research, 52(23):6899–

6917.

Mishra, K., Siwal, S. S., Nayaka, S. C., Guan, Z., and

Thakur, V. K. (2023). Waste-to-chemicals: Green so-

lutions for bioeconomy markets. Science of The Total

Environment, 887:164006.

Okunlola, A., Davis, M., and Kumar, A. (2023). Assess-

ing the cost competitiveness of electrolytic hydrogen

production from small modular nuclear reactor-based

power plants: A price-following perspective. Applied

Energy, 346:121290.

Singh, V., Buelens, L. C., Poelman, H., Saeys, M., Marin,

G. B., and Galvita, V. V. (2022). Carbon monoxide

production using a steel mill gas in a combined chem-

ical looping process. Journal of Energy Chemistry,

68:811–825.

Sollai, S., Porcu, A., Tola, V., Ferrara, F., and Pettinau, A.

(2023). Renewable methanol production from green

hydrogen and captured co2: A techno-economic as-

sessment. Journal of CO2 Utilization, 68:102345.

Venkataramanan, V. S. and Srinivasan, R. (2024). Agent-

based dynamic simulation for supply chain manage-

ment of lng import terminals. Industrial & Engineer-

ing Chemistry Research, 63(6):2750–2768.

Zhou, L., Chen, G., Zheng, M., Gao, X., Luo, C., and

Rao, X. (2024). Agent-based modeling methodology

and temporal simulation for natech events in chemi-

cal clusters. Reliability Engineering & System Safety,

243:109888.

Zhu, M., Yang, G., Jiang, Y., and Wang, X. (2023). Agent-

based modeling for water–energy–food nexus and its

application in ningdong energy and chemical base.

Sustainability, 15(14):11428.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

158