Next‑Generation Smart Grid Optimization: Integrating Edge AI for

Real‑Time and Decentralized Energy Management

Surya Narayan Sahu

, K. Ruth Isabels

, R. Gayathiri

, A. Nagamani

, V. Sriga

and Elumalai P.

Department of EEE, Centurion University of Technology and Management, Odisha, India

Department of Mathematics, Saveetha Engineering College (Autonomous), Thandalam, Chennai, Tamil Nadu, India

Department of Electrical and Electronics Engineering, J.J. College of Engineering and Technology, Tiruchirappalli, Tamil

Nadu, India

Department of Computer Science and Engineering MLR Institute of Technology, Hyderabad, Telangana, India

Department of Management Studies, Nandha Engineering College, Vaikkalmedu, Erode, Tamil Nadu, India

Department of CSE, New Prince Shri Bhavani College of Engineering and Technology, Chennai, Tamil Nadu, India

Keywords: Edge AI, Smart Grid, Energy Optimization, Federated Learning, Real‑Time Management.

Abstract: Nowadays, the development of smart grid systems require intelligent and real-time solutions in scalable

fashion. This paper presents a new approach that combines edge computing and AI for decentralized and

adaptable energy management in the context of distributed grid settings. Contrary to the typical cloud-based

models, the proposed model utilizes edge AI enabling local data processing, shortened latency and quick

decision-making. The approach is built based on federated learning and lightweight deep learning models, as

well as considerations on privacy preserving and grid resilience against dynamic load demands and system

failures. The performance of the system is additionally evaluated through simulation and benchmarked against

centralized approaches, with results showing that the proposed framework achieves higher efficiency,

scalability, and reliability in resource-limited edge environments. This research adds to the cornerstone of

future smart grids that can accommodate sustainable and self-sufficient energy ecosystems.

1 INTRODUCTION

Modern energy systems are becoming

increasingly complex due to rising demand, the

integration of renewable energy sources, and a shift

toward decentralized energy production, which has

forced traditional centralized power grids to continue

their transition to intelligent, adaptive state-of-the-art

power grids (i.e., smart grids). Such systems must not

only have efficient sharing mechanisms among

peers, but must also be responsive in real-time and

handle data securely across distributed nodes.

Traditional cloud-centric techniques, although

efficient, may suffer from high latency, bandwidth

limitation and privacy issues, particularly in the

context of the geographically distributed and

resource constrained scenarios.

In this regard, edge computing arises as a

disruptive paradigm that makes possible on-site

processing and decision-making near data generation

points. When complementing with AI, specifically

light weight models for edge, smart grids will have

the capability to forecast demand, find anomalies and

optimize energy transfer with minimal time

difference. This unification of Edge AI allows grid

nodes to be self-governing and responsive to

changing conditions, while sharing just enough

information to efficiently utilize energy, and

minimizing dependence on a central network

resource for all information.

But achieving this vision faces a number of

challenges, including keeping the models accurate

under diverse situations, secure in distributed

learning settings, and resource-efficient on edge

devices. In this work, we propose a holistic

framework that integrates Edge AI and federated

learning for smart grid operation improvement. With

the ability to empower distributed intelligence, it is no

longer necessary to transfer real-time decision-

making for control or optimisation out of a local

Fremework area or substation, which will result in

reduced latency and enhanced performance, and keep

316

Sahu, S. N., Isabels, K. R., Gayathiri, R., Nagamani, A., Sriga, V. and P., E.

Next-Generation Smart Grid Optimization: Integrating Edge AI for Real-Time and Decentralized Energy Management.

DOI: 10.5220/0013863300004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 1, pages

316-322

ISBN: 978-989-758-777-1

user data secure, creating a foundation for scalable

and sustainable smart energy systems of the future.

1.1 Problem Statement

The growing presence of renewable energy,

deployment of distributed energy resources (DERs)

and higher expectations for energy efficiency are

dramatically changing the operational requirements

of power grids. Next generation smart grids should

act as intelligent, self-adaptive infrastructures

dealing with decentralized energy production,

dynamic fluctuating consumption, and sophisticated

energy trading situations. Nevertheless, the current

grid architectures depend heavily on centralized

cloudbased infrastructure for data processing,

analytics, and decision-making. Although these

architectures provide computational capabilities, they

are inherently restricted due to high communication

latency, possible communication faults, congestion,

as well as higher risks on the data privacy and

security. These limitations are even more severe in

the case of emergency situations that require

immediate response and local control to ensure the

stability of the grid, and to avoid blackouts and

brownouts.

In addition, most AIenabled smart grid

optimization solutions that are deployed today are

computationally heavy and developed to be

implemented in cloud or centralized data centers such

that cannot be easily deployed at the edge with limited

processing power and memory. Inability to perform

local data processing and analysis slows down

decision making and limits the grids ability to

respond quickly to load or generation changes and

system anomalies. Moreover, centralized models of

data aggregation also present significant challenges

in terms of consumer privacy, data ownership and

system security especially when more and more users

join the demand response or peer-to-peer energy

trading schemes.

These limitations bring to surface a major gap in

the existing smart grid technology, i.e., the absence

of a decentralized, lightweight and privacy-

preserving mechanism that can leverage AI in situ to

the edge of the network. This ‘brain’ must be forged

if our visions of the completely flexible, smart, and

self-healing grid are to be realized. To do so, it is

necessary to provide new models combining Edge AI

and federated learning, which can achieve

decentralized decision, increase system robustness

and further secure and activate energy exchange in

real time. Closing this fall between the electronic and

physical world is crucial for the future of a smarter,

not to mention more sustainable, scalable, user-

centric, energy system.

2 LITERATURE SURVEY

The intersection of artificial intelligence (AI) and

edge computing has led to a new era for the

operations of smart grid that has transcended the real-

time and distributed grid control. With the

development of the smart grid to a more active, more

information-intensive network, many papers are

proposed for intelligent methods to optimize the

energy allocation. A seminal review is reported by

Biswal et al. (2025), who provide an extensive

overview of AI and edge technologies in power

systems in general, observing the growing

significance of distributed intelligence for grid

efficiency. But as inclusive as they are, they do not

delve into real-time applications.

Exploiting emerging networking models, Islam

et al. (2022), who suggest AI-supported architecture

over 6G networks for intelligent energy control, but it

cannot be used in the short term due to requirement

of the future infrastructure. Similarly, Arcas et al.

(2024) propose an edge offloading scheme in latency-

constrained controlnetworks, and it is a promising

approach, but still needs to be verified in practice for

practicality. Nandhakumar et al. (2023) introduce a

toolset for edge intelligence on energy applications,

with a focus on modularity and detailed

benchmarking in real settings.

Commercial voices including Shinde (2025) and

Habib (2025) clarify that the presence of edge-based

intelligence is increasingly evident, but their work has

remained a conceptual exercise without empirical

depth. On the other hand, a very limited review about

the smart grid digitization and lightweight

intelligence is presented in Biswal, Balamurugan,

and Sahoo (2024) but no new method is proposed.

A more technical analysis from Dileep (2021) that

reinforces the use of AI in predictive modeling and

demand forecasting. Yet it completely ignores the

latency and bandwidth limitations handled by edge

computing. Ullah and Khan (2022) consider a wider

angle, investigating edge computing issues in smart

grids, but they need more real applications and

configurations to verify their proposed models.

Applications of deep learning are developed by

Li et al. (2023) that propose the use of neural

networks for energy management, but do not consider

computational constraints common to edge. Yang et

al. (2022) continue this thread by surveying several

AI algorithms for smart grids, but only give high-

Next-Generation Smart Grid Optimization: Integrating Edge AI for Real-Time and Decentralized Energy Management

317

level comparisons between them without

recommendations for deploying them.

Regarding energy trading, Zhang et al. (2021)

use blockchain for peer to peer transactions, towards

decentralization, but from a theoritical point of view.

Similarly, Zhan et al. (2023) proposes the use of

federated learning for demand response, this solution

also with a privacy protection as it deems the

convergence of the model and consistency at the

nodes as issues.

Wang et al. (2021) presents an architecture that

integrates edge and AI for rapid grid response, though

real-time response in different situations is not fully

tested. Ahmed and Rehman (2022) address the short-

term load forecasting with AI for a voltage control

platform; although this is a crucial task, it has been

de-coupled from control plans needed in dynamically

changing grid situations.

Ghosh et al. (2023) propose a blockchain-AI

hybrid architecture at the edge of the grid, which

increases trust and autonomy, whereas Kumar and

Tripathi (2021) investigate reinforcement learning for

control optimization, however in simplified

simulation environment. Yu et al. (2024) use graph

neural networks for energy management, with high-

performance albeit demanding computational

resources that question the edge feasibility.

Continuing with real-time energy control theme,

Sun et al. (2021) use AI for dynamic decision-

making but test it with fixed sets of data. Tan and

Ramachandran (2023) use deep learning to detect

faults at substations, the proposed model achieves

high performance on a narrow domain and cannot be

well-extended to broader energy management

strategies, such as connectivity optimization.

Zhao et al. (2022), propose that intelligent

scheduling of edge resources can lead to efficiency

gains, however it needs a stronger practical

integration. the authors in Singh and Gupta (2024)

return to federated learning to maintain privacy in the

distributed environments, yet they merely loosen the

requirements of network latency and communication

overheads.

Luo et al. (2025) they propose multi-agent

system that supports collaborative control, which

brings forward the frontier of the decentralized grid

intelligence. However, the framework would benefit

from more treatment of fault tolerance and scalability

under load. Chatterjee et al. (2021), who concentrated

on AI-based anomaly detection that is essential to

grid resilience, but whose experiment results on

synthetic data are not fully trustworthy. Lastly, Liu et

al. (2023) analyze the optimization between the cloud

and the edge resources and provide solutions in terms

of workload assignment while not taking into

consideration the dynamism of the operating

conditions.

Collectively, they present a collection of work that

demonstrates the transformative nature of AI and

Edge computing toward smart grids. However, they

expose significant practical challenges in

deployment, real-time optimization and privacy-

preserving distributed learning. This highlights the

importance of developing an integrated Edge AI,

federated learning, and lightweight intelligence

framework for scalable, secure, and adaptive smart

grids.

3 METHODOLOGY

In this paper, we propose and evaluate a decentralized

Edge AI based energy management framework for

SoGs, utilizing hybrid design and simulation

procedure. The approach is based on the combination

of federated learning and lightweight artificial

intelligence models with fast response time, intended

to be deployed in edge devices in the smart grid. The

architecture of the proposed system enables

processing of real-time data, local decision-making,

and adaptive control of energy resources without

relying on centralized cloud servers.

The framework is open into three interrelated

architecture tiers: data acquisition layer, edge-based

intelligence layer and federated coordination layer. At

the bottom, smart meters and IoT sensors spread on

the grid are gathering metered and real-time energy

production/consumption, load variations and all sorts

of sensed environmental features. These streams are

processed in real-time at substations and energy

nodes that include low-power embedded systems.

These edge units come with pre-trained deep learning

models, such as CNNs and LSTM forms, which have

been pruned and quantized to enable inference under

limited computational budget. Table 1 shows the

evaluation metrics used for model assessment.

Instead of pushing raw data to a server, each edge

device trains its model locally on-the-fly, capturing

localized patterns in energy dynamics. To maintain

privacy and scalability, federated learning is used to

periodically average the updates of learned-model

(not raw data) of distributed nodes into a global

model. This coordination is overseen by a light-

weight orchestration algorithm that dynamically

aggregates participating nodes dependent on their

quality of data, availability of network, and the

remaining energy. The federated model is

subsequently dispatched to edge devices, this

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

318

continuous loop learning helps dynamically adjusting

changes in grid environment such as user privacy and

communication overhead.

Table 1: Evaluation metrics used for model assessment.

Metric Description Reason for Use

Accuracy

(%)

Correct

predictions over

total samples

General

effectiveness of

classification

Precision

True positives /

(true + false

positives)

Avoid false alarms

in fault detection

Recall

True positives /

(true + false

negatives)

Ensure actual faults

aren't missed

F1 Score

Harmonic mean of

precision and

recall

Balanced

evaluation of model

Inference

Latency

(ms)

Time taken to

make prediction

Important for real-

time edge

deployment

Figure 1: Inference time by model type.

A simulation of the system that mimics a real smart

grid setup is built with GridLAB-D and TensorFlow

Lite, which is used to validate the system’s

performance. The testing involves an amount of

consumption of electricity, solar and wind

generation, and peak load to check the willingness of

users to respond. Performance metrics including

latency, energy distribution accuracy, prediction

error rate, system robustness, and communication

load are compared with a conventional cloud-based

model and a local-only baseline. The findings are

employed to validate the model and evaluate its

practical application. Table 2 and figure 1 shows

inference time by model type.

Table 2: AI model comparison for edge inference.

Model Architecture

Latency

(ms)

Accurac

y (%)

Resourc

e Usage

(MB)

CNN

3 Conv + 2

Dense

120 92.3 45

LSTM

2 LSTM +

Dense

145 94.1 50

MLP

4 Dense

Layers

85 89.6 30

Through this methodology, the research not only

demonstrates the viability of edge-based AI in smart

grid contexts but also establishes a privacy-

preserving, scalable, and adaptive foundation for

next-generation energy systems. Figure 2 shows the

system workflow of edge-intelligent smart grid

energy management framework.

Figure 2: System workflow of edge-intelligent smart grid

energy management framework.

4 RESULTS AND DISCUSSION

Due to the fact that the experimental evidence of the

proposed Edge AI-based decentralized energy

management framework showed significant

performance enhancements in terms of

performance/efficiency, scalability, responsiveness

as compared to traditional centralized approaches.

When different demand-load patterns and different

distributed energy situations were simulated, the edge

models showed a much faster reaction to demand

fluctuations as compared to cloud-dependent models.

Next-Generation Smart Grid Optimization: Integrating Edge AI for Real-Time and Decentralized Energy Management

319

Average decision delay decreased by 42%, so making

almost real-time decisions on energy allocation

during the busiest hours of the day or moment of

renewable generation. Figure 3 shows the latency

comparison: centralized vs edge AI.

Figure 3: Latency comparison: centralized vs edge AI.

Prediction precision was also enhanced as the global

model with federated learning outperformed

individual edge models in the mean absolute error

(MAE) of 3.7% versus 6.1% for isolated setups. This

highlights the importance of collaborative learning

among edge nodes such that data privacy does not get

compromised. Crucially, the system sustained model

stability throughout rounds of training, including

when training on non-IID (non-independent and

identically distributed) data – which is a common

problem in federated systems. Dynamic node

selection and update—Selecting a subset of active

nodes that is constantly varied to perform calculations

and update was also a part of the force aggregation

strategy, and it helped to converge the model with

much less training steps.

Efficiency of communication was another key

measure investigated. Since sending only model

weights and updates, not the raw data, the system

reduced the network traffic by more than 60% and

thus became feasible for network-constrained

environments that are typical in rural or developing

grid regions. The model synchronization mechanism

was still robust even with such overhead of

communication, modules could all still stay up-to-

date on system-wide decision records. Table 3

represents the federated learning cycle timing.

Table 3: Federated learning cycle timing (example results).

Cycle

Local

Trainin

g Time

(s)

Uploa

Time

(s)

Aggregatio

n Time (s)

Total

Time

per

Round

(s)

1 20 5 8 33

2 18 4.8 7.9 30.7

3 21 5.1 8.2 34.3

Decentralized architecturally was highly

advantageous from a resilience perspective point as

well. Local edge nodes could still work

independently, even in the case of transient loss in

communication with the aggregator. This

independence enabled the network to continue to

operate its fundamental functions–load balancing,

fault recognition, and energy re-routing–

uninterrupted and some degree of grid fault tolerance

could be supported. Traditional cloud systems, on

the other hand, had a degraded performance or simply

were not functional under similar network outages.

Figure 4 shows the bandwidth usage.

Figure 4: Bandwidth usage: centralized vs edge AI.

Finally, the results demonstrate the scalability of

our model. Performance improvements with

increased numbers of edge devices were also

sustained, with the system behavior holding and not

degrading as the simulation size is increased,

demonstrating the potential scalability of the

architecture for deployment across a larger region, or

even across a national grid. Energy consumption at

device level was also kept at a manageable level since

the AI models were lightweight models that were

tuned for edge inference purpose. Table 4 represents

the communication latency across grid layers.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

320

Table 4: Communication latency across grid layers.

Communicati

on Layer

Average

Latency

(ms)

Max

Latenc

y (ms)

Technolog

y Used

Edge to

Aggregator

50 120

Wi-Fi 6 /

LTE

Aggregator to

Cloud

90 160

Fiber

Optic

Sensor to

Edge Device

20 45

Zigbee /

Bluetooth

Edge Device

to Control

35 60

LAN /

MQTT

In conclusion, we combined the edge AI and

federated learning in this research to form a smart grid

management system, which is shown to become a

system of both fast/precise as well as keeping privacy

and resilience to the failures. These features are very

significant in addressing the main bottlenecks

identified on the state of the art where centralised

solutions were not scallable nor secure enough to

allow processing of sensitive data. The results

confirm that the framework is now ready to be tested

in pilot implementations and that a promising future

lies ahead for this concept of intelligent, distributed

energy systems.

5 CONCLUSIONS

This paper proposes a prospective framework, which

combines the AI at the Edge and the Federated-

Learning techniques, to solve the fundamental issues

in the current Smart-Grid based energy management.

The proposed system moves the intelligence near the

data generation and consumption points and thereby

facilitates the autonomous and real-time decision-

making independent from the central cloud

infrastructure. The results of our study affirm that

this approach can lead to a substantial reduction in

latency, raise prediction accuracy, and make the grid

more resilient and in a private fashion through

federated model training.

Using small AI models combined with adaptive

learning and effective communication, the framework

can learn to dynamically allocate energy across a

range of different, dynamic contexts. In contrast to

common designs, which are hindered by potentially

low-bandwidth network links and concerned with

privacy issues, the edge-based approach shows a

good scalability, low communication overhead, and

resilience in the presence of network anomalies.

This way, the present work brings together not

only an innovative architectural scheme, but also

outlines an actionable path towards sentient energy

systems. With the worldwide momentum towards

decentralization, sustainability, and digitalization,

embedding Edge AI in smart grids will remain a

paramount requirement to support efficient, secure,

and future-ready energy systems.

REFERENCES

Ahmed, M., & Rehman, M. H. (2022). Machine learning-

based energy load forecasting in smart grid: A survey.

Journal of Cleaner Production, 328, 129532.

https://doi.org/10.1016/j.jclepro.2021.129532

Arcas, G. I., Cioara, T., Anghel, I., Lazea, D., & Hangan,

A. (2024). Edge offloading in smart grid. arXiv preprint

arXiv:2402.01664. https://arxiv.org/abs/2402.01664

Biswal, P., Balamurugan, S., & Sahoo, S. (2024). Role of

artificial intelligence in smart grid—a mini review.

PubMed Central (PMC11832663).

https://www.ncbi.nlm.nih.gov/pmc/arti-

cles/PMC11832663/

Biswal, P., Rashid, A., Al Masum, A., Al Nasim, M. A.,

Ferdous, A. S. M. A., Gupta, K. D., & Biswas, A.

(2025). An extensive and methodical review of smart

grids for sustainable energy management—Addressing

challenges with AI, renewable energy integration, and

leading-edge technologies. arXiv preprint

arXiv:2501.14143. https://arxiv.org/abs/2501.14143

Chatterjee, R., Bose, S., & Misra, S. (2021). Edge-enabled

AI for anomaly detection in smart grids. Computers &

Electrical Engineering, 91, 107010.

https://doi.org/10.1016/j.compeleceng.2021.107010

Dileep, G. (2021). A survey on smart grid technologies and

applications. Renewable Energy, 146, 2589–2625.

https://doi.org/10.1016/j.renene.2019.08.092

Ghosh, S., Sengupta, A., & Datta, S. (2023). A blockchain-

and AI-based secure smart grid framework using edge

nodes. Sustainable Computing: Informatics and Sys-

tems, 40, 100856. https://doi.org/10.1016/j.sus-

com.2023.100856

Habib, M. (2025). Edge AI: Transforming the energy indus-

try with smart, sustainable solutions. LinkedIn Articles.

https://www.linkedin.com/pulse/edge-ai-transforming-

energy-industry-smart-solutions-habib-k58hf

Islam, S., Zografopoulos, I., Hossain, M. T., Badsha, S., &

Konstantinou, C. (2022). A resource allocation scheme

for energy demand management in 6G-enabled smart

grid. arXiv preprint arXiv:2207.00154.

https://arxiv.org/abs/2207.00154

Kumar, N., & Tripathi, A. (2021). Reinforcement learning

for energy management in smart grid. Energy AI, 5,

100072. https://doi.org/10.1016/j.egyai.2021.100072

Li, Y., Wang, Y., & Li, L. (2023). Intelligent energy man-

agement system for smart grid using edge computing

Next-Generation Smart Grid Optimization: Integrating Edge AI for Real-Time and Decentralized Energy Management

321

and deep learning. Energy Reports, 9, 1203–1214.

https://doi.org/10.1016/j.egyr.2023.01.095

Liu, F., Huang, Y., & Tao, L. (2023). Joint optimization of

computation and communication for edge AI in energy

grids. Ad Hoc Networks, 139, 103103.

https://doi.org/10.1016/j.adhoc.2023.103103

Luo, Y., Zhang, T., & Zhang, Q. (2025). A decentralized

multi-agent system for smart energy distribution using

edge AI. Journal of Energy Storage, 75, 108124.

https://doi.org/10.1016/j.est.2024.108124

Nandhakumar, A. R., Baranwal, A., Choudhary, P., Golec,

M., & Gill, S. S. (2023). EdgeAISim: A toolkit for sim-

ulation and modelling of AI models in edge computing

environments. arXiv preprint arXiv:2310.05605.

https://arxiv.org/abs/2310.05605

Shinde, D. B. (2025). Energy grid optimization—AI & dig-

ital technologies for improving efficiency. Cyient Blog.

https://www.cyient.com/blog/energy-grid-optimiza-

tion-ai-digital-technologies-for-improving-efficiency

Singh, A., & Gupta, R. (2024). Demand response optimiza-

tion using federated learning at the grid edge. IEEE

Transactions on Industrial Informatics.

https://doi.org/10.1109/TII.2024.3334412

Sun, H., Wang, Z., & Liu, C. (2021). Edge-based data-

driven approach for real-time smart grid management.

Electric Power Systems Research, 196, 107257.

https://doi.org/10.1016/j.epsr.2021.107257

Tan, J., & Ramachandran, G. (2023). AI-enabled fault de-

tection in smart grid substations using edge analytics.

Sensors, 23(5), 2479.https://doi.org/10.3390/s2305247

Ullah, Z., & Khan, I. (2022). Edge computing for smart

grid: A comprehensive survey. IEEE Access, 10,

50015–50045. https://doi.org/10.1109/AC-

CESS.2022.3172471

Wang, K., Chen, C., Liu, Y., & Guo, S. (2021). Edge com-

puting for smart grid: An overview on architecture and

key technologies. IEEE Network, 35(5), 56–63.

https://doi.org/10.1109/MNET.011.2000274

Yang, J., Zhang, H., Zhou, T., & Xu, W. (2022). AI-enabled

optimization in smart grid operation: A review. Applied

Energy, 308, 118302. https://doi.org/10.1016/j.apen-

ergy.2021.118302

Yu, R., Xie, S., & Li, Y. (2024). Smart grid energy distri-

bution optimization using graph neural networks. IEEE

Transactions on Industrial Informatics, 20(2), 1035–

1044. https://doi.org/10.1109/TII.2023.3298427

Zhan, C., Lu, J., & Wang, J. (2023). Federated learning in

smart grids: Privacy-preserving demand prediction with

edge devices. IEEE Transactions on Smart Grid, 14(1),

112–124. https://doi.org/10.1109/TSG.2022.3198710

Zhang, C., Wu, J., Zhou, Y., Cheng, M., & Long, C. (2021).

Peer-to-peer energy trading in a microgrid with edge

computing and blockchain. Energy Procedia, 159, 261–

266. https://doi.org/10.1016/j.egypro.2019.01.179

Zhao, J., Wang, J., & Lin, Y. (2022). Energy-efficient re-

source scheduling for smart grid IoT with edge compu-

ting. Future Generation Computer Systems, 128, 219–

230. https://doi.org/10.1016/j.future.2021.10.018

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

322