Hybrid-Oriented Intelligent Operational and Architectural Foundations of IoT-Enabled Smart Grids: A System-Level Review and Challenge-Oriented Comparative Synthesis

Abstract

The rapid digitalization of energy systems and the increasing integration of distributed energy resources, renewable energy technologies, and prosumer-oriented infrastructures have accelerated the development of IoT-enabled Smart Grids as a foundation for intelligent and adaptive energy management. Modern Smart Grids increasingly depend on the coordinated interaction of IoT architectures, artificial intelligence, distributed analytics, and decentralized control mechanisms to ensure reliability, scalability, and real-time operational flexibility. Despite extensive research activity, existing studies remain predominantly technology-centric, focusing on isolated architectural layers or individual intelligent methods without providing a unified system-level perspective on their coordinated operation and interoperability. This article presents a system-level integrative review and challenge-oriented comparative synthesis of intelligent operational and architectural foundations of IoT-enabled Smart Grids. The study analyzes data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented intelligent paradigms within multi-layer IoT energy infrastructures. In addition, the research establishes a cross-layer mapping between Smart Grid operational challenges, enabling technologies, and corresponding analytical approaches while identifying interoperability constraints, scalability limitations, and coordination challenges associated with decentralized energy ecosystems. The conducted synthesis demonstrates that hybrid-oriented intelligent approaches represent the most promising direction for future Smart Grid evolution due to their ability to integrate AI, ML, digital twins, semantic reasoning, and decentralized multi-agent coordination within unified IoT architectures. The conducted comparative synthesis identifies the ongoing transition from isolated intelligent solutions toward integrated hybrid cyber–physical energy ecosystems and highlights key characteristics of future adaptive, interoperable, scalable, and explainable Smart Grid architectures.

1. Introduction

1.1. Operational Relevance and Technological Evolution of IoT-Enabled Smart Grids

Current trends in the rapid development of intelligent approaches to improving energy systems are driven by a wide range of global challenges related to energy security, decarbonization, and the digital transformation of production and business processes. Traditional power systems that rely on centralized generation technologies and one-way reactive power control are proving to be insufficiently efficient for the integration of renewable energy sources (RES) and decentralized energy resources (DER). In this context, the concept of the Smart Grid, integrated with Industry 4.0 technologies, in particular the Internet of Things (IoT), artificial intelligence (AI), machine learning (ML), and digital twins (DT), is becoming a priority area for the development of the modern energy sector [1,2,3,4]. The rapid development of approaches to the intellectualization and decentralization of energy systems and complexes is driven by their significant impact on economic, environmental, technological, and social outcomes.

The ongoing digital transformation of energy infrastructures is increasingly driven by the integration of IoT technologies, DERs, intelligent analytics, and adaptive control mechanisms. These developments significantly enhance the observability, flexibility, and operational efficiency of modern power systems while simultaneously increasing their architectural complexity and coordination requirements. As a result, Smart Grids have emerged as a fundamental technological foundation for the transition toward intelligent, resilient, and sustainable energy ecosystems [5,6,7,8].

In addition to operational benefits, Smart Grid technologies contribute to sustainability objectives by facilitating the integration of renewable energy, reducing transmission losses, improving energy efficiency, and supporting the decarbonization of energy systems [9,10].

It is also worth noting that the Smart Grid serves as the technological foundation for the development of smart cities and digital infrastructure in many countries worldwide. The integration of smart and digital solutions into energy grids brings numerous technological and social benefits. The main benefits are as follows: improved reliability of power systems (e.g., self-healing grids and predictive maintenance) [11], the empowerment of consumers (prosumers) [12], the practical implementation of two-way ‘consumer–grid’ interaction [13], enhanced safety and resilience of power systems [14], as well as the support and innovative development of industrial structures [15].

Such a comprehensive impact resulting from the creation and implementation of digitalized and intelligent energy systems is made possible, in particular, by the dynamic development and rapid adaptation of practical approaches to decentralized electricity generation. In particular, these include the integration of cyber–physical systems and the IoT, the use of advanced AI and ML algorithms for predictive and optimization data analytics, the active integration of RES and electric vehicles (EVs) into the power grid, as well as the transition to autonomous energy management systems [16,17,18]. Taken together, these trends enable the practical implementation of the conceptual paradigm of an intelligent power system capable of adapting to changes in influencing parameters and factors in real time and of rationally balancing the modes of electricity generation, storage, and consumption. This development is driven by global and local initiatives, particularly those set out in regulatory and analytical documents that facilitate the harmonization and standardization of approaches, as well as the scaling up of best international practice, as systematically illustrated in Figure 1.

An analysis of the strategic initiatives presented indicates a clear convergence in approaches to the development of smart energy systems at all levels. A common feature across all levels is a focus on decarbonizing the energy sector, integrating RES, improving energy efficiency, and implementing the Smart Grid concept as the core platform for modernizing electricity systems. At the same time, there is a clear trend toward the systematic strengthening of the role of digital and intelligent technologies, which provide the technological foundation for the practical implementation of Industry 4.0 conceptual principles in the energy sector. The differences between these levels lie primarily in the level of detail and in the degree of practical implementation of the identified priorities. Global initiatives are more general in nature and define strategic development directions, thereby establishing the framework for the transformation of energy systems. The European level is characterized by greater regulatory specificity, including mechanisms for the functioning of the electricity market, the integration of distributed energy resources (DER), the development of smart metering, and the promotion of innovation through targeted programs. Thus, the alignment of strategic priorities, coupled with varying degrees of implementation, confirms a sustained global trend toward the digital transformation of the energy sector and the development of smart grids. This, in turn, demonstrates the relevance and necessity of research into IoT technologies to enhance Smart Grid efficiency.

1.2. System-Level Challenges and Technological Transformation Drivers in Smart Grids

The current development of energy systems is taking place against a backdrop of rapidly increasing functional complexity. This trend is driven by the integration of DERs and by prosumers’ active involvement. Traditional centralized approaches to energy system management are proving insufficient to ensure the necessary levels of flexibility, adaptability, and reliability, particularly given the high share of RES and the growing role of electric mobility. In this context, global challenges are emerging regarding technology interoperability [19,20], real-time decision-making [21,22], cybersecurity [19,23,24], DER integration, and the effective coordination of energy flows [25,26,27]. Overcoming these challenges requires the systematic implementation of IoT technologies that ensure the continuous acquisition, processing, and exchange of data across energy system components [28,29].

Figure 2 illustrates the systemic logic underlying the transformation of modern energy systems toward an intelligent, IoT-based Smart Grid infrastructure through the sequential interaction of challenges, solutions, technologies, and their systemic effects.

At the initial level, the key challenges that significantly hinder the digital development of modern power systems have been identified. These include limited technological interoperability, insufficient efficiency in real-time decision-making, fragmented integration of AI, the complexity of integrating and coordinating DER, compatibility issues among heterogeneous energy sources, cybersecurity risks, limited prosumer integration, and insufficient integration of electric vehicles (EVs) as energy assets. These issues are widely discussed in contemporary Smart Grid research, highlighting the increasing complexity of managing distributed energy systems and the need to transition to decentralized models [30].

The next level outlines possible solutions to the aforementioned problems, which primarily include the standardization and unification of technological solutions [31,32], the intelligent coordination of DER [33], the implementation of intelligent control and optimization, the development of unified energy flow monitoring and control systems, the use of secure protocols and architectures [19], as well as the improvement of decision support systems and the implementation of vehicle-to-grid (V2G) mechanisms [34]. An analysis of relevant scientific sources demonstrates that the integration of such approaches and the implementation of corresponding technological solutions enable a transition from traditional centralized networks to adaptive, self-regulating power systems. In particular, the priority technologies that deliver a significant positive impact on the digitalization of energy systems include information and communication technologies (ICT), machine learning (ML) algorithms and model predictive control (MPC), DT, Big Data analytics, explainable artificial intelligence (XAI), Generative AI (GenAI), and related approaches [35,36,37]. Recent studies emphasize that the integration of such technologies into a multi-level IoT network infrastructure is essential for ensuring real-time operation, prediction, and optimization of energy systems.

Thus, an approach based on the practical development of an IoT-based Smart Grid infrastructure enables bidirectional energy and data exchange, the integration of DER, and adaptive, autonomous control of energy processes. Such a transformation leads to significant improvements in the reliability, efficiency, and resilience of energy systems, while also ensuring the active participation of prosumers in energy processes.

1.3. Review Motivation

Despite the rapid growth of research on IoT-enabled Smart Grids, the existing scientific literature remains fragmented and predominantly technology-centric. Most studies focus on isolated technological components, communication infrastructures, data processing mechanisms, or specific optimization algorithms without sufficiently addressing their systemic interdependencies within complex cyber–physical Smart Grid foundations. As a result, the interaction between IoT architectural layers and intelligent decision-making mechanisms is often analyzed separately, resulting in limited interoperability among operational, analytical, and control processes.

More critically, contemporary research demonstrates a substantial imbalance between the rapid evolution of enabling technologies and the development of unified methodological paradigms capable of supporting adaptive and scalable Smart Grid intelligence. Existing approaches based on data-driven analytics, model-driven optimization, knowledge-driven reasoning, and agent-based coordination are typically investigated independently and remain fragmented at the architectural level. Such fragmentation significantly constrains the development of coherent intelligent energy systems capable of simultaneously ensuring real-time responsiveness, explainability, interoperability, resilience, and decentralized coordination under conditions of uncertainty and dynamic operational variability.

Furthermore, the increasing penetration of distributed energy resources, renewable energy sources, electric vehicles, prosumer-oriented infrastructures, and AI-driven decision-making mechanisms fundamentally changes the operational logic of modern energy systems. Under these conditions, traditional management models become insufficient for handling the growing complexity, heterogeneity, and stochasticity of Smart Grid environments.

Therefore, there is a strong need for a system-level, integrative perspective that bridges the conceptual gap between IoT infrastructure architectures and intelligent control paradigms. Such an approach should not only classify existing technological solutions but also establish a unified operational framework for the coordinated orchestration of heterogeneous intelligent mechanisms within future adaptive Smart Grid ecosystems.

Based on the identified limitations of the existing literature, the principal research gap addressed in this study can be formulated as the lack of a system-level integrative understanding of how different intelligent Smart Grid paradigms complement each other and contribute to addressing the operational and architectural challenges of IoT-enabled energy ecosystems. Such an approach should not only classify existing technological solutions but also establish a unified operational framework for the coordinated orchestration of heterogeneous intelligent mechanisms within future adaptive Smart Grid ecosystems.

So, unlike most existing review studies on Smart Grids, which typically focus on individual technologies, specific operational functionalities, or isolated analytical paradigms, the present work adopts a system-level perspective that simultaneously considers the architectural, operational, and intelligence-related dimensions of Smart Grid evolution. In addition to reviewing data-driven, model-driven, knowledge-driven, agent-based, and hybrid approaches, the study provides a challenge-oriented comparative synthesis that explicitly links Smart Grid operational challenges with the capabilities, limitations, and applicability domains of different intelligent paradigms. This integrated analytical perspective enables a more comprehensive understanding of how heterogeneous intelligent mechanisms can be orchestrated within modern IoT-enabled Smart Grids.

1.4. Review Scope, Objectives and System-Level Analytical Approach

The primary aim of this study is to develop a system-level integrative understanding of intelligent operational and architectural foundations of IoT-enabled Smart Grids through the comparative analysis, functional decomposition, and challenge-oriented synthesis of modern intelligent paradigms operating within cyber–physical energy ecosystems. In contrast to existing technology-centric review studies that predominantly investigate isolated analytical methods or individual infrastructure components, this work seeks to establish a unified conceptual perspective that integrates IoT architectures with intelligent monitoring, prediction, optimization, orchestration, and decentralized decision-making mechanisms. To achieve this aim, the following research objectives are defined: to perform a systematic review and comparative synthesis of data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented intelligent approaches applied in Smart Grid environments; to identify the functional capabilities, interoperability constraints, scalability limitations, coordination challenges, and architectural dependencies associated with different intelligent paradigms; to conduct a cross-layer system-level mapping of IoT-enabled Smart Grid architectures, including sensing, communication, computational, orchestration, and intelligent control layers; to establish a challenge-oriented mapping between Smart Grid operational problems and the most suitable intelligent analytical approaches; to investigate the role of hybrid-oriented intelligent ecosystems as an integrative foundation for future adaptive and decentralized Smart Grid operation; and to formulate a unified system-level analytical perspective supporting the interpretation of interoperability, explainability, scalability, resilience, and adaptive orchestration within next-generation intelligent energy infrastructures.

The research process includes the identification and classification of relevant scientific sources, the comparative evaluation of intelligent paradigms, the functional interpretation of operational Smart Grid requirements, and the system-level synthesis of hybrid orchestration mechanisms. Furthermore, the conducted analysis incorporates architectural and operational perspectives to establish relationships between IoT infrastructures, distributed intelligence mechanisms, and cyber–physical coordination processes within modern Smart Grid ecosystems. The proposed research approach enables a transition from fragmented, technology-oriented analysis to an integrated understanding of the evolution of intelligent Smart Grids, emphasizing the growing importance of hybrid architectures, decentralized coordination mechanisms, semantic interoperability, and adaptive, intelligent orchestration for future IoT-enabled energy systems.

1.5. System-Level Contributions and Applied Impact

The main contribution of this study lies in providing a system-level integrative understanding and systematization of intelligent IoT-enabled Smart Grid foundations through the coordinated synthesis of architectural, operational, and analytical dimensions. The study further offers a comparative synthesis of data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented intelligent paradigms, highlighting their complementary roles, capabilities, and applicability within multi-layer IoT-enabled energy infrastructures. A systematized graphical interpretation of the main contributions of this article’s results is shown in Figure 3.

2. Methodology

The conducted methodology includes a comprehensive analysis of contemporary architectural paradigms, intelligent orchestration mechanisms, digital transformation strategies, and emerging technologies such as AI, ML, DT, edge/fog computing, Big Data analytics, and V2G integration within modern Smart Grid infrastructures. Particular attention is devoted to identifying functional limitations, interoperability challenges, cross-layer coordination issues, and scalability constraints associated with intelligent energy systems, with a specific focus on data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented approaches for monitoring, predictive analytics, optimization, and decentralized energy management.

Considering the rapid evolution of intelligent energy ecosystems and the interdisciplinary nature of Smart Grid digitalization, a system-level methodological strategy combining bibliometric analysis, comparative synthesis, architectural decomposition, and challenge-oriented interpretation was adopted. Such an analysis is necessary to identify the main scientific directions, the evolution of research trends, and the interrelationships among technologies that shape the current Smart Grid development paradigm. The methodological structure of this research is organized in a sequential decomposition of the investigated subject area into several interconnected stages, as graphically illustrated in the block diagram shown in Figure 4. The criteria and characteristics of the bibliographic search and analysis of scientific sources considered in this study are presented in

Table 1. Informative criteria and characteristics of bibliographic search and analysis of scientific sources.

Category	Criteria for the Selection and Evaluation of Scientific Literature
Scientometric databases	Scopus, Web of Science Core Collection
Publication time range	from 2010 to 2026
Types of documents	Scopus: conference paper, article, book chapter, review, book; Web of Science: article, proceeding paper, review article, book chapter, data paper
Language of documents	English
Type of search query	keywords
Search query combinations	“smart grid” AND “internet of things” AND “energy systems” “smart grid” AND (“IoT” OR “AI” OR “machine learning”) AND “Industry 4.0” “smart grid” AND (“digital twin” OR “big data” OR “edge computing”) AND “energy management”
Type of analysis	co-occurrence
Unit of analysis	all keywords
Counting method	full counting
Minimum number of keywords	3

.

After executing the bibliometric search queries, the retrieved records were subjected to a multi-stage screening process. The bibliographic search initially identified 660 publications in Web of Science (WoS) and 302 publications in Scopus, which constituted the primary sources for the subsequent screening and selection procedure. First, duplicate publications identified across databases were removed. Subsequently, titles, abstracts, and keywords were examined to assess their relevance to the scope of IoT-enabled Smart Grid intelligence, intelligent energy management, distributed energy resources, and advanced operational paradigms. Studies primarily focused on unrelated communication technologies, conventional power system analysis without intelligent control mechanisms, and non-energy IoT applications were excluded. The final corpus of literature was used to support the bibliometric analysis, the comparative assessment of intelligent paradigms, the architectural decomposition, and the challenge-oriented synthesis presented in this study. A graphical interpretation of the results of the bibliographic analysis of the thematic taxonomy of research trends in IoT-based Smart Grids is shown in Figure 5. For clarity, the proposed taxonomy in Figure 5 should be interpreted from top to bottom. The upper layer represents the principal thematic domains identified through the literature analysis, while the lower layer summarizes the corresponding bibliometric evidence and emerging research patterns. This layered structure enables the systematic tracing of relationships between technological foundations, intelligent paradigms, and future Smart Grid development directions.

The results, shown in Figure 5, reveal a high degree of consistency in the structure of scientific research in the field of the digitalization and intellectualization of Smart Grid systems. The graphical representations demonstrate that the Smart Grid serves as a practice-oriented platform for implementing innovative solutions grounded in the conceptual principles of IoT. The high density of inter-node connections in graph models centered on IoT and Industry 4.0 confirms their status as global integrative concepts for the digitalization and intellectualization of the electricity sector. In turn, the IoT is often regarded as a technical and functional enabler of the conceptual foundations of Industry 4.0, serving as an integrating element across a range of technological clusters, including AI, ML, Big Data, predictive models, edge and cloud computing, DT, cybersecurity, and related domains. Thus, the conducted bibliographic analysis indicates that the current trend in Smart Grid development is focused on the creation of autonomous, self-adaptive, and predictive energy systems that integrate various energy sources and means of monitoring and controlling their status in real-time. This finding highlights the necessity and importance of further research into the criteria-based analysis and logical generalization of methodological principles and technological solutions to enhance the level of digitalization and intellectualization of Smart Grids.

3. Architectural Decomposition of a Multi-Layer IoT-Based Smart Grid Infrastructure

3.1. Multi-Layer IoT Models

In current practice, the synthesis of multi-layer architectures is a fundamental approach to designing IoT-based systems, particularly in Smart Grids. This approach ensures the modularity, interoperability, and scalability of complex technological infrastructures. The decomposition of a system into functional levels enables the distinction between data acquisition and processing, information and communication processes, and application services, thereby significantly simplifying the integration of new technologies and improving the efficiency of energy process management. Relevant scientific and applied research emphasizes that the IoT enables the digitalization of energy assets, data collection, and subsequent analytics, which form the conceptual basis for Smart Grid operation [38,39]. In the context of the development of IoT-based solutions, several reference architectural models have been established, the most widely used being the International Telecommunication Union (ITU-T) Y.2060 model [40] and the model developed based on the findings of the IoT World Forum (IWF) [41]. These models formalize the structure of IoT systems by defining the main layers and their interactions, which is particularly important for decentralized technological infrastructures such as the Smart Grid.

The ITU-T Y.2060 model (International Telecommunication Union, Geneva, Switzerland) describes IoT system solutions as a four-layer architecture comprising device, network, service support, and application layers. This approach provides a high level of abstraction and versatility, enabling its application across various domain areas. In turn, the IWF model offers a more detailed seven-layer structure, in which the levels of data processing, storage, and abstraction, as well as business logic, are distinguished, making it more suitable for complex industrial IoT systems [40,41,42]. The results of a detailed comparison of the main technological and functional characteristics of these reference IoT models are presented in

Table 2. Results of a comparative analysis of reference IoT models.

Technical and Functional Characteristics	ITU-T Y.2060 Model	IWF Model
Number and types of functional levels	4 main levels (device, network, service support, and application levels) and 2 cross-levels (control and security capabilities)	7 main levels (devices and controllers, communication, peripheral computing, data accumulation, data abstraction, applications, and interactions and processes levels)
Level of abstraction	high	detailed
Applied focus	generalized	industrial
Data processing	focused on the service support level	distributed by levels
Business logic	partially taken into account	listed separately at the top level
Suitability for Smart Grids	conceptual model	detailed synthesis and design

.

Research papers indicate that multi-layer IoT architectures in electrical engineering applications typically include data collection layers (e.g., sensors and smart meters), information and communication networks, and computing services and platforms, which together enable the monitoring and control of energy systems [43,44]. The choice of an architectural model for IoT-based Smart Grid infrastructure depends on the required level of detail. Generalized models, such as ITU-T Y.2060, are suitable for conceptual analysis and standardization, whereas more detailed approaches, particularly those based on the IWF model, allow for the description of real-world data processing and component interactions within energy systems.

The implementation of IoT architectures in Smart Grids is characterized by several specific features, including the need to ensure bidirectional flows of data and energy, the integration of a large number of heterogeneous devices, and stringent requirements for reliability, cybersecurity, and real-time data processing. Furthermore, the integration of IoT into Smart Grids involves the use of distributed computing models and standardized protocols to ensure interoperability and the efficient management of energy resources. This determines the feasibility of using multi-layer architectures with a clear division of functions between levels, which is a key prerequisite for creating scalable and resilient modern energy systems [45,46].

The subsequent analysis in this section focuses on the four-layer architecture. This choice is motivated by its ability to ensure an optimal balance between abstraction and detail within the Industry 4.0 conceptual framework, clearly distinguishing the key functional domains of data acquisition, transmission, processing, and application. This decomposition is sufficiently functional and objective when aligned with existing reference IoT models, while retaining practical applicability for the analysis and design of Smart Grid systems.

3.2. Level of Measurement Data Collection

This level of the IoT-based Smart Grid infrastructure represents the foundational layer. It is responsible for directly collecting information from the physical environment using sensors and measuring devices. It generates the primary data stream, which is subsequently used to monitor and control energy systems.

At this level of the Smart Grid, key electrical and operational parameters are measured, including voltage, current, frequency, active and reactive power, energy consumption, and power quality parameters (e.g., harmonics, overvoltage, fluctuations, and others). In addition, equipment condition monitoring is carried out (e.g., transformer temperature, vibrations, partial discharges, pressure, and others), which is critical for implementing predictive maintenance and improving grid reliability. Typical components at this level include smart meters, electrical sensors, equipment status sensors, identification components, GPS, and other devices [47,48,49,50].

In the context of the Smart Grid, and given the conceptual principles underlying the design of IoT systems, the data collection layer is characterized by high device heterogeneity and the need to ensure high measurement accuracy and reliability. These characteristics are further complemented by the conceptual requirements of Industry 4.0 for real-time operation [51] and by the ability to operate within a distributed energy infrastructure encompassing generation, transmission, and consumption of electricity [52,53]. Consequently, this level plays a functionally important role in the digitalization of energy systems, ensuring the continuous collection of high-precision data on the state of the grid and its components, which serves as the basis for implementing intelligent monitoring and control in the Smart Grid.

3.3. Network Level

The network level enables bidirectional data and information flow within the IoT infrastructure of the Smart Grid. The Smart Grid’s information and communication infrastructure relies on both wired and wireless technologies. Wired solutions include fiber-optic networks [54,55] and power line communication [56,57], which provide high bandwidth and reliable data transmission. Wireless technologies such as ZigBee, LoRaWAN, NB-IoT, Wi-Fi, and 5G enable flexible connectivity for large numbers of distributed devices [58,59]. The key parameters characterizing this level include bandwidth, transmission latency, reliability, scalability, and energy efficiency [60]. Low latency and high reliability are particularly important in Smart Grids, as the network infrastructure is used not only for monitoring but also for real-time control of energy processes. An equally important feature of the IoT-based Smart Grid infrastructure at this level is the need to ensure cybersecurity and data protection, given the critical nature of the energy infrastructure. This includes the use of encryption, authentication, and access control mechanisms, as well as ensuring resilience against cyberattacks and failures [61,62].

From an architectural perspective, this level implements a multi-layer network hierarchy comprising the Home Area Network, Neighborhood Area Network, and Wide Area Network, which facilitates data transmission from end users to control centers [38,63]. This approach enables effective data aggregation and ensures system scalability.

Thus, the network level serves as an integrating component of the IoT-based Smart Grid infrastructure, ensuring the efficient exchange of data and information between all components and layers of the system, thereby providing a functional foundation for the implementation of intelligent, decentralized monitoring and control of energy processes.

3.4. Level of Data Processing

This level in the IoT-based Smart Grid infrastructure is responsible for comprehensive data processing and analytics. At this level, unstructured measurement data undergoes a fundamental transformation into useful information to support decision-making and the implementation of intelligent control of the energy system. From an architectural perspective, this level is realized through a combination of distributed computing paradigms: edge, fog, and cloud computing [64,65]. The edge layer provides primary data processing directly at the sources of generation, minimizing delays and reducing the volume of data transmitted. The fog level performs intermediate aggregation and coordination functions between local nodes and the cloud infrastructure, while the cloud level provides scalable storage, deep analytics, and integration with application services [66].

The main procedures implemented at this level in IoT systems, particularly when applied in Smart Grids, include data pre-processing (filtering, normalization, standardization, and removal of noise and anomalies); the aggregation and integration of data from various sources; real-time analytics and batch processing; the application of AI and ML algorithms for prediction and optimization; the generation of control signals for the upper and lower levels of the system; and the storage of large volumes of data.

In the context of Smart Grids, this level processes both real-time data and historical information, taking into account formalized factors that determine the technical and functional performance of specific energy systems. This enables a wide range of tasks, including load prediction [67], anomaly detection [68], energy distribution optimization [69], predictive equipment maintenance [70], and others. The utilization of distributed computing resources facilitates harmonization between the competing requirements of performance and the computational complexity of processing data streams, particularly in dynamic energy systems. Similar to the network level, this level is also characterized by heightened cybersecurity requirements, particularly through the implementation of access control mechanisms, encryption, and related measures [71].

Therefore, the data-processing level serves as the computational and analytical core of the IoT-based Smart Grid infrastructure, providing practical mechanisms for intelligent monitoring and control of energy processes through the effective integration of data analytics and distributed computing.

3.5. Application Level

The application level is the uppermost level of the IoT-based Smart Grid infrastructure and is responsible for the interaction of functional services designed to facilitate user-friendly interaction with all participants in the processes of electricity generation, storage, distribution, and consumption. The primary functional purpose of this level is to interpret the results of data processing and transform them into application-level solutions for the monitoring and control of energy processes. The application level should also provide data visualization, decision support, automated control, and integration with energy system business processes [72,73]. These functions are implemented through applications such as SCADA, EMS, MRP, and demand-management platforms. At the system level, control mechanisms are distributed throughout the architecture and include embedded local control at the device level, transmission of control commands through communication infrastructures, formation of control actions based on analytical processing, and supervisory control and decision support at the application level. Currently, the application level of Smart Grids provides a wide range of services, including real-time monitoring of the energy grid status; load and generation prediction; optimization of energy resource distribution; demand management and energy system balancing; and user interaction via user-friendly interfaces [74,75].

A distinctive feature of this level is its integration with enterprise information systems and energy markets. This facilitates effective resource management and support for emerging energy consumption models, such as the prosumer paradigm [76]. Furthermore, the application level actively leverages analytics and intelligent models to enhance decision-making efficiency, particularly through the use of GenAI approaches [77].

3.6. Cross-Level Interaction and Key Issues in Multi-Layer Architecture

To summarize the functional structure of the analyzed four-layer architecture of the IoT-based Smart Grid infrastructure, it is reasonable to present it in the form of a matrix that illustrates the correspondence between the architectural levels and the main technical and functional characteristics. In contrast to conventional descriptive methodologies, this form of presentation (see

Table 3. Functional matrix of IoT-based Smart Grid architecture.

Functional Level	Data Collection	Data Transmission	Data Processing	AI Analytics	Control	Data Types
Device level	Voltage, current, frequency, active/reactive power, energy consumption, voltage quality (harmonics, overvoltages), equipment status (temperature, vibrations, partial discharges, pressure)	Local transmission at the level of wireless sensor networks	Preliminary analog and digital processing of sensor output signals	Not applicable	Local control actions (embedded control)	Raw data, telemetry data, alarm and/or event signals
Network level	Not applicable	Data transmission using a wide range of wired and wireless technologies at various hierarchical levels	Buffering, routing, packet aggregation	Not applicable	Transmission of control signals	Streaming data, control messages, packets
Processing level	Aggregation from distributed sources	Integration of data streams	Filtering, normalization, cleaning, storage	Load forecasting, anomaly detection, optimization, predictive maintenance	Formation of controlling influences	Aggregated data, processed data, historical information, formalized models
Application level	Not applicable	Integration with external information technology and enterprise management systems	Interpretation of the results	Decision support, scenario analysis	Global control (SCADA, EMS, Demand response, HMI)	Decision information, control, commands, and user-level data

) allows for the systematic integration of technological aspects, data types, and functional indicators that are important when designing IoT-based Smart Grid solutions.

In addition to the functional decomposition presented in Table 3, it is also reasonable to analyze the key challenges identified in the review of relevant scientific sources cited in Section 3.2, Section 3.3, Section 3.4 and Section 3.5 of this article. This allows for a more comprehensive assessment of the limitations of existing IoT-based Smart Grid infrastructures and the identification of potential avenues for their further development, as summarized in

Table 4. Results of the analysis of challenges and prospects for the development of IoT-based Smart Grid infrastructure by functional level.

Functional Level	Key Challenges and Constraints	Promising Areas for the Development
Device level	Significant heterogeneity of devices, power consumption, the need for high measurement accuracy, and degradation of sensitive components due to exposure to harsh environments	The use of energy-efficient IoT devices, a widespread transition to smart sensor technologies with built-in self-diagnostic algorithms, the development of interoperability standards, and improvements to protective materials for sensitive components
Network level	Ensuring low latency and high reliability, network overload, and cyber threats	Large-scale deployment of 5G/6G and LPWAN technologies, use of software-defined networking and network functions virtualization, optimization of quality of service, implementation of modern cryptographic methods
Processing level	Processing large volumes of data, optimizing the balance between latency and computational complexity, and ensuring data and information security	Scaling edge and fog computing technologies, refining and optimizing AI and ML models, and implementing modern cryptographic protocols
Application level	System interoperability, data confidentiality, decision-making complexity, and integration with energy markets	The use of open standards, the implementation of multimodal large language model (LLM)-based decision support systems, and the standardization of energy management platforms

.

Consequently, summarizing the results presented in Table 3 and Table 4 has enabled a systematic analysis of the functional distribution of processes and the problem-oriented aspects of implementing the IoT-based Smart Grid infrastructure. The analysis of the functional distribution demonstrates that the mandatory procedures for data collection, transmission, and primary processing are decentralized, whereas intelligent analytics and decision-making are concentrated at the upper levels of the architecture. Accordingly, the range of challenges is distributed across the levels. At the device level, hardware and metrological constraints dominate; at the network level, latency and reliability requirements prevail, whereas at the processing and application levels, issues of scalability, data integration, and the complexity of analytical models are decisive. It is worth noting that a common challenge across levels of IoT architecture is cybersecurity and information protection.

The promising areas for development outlined in Table 4 demonstrate that the technology stack is evolving in a coordinated manner, aligned with the identified challenges and constraints. The prevailing trends involve shifting some analytical and control functions closer to data sources, thereby reducing latency, enhancing resilience, and improving cybersecurity for the IoT-based Smart Grid infrastructure.

Therefore, the results of the analysis confirm that the effective implementation of any IoT-based Smart Grid infrastructure requires scientific substantiation of a comprehensive methodological approach to the design of digitized and intelligent energy systems, achieved through the systematic integration of the latest advances in sensor, computing, networking, and information technologies. On this basis, a generalized diagram of the cross-level interaction of the IoT-based Smart Grid architecture was developed, as shown in Figure 6.

4. System-Level Approaches for Intelligent Monitoring and Predictive Control of Smart Grids

4.1. General Classification of Approaches

In contemporary IoT-based Smart Grid infrastructures, a single approach to monitoring and control is insufficient due to the considerable structural complexity and the dynamic nature of processes within energy systems. This necessitates the use of various conceptual approaches to the intelligentization and digitalization of Smart Grid operational processes. Each approach should be oriented toward solving a specific class of problems, such as intelligent analytics of large volumes of spatiotemporal data, physics-informed predictive control, decentralized and reliable interaction among system components, and others.

In contemporary scientific and technical practice, the following approaches have gained the most traction in the context of the intellectualization and digitalization of energy systems: data-driven [78,79], model-driven [80,81], knowledge-driven [82], agent-based [83,84,85], and hybrid-oriented approaches [86,87,88,89]. The generalized technical and functional characteristics of these approaches are summarized graphically in Figure 7.

Therefore, the classification of approaches presented in Figure 7 forms, within the scope of this study, a conceptual framework for the further analysis and logical generalization of modern system-level approaches and technologies for intelligent monitoring and predictive control in IoT-based Smart Grid infrastructures.

4.2. Data-Driven Approach

This approach employs AI and ML methods, particularly deep learning, to analyze large volumes of data generated by IoT devices in Smart Grids. At the core of this approach lies the intelligent analysis of spatiotemporal data, enabling the identification of hidden patterns in measurement data streams. In Smart Grids, the data-driven approach is used to predict electricity generation and consumption, as well as to detect anomalies and equipment failures. This improves the efficiency of monitoring and provides the information basis for adaptive control of energy processes. The results of the analysis of known scientific studies on the development and application of the data-driven approach in Smart Grids are presented in

Table 5. Results of the analysis of research into the data-driven approach within the context of Smart Grids.

Subject of the Research	Technologies and Approaches Used	Scientific and Practical Outcomes Achieved	References
A data-driven, cyber-secure optimization approach for dynamic energy management in EV-coupled Smart Grids	Electric vehicle–grid communication, CatBoost, a lightweight blockchain-inspired security protocol	Development and experimental validation of a cyber-resilient, data-driven real-time framework for energy management in EV-integrated Smart Grids, which combines optimization, adaptive forecasting under incomplete data, and a lightweight blockchain-based approach to cybersecurity, thereby ensuring simultaneous improvements in the efficiency, accuracy, and security of energy processes	[90]
A renewable-aware, data-driven framework for zonal power quality monitoring in Smart Grids	Clustering (K-means), statistical analysis, and adaptive monitoring strategies	Development and validation of an adaptive power quality monitoring model based on a Renewable Variability Index (RVI), enabling risk-based zonal classification of power networks and reducing data acquisition requirements without compromising disturbance detection performance	[91]
Data-driven architecture of smart renewable energy microgrids in non-interconnected zones	IoT, big data, web technologies	Development and implementation of a data-driven architecture for smart renewable energy microgrids in non-interconnected zones, addressing contextual constraints and enabling efficient, scalable energy management in isolated regions	[92]
Data-driven methods and technologies for energy optimization in smart building systems	Big data, AI, ML, IoT, edge and cloud computing, wireless sensor networks (WSN), DT, blockchain, and geographic information systems	Systematic synthesis and analysis of data-driven technologies for energy optimization in smart buildings, identifying key technological enablers, adoption barriers, and business models to support efficient, sustainable, and user-centric energy management	[93]
Data-driven methodological framework for identifying electricity consumption typologies from smart meter data	Smart meter, time-series feature extraction, unsupervised learning, clustering algorithms, statistical analysis, expert-in-the-loop validation.	Development and validation of a data-driven methodology for extracting and clustering electricity consumption typologies from large-scale smart meter data, enabling more accurate demand modeling and supporting adaptive energy planning and tariff design.	[94]
Graph-based deep learning methods for anomaly detection in Smart Grid time-series data	Graph neural networks, graph deviation networks, deep learning, and ML	Development and validation of a semi-supervised graph deviation network–based approach for real-time anomaly detection in Smart Grid time-series data, enabling robust pre-filtering of corrupted data and significantly enhancing the efficiency and reliability of state estimation	[95]
Data-driven predictive modeling of photovoltaic energy for Smart Grid systems	Time-series data, deep learning, artificial neural networks (ANNs), statistical analysis	Development and comparative evaluation of data-driven time-series models for photovoltaic (PV) energy forecasting, demonstrating high predictive accuracy and supporting reliable integration of renewable generation into Smart Grid operations	[96]
A comprehensive review of deep learning–based detection and diagnosis of short-circuit faults in power distribution networks	Deep learning, Deep reinforcement learning, AI and ML models, XAI, federated and distributed learning	Comprehensive analysis and systematization of deep learning methods for the detection, classification, and localization of short-circuit faults in power distribution networks, identifying key technological trends, performance trade-offs, and research gaps for scalable and reliable Smart Grid fault diagnostics	[97]

.

The data-driven approach enables the identification of hidden patterns and predictive relationships directly from the heterogeneous data streams of Smart Grids, without relying on explicit physical models of the system, in accordance with the following formalized description:

$$y\left(t\right)=f_{DD}\left(X\left(t\right),\mathsf{\Theta }\right)+\mathsf{\epsilon }\left(t\right),$$

(1)

where y(t)—control output; f_DD—function of data-driven mapping; X(t)—IoT data streams; $\mathsf{\Theta }$—a set of the data-driven model parameters; $\mathsf{\epsilon }\left(t\right)$—uncertainty and/or destabilizing influences.

As shown in the analysis of the scientific papers presented in Table 5, the main advantage of this approach is its capacity to adapt to changing system operating conditions without requiring explicit modeling of physical processes. Depending on the application domain, supervised learning techniques are typically employed for load forecasting and anomaly detection, unsupervised learning methods support clustering and pattern discovery, while reinforcement learning algorithms enable adaptive optimization and autonomous decision-making under dynamic operational conditions. At the same time, the approach is characterized by general functional limitations stemming from its dependence on the quality, completeness, and representativeness of input data, and, in certain cases, requires significant computational resources. In the context of an IoT-based Smart Grid architecture, the data-driven approach is primarily implemented at the data processing and application levels, where analytics and result interpretation are performed.

So, from a system-level perspective, the growing adoption of the data-driven approach reflects a broader transition from deterministic energy management to adaptive intelligence that continuously learns from operational data streams. However, the strategic value extends beyond forecasting and optimization tasks. Data-driven mechanisms increasingly function as operational intelligence layers that transform heterogeneous Smart Grid observations into actionable knowledge, thereby enabling dynamic situational awareness and proactive decision support across multiple infrastructural levels.

4.3. Model-Driven Approach

This approach is based on physically grounded mathematical and computer models of energy systems that describe their dynamic and static behavior. Key tools include models for the distribution of energy flows, the assessment of energy system stability, and the representation of electromechanical and electromagnetic processes in networks. In the context of Smart Grids, this approach provides a formalized description of the interaction between electricity generation, transmission, storage, and consumption. Currently, a well-established implementation of the model-driven approach is DTs [98,99,100], which enable the creation of dynamic digital replicas of energy systems, allowing for accurate and adaptive analysis and optimization of operating modes in real time, thereby enhancing the reliability and efficiency of Smart Grids. Technically, DTs are commonly implemented through continuous bidirectional data exchange mechanisms linking physical assets, IoT sensing infrastructures, simulation engines, and analytical modules. Such solutions support state estimation, predictive maintenance, scenario analysis, and adaptive control by maintaining an up-to-date virtual representation of physical system behavior. From a system-level perspective, DTs can be viewed as an intermediary orchestration layer that bridges data-driven and model-driven intelligence. This capability becomes increasingly important in Smart Grid environments characterized by high uncertainty, distributed energy resources, renewable energy variability, and dynamic operational conditions. Consequently, DTs are evolving from standalone simulation tools toward foundational components of intelligent cyber–physical energy ecosystems capable of supporting adaptive, resilient, and autonomous grid operation. MPC technology has also gained significant traction in the development of model-based control techniques, enabling the optimization of control actions while accounting for system constraints and predicted operating modes [101,102,103].

The results of the analysis of relevant scientific studies on the development and application of the model-driven approach in energy and electrical systems are presented in

Table 6. Results of the analysis of scientific research into the model-driven approach within the context of energy and electrical systems.

Subject of the Research	Technologies and Approaches Used	Scientific and Practical Outcomes Achieved	References
A goal-oriented and model-driven framework for conceptual design and system-level specification of Smart Grid services	System-of-systems, Petri nets, linear temporal logic, service-oriented design	Development of a goal-oriented and model-based methodology for designing flexible and user-centric Smart Grid services, enabling effective integration of distributed cogeneration and adaptive energy management in heterogeneous and isolated energy systems	[104]
Design and implementation of a programmable DT framework based on IEC 61850 communication for Smart Grid systems	DT, IEC 61850, information and communication protocols GOOSE, MMS, embedded platform Raspberry Pi, open-source stack libIEC61850	Development and experimental validation of an IEC 61850-compliant programmable platform for DT applications, enabling high-speed, interoperable, and bidirectional real-time interaction between physical and virtual Smart Grid components	[105]
A DT–based modeling framework for intelligent V2G systems with AI-driven energy management	DT, physics-based modeling, AI, ML, simulation platforms, V2G energy management algorithms, Kalman filtering	Development and validation of a high-fidelity multi-physics electric vehicle model for DT–enabled V2G systems, enabling highly accurate energy state prediction and supporting AI-driven optimization of energy management in Smart Grid environments	[106]
Framework design and implementation of DT systems for Smart Grid-integrated EV	DT, modular multi-agent-based architecture, Kalman filtering, V2G, predictive analytics	Development of a DT–based integrated framework for EV and charging infrastructure coordination, enabling decentralized decision-making and incentive-driven participation in grid services to enhance operational efficiency and grid stability	[107]
Theoretical development and case study of an extended MPC framework for energy management in renewable-based smart microgrids with hydrogen backup systems	MPC, state-space and nonlinear microgrid models, multi-criteria optimization, AI-based tuning	Development and experimental validation of an extended MPC-based energy management framework for renewable microgrids with hydrogen backup, enabling multi-objective optimization of energy distribution while accounting for system dynamics, degradation, and operational constraints	[108]
Methodological approaches to enhancing the energy efficiency of operating regimes in distribution networks with integrated PV generation	Spectral analysis, correlation analysis, mathematical and physical modeling, Mdaq-14, LabVIEW	Experimental identification of harmonic distortion patterns in photovoltaic inverters and development of a control method to reduce electromagnetic interference, enabling improved electromagnetic compatibility and enhanced energy efficiency of distribution networks with PV integration	[109]
Computer models for predictive energy-efficient control with multiparameter optimization of electromechanical systems	MPC, Matlab & Simulink R2025, GRAMPC	Development and validation of a predictive control–based optimization methodology for induction motor drives, enabling improved energy efficiency, control accuracy, and computational performance through multi-criteria tuning of MPC algorithms	[110,111]
Development and application of MPC strategies for enhancing the resilience of electric power grids	MPC, reduced-order modeling, multi-objective functions, collocation methods, orthogonal spline collocation, Bernstein polynomials, numerical simulation	Development and validation of a generalized MPC-based control framework for power systems, enabling resilient integration and optimal management of intermittent energy resources through reduced-order modeling and advanced optimization techniques	[112]
Development of a hierarchical two-layer MPC–supervisory framework for efficient operation of inverter-dominated small-scale microgrids	Hierarchical MPC, droop and PI-based control, d–q transformation, optimization algorithms, DER integrative approaches	Development and validation of a hierarchical two-layer MPC-based control framework for inverter-dominated microgrids, enabling optimized power sharing, enhanced stability, and improved integration of RES under dynamic operating conditions	[113]
Development of an MPC–based energy management framework for cooperative optimization of connected microgrids	MPC, distributed and cooperative optimization, multi-microgrid coordination, state-space modeling, energy management systems (EMS), DER integrative approaches	Development and validation of an MPC-based energy management system for cooperative microgrid operation, enabling cost-efficient and stable optimization of distributed energy resources through predictive coordination and network-level interaction	[114]

.

The model-based approach captures the behavior of Smart Grids using formalized mathematical and computer models grounded in physical principles, enabling predictive control, optimization, and stability analysis in accordance with the following:

$$x_{formalized}\left(t\right)=f_{MD}\left(s\left(t\right),u\left(t\right),p,t\right),$$

(2)

where x_formalized(t)—formalised description of Smart Grid processes; f_MD—function of model-driven mapping of Smart Grid dynamics; s(t)—Smart Grid states; u(t)—control actions; p—physical parameters; t—time.

As shown in Table 6, the main advantages of the model-driven approach are the high interpretability of results and their physical validity. This is confirmed by the significant development and use of formalized models, optimization methods, and predictive control techniques. The principles for the practical implementation of digitalization approaches have also advanced considerably, particularly those based on DTs, which ensure the cyber–physical coordination of Smart Grid system components. This accounts for the considerable effectiveness of this approach in monitoring and controlling energy systems with high reliability, security, and adaptability requirements. At the same time, the model-driven approach is characterized by certain functional limitations, mainly related to the complexity of constructing adequate models for large-scale, heterogeneous systems and limited scalability in dynamic IoT-based Smart Grid infrastructures. This motivates the integration of this approach with advanced technologies to enhance the intelligence of technological processes within the framework of hybrid approaches.

Also, the evolution of the model-driven approach demonstrates a gradual shift from static engineering representations toward continuously synchronized cyber–physical system models. In this context, DTs extend the traditional role of simulation environments by enabling real-time interaction between physical assets and virtual counterparts. Consequently, model-driven intelligence increasingly serves not only as a tool for optimization and validation but also as a foundation for predictive and self-adaptive Smart Grid operation.

4.4. Knowledge-Driven Approach

In most practical applications, this approach relies on formalized expert knowledge of the subject domain, expressed through rule-based logic and ontologies. This enables modeling cause-and-effect relationships and ensures the explainability of decisions. In the context of Smart Grids, these solutions are applied in tasks such as relay protection, fault diagnosis, fault detection, and automated control of electrical equipment. The results of the analysis of relevant scientific studies on the development and application of the knowledge-driven approach in energy and electricity systems are presented in

Table 7. Results of the analysis of scientific research on the knowledge-driven approach within the context of energy and electricity systems.

Subject of the Research	Technologies and Approaches Used	Scientific and Practical Outcomes Achieved	References
Development and evaluation of rule-based energy management strategies for enhancing PV self-consumption in building energy systems	Rule-based control (if–then logic), EMS, and algorithms for optimizing PV self-consumption	Design and real-world validation of a rule-based energy management system for PV-driven buildings, enabling high self-consumption rates through IoT-based control of flexible loads while maintaining user comfort	[115]
Development of a modular rule-based energy framework for management and coordinated operation of hybrid AC/DC microgrids	Rule-based control (modular if–then logic), DER coordination, power balancing algorithms	Development and validation of a modular rule-based energy management system for hybrid microgrids, enabling adaptive optimization of diverse configurations through dynamic coordination of DER and system constraints	[116]
Development and stochastic evaluation of a rule-based energy management framework for EV charging station nanogrids	Rule-based energy management, stochastic modeling, DER integration, energy flow optimization methods	Development and validation of a rule-based energy management system for EV charging nanogrids, enabling cost-efficient and reliable operation through stochastic modeling, renewable prioritization, and forecast-aware decision-making under uncertainty	[117]
Development of rule-based data transformation frameworks for structured information processing in Smart Grids	Jena rule language, semantic web rule language, Smart Grid data models	Development of a rule-based data transformation framework using semantic rule languages, enabling interoperability between heterogeneous Smart Grid data models without reliance on custom integration solutions	[118]
Development of an ontology matching framework for semantic interoperability in next-generation Smart Grid systems	Ontology matching algorithms, semantic web technologies, and rule-based reasoning	Development and validation of an advanced ontology matching system for Smart Grids, enabling automated detection of complex semantic correspondences between heterogeneous data models to enhance interoperability of intelligent energy systems	[119]
Development of an ontology-driven energy management framework for intelligent smart home systems	Intelligent reasoning, domain ontology, decision-making, semantic web rule language	Development and implementation of an ontology-based smart home energy management system, enabling context-aware optimization of electricity consumption through semantic reasoning and achieving measurable energy savings	[120]

.

The knowledge-driven approach generates explainable decisions by leveraging semantic reasoning mechanisms that operate on formalized expert knowledge and ontological representations in accordance with the following formalized logical–semantic representation:

$$K\cup R\Rightarrow D,$$

(3)

where K—knowledge base (ontologies, expert knowledge); R—a set of reasoning rules (rule-based logic, semantic inference, ontology reasoning); D—inferred decision; $\cup$—union; $\Rightarrow$—logical implication.

As shown in the analysis of relevant scientific literature, the main advantages of the knowledge-driven approach are its high interpretability and transparency of the decision-making logic. This approach is also effective when integrating expert knowledge via rule-based mechanisms and ontologies, as confirmed by practical applications in energy management, interoperability, and the semantic integration of heterogeneous energy systems. At the same time, limitations arise from the complexity of formalizing the domain, particularly when analyzing large-scale, dynamic energy systems, as well as from insufficient adaptability to uncertain operating conditions, even when stochastic methods are used.

An important characteristic of the knowledge-driven approach is its ability to preserve semantic transparency throughout the decision-making process. Unlike purely statistical approaches, knowledge-based reasoning enables explicit representation of causal relationships, operational constraints, and expert knowledge. This capability becomes increasingly valuable in Smart Grid environments, where explainability, regulatory compliance, and trustworthy autonomous decision support are emerging as critical operational requirements.

4.5. Agent-Based Approach

This approach is based on multi-agent systems, in which autonomous software components interact to achieve global objectives through local decision-making. In practice, this approach enables decentralized, adaptive, and scalable control of energy systems and electricity networks. In Smart Grids, the agent-based approach is widely used in microgrids and prosumer-oriented systems, where individual agents represent software entities associated with consumers, generation sources, energy storage devices, and operators, coordinating balancing, electricity distribution, and participation in market mechanisms. The results of the analysis of relevant scientific studies on the development and application of the agent-based approach within power systems are presented in

Table 8. Results of the analysis of scientific research on the agent-based approach within the context of energy and electricity systems.

Subject of the Research	Technologies and Approaches Used	Scientific and Practical Outcomes Achieved	References
Systematic exploration of ontology-enabled architectures and semantic coordination mechanisms in multi-agent energy systems	Scoping review, multi-agent systems, ontology	Systematization of ontology-driven multi-agent system design in the energy domain, enabling improved interoperability, knowledge representation, and coordinated operation of complex distributed energy systems	[121]
Development of a hierarchical decentralized multi-agent architecture for sustainable energy management in Smart Grid systems	Multi-agent systems, agent coordination algorithms, energy management optimization, DER integration	Development and validation of a decentralized hierarchical multi-agent energy management framework, enabling scalable real-time coordination and multi-objective optimization of smart grids with high renewable and prosumer integration	[83]
Comprehensive analysis of agent-based modeling approaches for Smart Grids and electricity markets	Scoping review, agent-based simulation and modeling, agent-based computational approaches	Systematization of agent-based modeling and simulation approaches for smart grids, enabling comprehensive analysis of multi-agent interactions and supporting decision-making in complex energy systems and market environments	[84]
Development of an agent-based planning framework for distribution grids within a socio-technical system paradigm	Multi-agent modeling, behavioral economics, agent-based modeling of consumers, energy flow analysis, causal loop diagrams	Development and application of an agent-based planning model for distribution grids, enabling the incorporation of heterogeneous consumer behavior to improve the accuracy of load forecasting and support socio-technical energy system design	[122]
Development of a multi-agent control framework for adaptive management of Smart Grid systems	Multi-agent systems, intelligent agents, decentralized control, agent coordination algorithms, and real-time decision-making mechanisms	Development and application of a multi-agent–based simulation framework for smart grid management, enabling analysis of market interactions, coordinated use of distributed resources, and adaptive decision-making under dynamic electricity market conditions.	[123]
Development and application of an agent-based simulation environment for modeling and analysis of Smart Grid systems	Agent-based modeling, power flow modeling, load, and DER models	Development and application of an agent-based simulation platform (GridLAB-D), enabling integrated time-series modeling of power systems, markets, and distributed resources for comprehensive analysis and design of smart grid technologies	[124]
Development of an adaptive multi-agent control and optimization framework for intelligent management of Smart Grid systems	Adaptive control, stochastic learning, reinforcement learning, multi-agent systems	Development of an adaptive multi-agent reinforcement learning–based energy management approach, enabling coordinated control of DER and energy storage under renewable variability through decentralized learning and real-time interaction	[125]
Development of agent-based modeling frameworks for simulation and analysis of Smart Grid market operations	Agent-based modeling, market-based trading, decentralized management, energy market modeling	Development and evaluation of agent-based market mechanisms for Smart Grids, enabling efficient decentralized energy trading and coordination among self-interested participants under uncertainty and resource constraints	[85]
Development of an intelligent multi-agent autonomous control framework for energy management in microgrid systems	Agent-based models for real-time prediction and correction, simulation modeling	Development and validation of a multi-agent–based energy management architecture for microgrids, enabling adaptive and efficient coordination of DER under uncertainty through prediction and real-time correction mechanisms	[126]

.

The agent-based approach enables decentralized coordination through autonomous local interactions between distributed system entities in accordance with the following formalized representation:

$$A_i\left(t+1\right)=F_i\left(A_i\left(t\right),N_i\left(t\right),E\left(t\right)\right),$$

(4)

where A_i—state and/or action of an agent; F_i—logical decision policy; N_i—neighboring agents; E—state of the Smart Grid.

The main advantages of the agent-based approach are its high adaptability and fault tolerance, resulting from its decentralized architecture. This is confirmed by a wide range of practical applications. As shown in Table 8, the main application areas of this approach include microgrid energy management, DER coordination, modeling of market mechanisms and consumer behavior, and planning of grid operating modes. Furthermore, the use of semantic and ontological algorithms improves interoperability, aligns knowledge representations, and enables effective coordination among agents in heterogeneous environments. However, the main limitations of this approach include the complexity of coordinating and harmonizing agents’ actions, the risks of unstable behavior when local objectives conflict, and increased demands on the information and communication infrastructure. A current trend in the development of the agent-based approach is the integration of AI algorithms into multi-agent systems, transforming this approach into hybrid solutions with enhanced capabilities for self-adaptation, prediction, and decision-making under conditions of uncertainty and in dynamic electricity market environments.

The growing interest in agent-based architectures reflects the broader decentralization of modern energy ecosystems driven by DERs, prosumer participation, and local energy communities. In this context, intelligent agents can be viewed as fundamental building blocks of distributed operational intelligence, enabling autonomous coordination across multiple decision-making levels without relying on centralized control structures. Such capabilities are particularly important for modern Smart Grids characterized by high operational variability and geographically dispersed assets.

4.6. Hybrid-Oriented Approach

This approach integrates and combines the diverse approaches analyzed above, including AI and MPC, DTs in conjunction with ML or deep learning, physics-informed ML, and data-model fusion. This enables a more adequate, adaptive, and accurate formalization of the processes occurring in power systems, particularly in Smart Grids. Such an approach facilitates the practical integration of physics-based modeling, the predictive capabilities of AI algorithms, and digital platforms for the decentralized control of electrical and information flows. The results of the analysis of relevant scientific studies on the development and application of the hybrid-oriented approach within electricity and energy systems are presented in

Table 9. Results of the analysis of research into the hybrid-oriented approach within the context of electricity and energy systems.

Subject of the Research	Technologies and Approaches Used	Scientific and Practical Outcomes Achieved	References
Unified hybrid methodology integrating data-driven and model-driven approaches for distributed optimal control of microgrid systems	Data-driven approach, model-driven approach, optimization algorithms, ML, ICT, Python 3.11, MATLAB R2023	Development and validation of a hybrid data-driven and model-based distributed control framework, enabling accurate voltage and frequency restoration in standalone microgrids through consensus algorithms and ML–based renewable prediction	[127]
Systematic investigation of AI–driven hierarchical control architectures for advanced management of microgrid systems	Comprehensive review of ML, ANN, deep learning, reinforcement learning, fuzzy logic, meta-heuristic algorithms, MPC, and hybrid AI controllers	Systematization and comprehensive analysis of AI-driven hierarchical control strategies for microgrids, enabling improved efficiency, adaptability, and stability in the integration of distributed RES	[128]
Development of an integrated energy management framework for residential microgrids based on MPC using Shapley value and reinforcement learning	MPC, reinforcement learning, hybrid MPC–reinforcement learning approach, Shapley value, optimization algorithms, microgrid modeling	Development and validation of a hybrid MPC–reinforcement learning energy management strategy with Shapley value–based allocation, enabling cost-optimal operation of residential microgrids under uncertainty and fair distribution of cooperative economic benefits among prosumers	[129]
Comparative analysis of ML– and MPC–based approaches for optimal operation of residential battery energy storage systems	MPC, mixed-integer linear programming, nonlinear programming, supervised learning, reinforcement learning, imitation learning, neural networks, hybrid MPC–ML control approaches	Comparative evaluation of model-based MPC and ML control strategies for home energy management systems, demonstrating improved operational efficiency over rule-based methods while revealing trade-offs between optimization performance, constraint satisfaction, and computational complexity	[130]
Development of an integrated hybrid methodology combining data-driven learning and physics-based modeling for reliability assessment of distribution networks with a high proportion of renewable energy	Hybrid data-driven and model-driven approach, conditional Wasserstein generative adversarial network, clustering algorithms, mixed-integer linear programming, Monte Carlo simulation	Development and validation of a hybrid data-driven and model-based reliability assessment framework, enabling accurate quantification of distribution network reliability under renewable uncertainty through advanced scenario generation and optimization-based fault analysis	[131]
Development of a unified hybrid framework for accelerated power system state estimation via integration of model-based and data-driven techniques	Hybrid data-driven and model-driven approach, state estimation, physically grounded network models, deep learning, optimization algorithms, SCADA, accelerated computing methods	Development and validation of a hybrid model-driven and graph neural network–based state estimation framework, enabling fast and accurate monitoring of power system dynamics by combining physical topology awareness with data-driven feature learning	[132]
Development of an advanced MPC framework augmented by ANNs for optimized high-rate charging of lithium-ion battery systems	MPC, ANNs, models of lithium-ion battery systems	Development and validation of an ANN-augmented MPC framework for lithium-ion battery charging, enabling MPC-level predictive performance with drastically reduced computational complexity to support real-time control	[133]
Development of a data-driven multi-agent reinforcement learning–based approach for home energy management	Decision making, reinforcement learning, optimization, neural network, finite Markov decision process, Q-learning algorithm	Development and validation of a multi-agent reinforcement learning–based home energy management framework, enabling cost-efficient and user-aware demand response optimization under uncertainty through data-driven forecasting and adaptive scheduling	[134]
Development of an ontology-driven framework for integrated network model management and semantic data interoperability in Smart Grid systems	Unified modeling language; time series analysis; semantics; ontologies; common information model; data models	Development and validation of an ontology-based data management framework extending common information model semantics, enabling efficient integration, semantic interoperability, and advanced querying of heterogeneous Smart Grid data	[135]
Development of a data-driven and retrieval-augmented generation (RAG)–enabled framework for digital energy infrastructure systems	Decision-making, RAG, DT, natural user interface, API-based services, ML	Development and implementation of RAG and knowledge graph–enhanced DT framework, enabling advanced predictive analytics and intelligent decision support for energy infrastructure management through integration of ML and LLMs	[136]
Development of an integrated deep learning–enhanced MPC framework for advanced frequency regulation in wind farm systems	Hybrid deep learning and MPC, bi-level control architecture, deep neural networks, particle swarm optimization, physics-based modeling, MATLAB and Simulink	Development and validation of a hybrid deep learning–MPC framework for wind farm control, enabling improved frequency stability and dynamic response through coordinated data-driven inertia estimation and constrained optimal power allocation	[137]

and graphically interpreted in Figure 8.

The hybrid-oriented approach integrates complementary intelligent paradigms to simultaneously achieve adaptability, explainability, predictive capability, and decentralized coordination of the Smart Grid in accordance with the following formalized representation:

$$H\left(t\right)={\mathsf{\alpha }}_1\cdot k_{DD}\cup {\mathsf{\alpha }}_2\cdot k_{MD}\cup {\mathsf{\alpha }}_3\cdot k_{KD}\cup {\mathsf{\alpha }}_4\cdot k_{AB},$$

(5)

where H(t)—the integrated hybrid intelligence state of the Smart Grid ${\mathsf{\alpha }}_1$, ${\mathsf{\alpha }}_2$, ${\mathsf{\alpha }}_3$, ${\mathsf{\alpha }}_4$—adaptive contribution weights; k_DD—data-driven component; k_MD—model-driven component; k_KD—knowledge-driven component; k_AB—agent-based component; $\cup$—conceptual union (orchestration function).

The practical implementation of hybrid-oriented intelligence commonly involves multi-layer orchestration frameworks that integrate ML models, physical simulation engines, semantic reasoning systems, and autonomous agents via standardized communication interfaces. Such integration enables the coordinated exchange of operational knowledge, predictive insights, and control actions across heterogeneous analytical modules. Hybrid architectures thus function as distributed intelligence ecosystems rather than isolated decision-support mechanisms.

As shown in the analysis presented in Table 9 and Figure 8, the hybrid approach is the dominant theoretical and practical concept for the digitalization and intelligentization of modern Smart Grid systems. The growing prevalence of hybrid-oriented architectures demonstrates an ongoing transition from isolated analytical solutions toward integrated cyber–physical intelligence foundations capable of supporting adaptive and distributed Smart Grid operation. Hybrid AI–MPC solutions, in particular, have gained significant popularity, as has the combined application of data-driven and model-driven approaches integrated with ML and deep learning algorithms, ontologies, and GenAI models.

The integration of MPC with AI-based analytical mechanisms reflects a broader shift toward predictive, context-aware energy orchestration, in which physical system constraints are dynamically coordinated with adaptive, data-centric intelligence. The convergence of DT, semantic knowledge representations, and RAG mechanisms enables the creation of continuously adaptive cyber–physical environments characterized by improved situational awareness, explainability, and operational responsiveness. At the same time, the application of physics-informed and hybrid ML approaches helps align models with physical laws and reduce dependence on large data volumes. Consequently, hybrid-oriented paradigms emerge not merely as combinations of heterogeneous analytical methods, but as integrative orchestration mechanisms capable of simultaneously ensuring adaptability, resilience, explainability, and operational scalability in decentralized Smart Grid ecosystems. Furthermore, this approach demonstrates high effectiveness in conditions of incomplete, noisy, and/or heterogeneous data, as well as in tasks involving the coordination of DER and prosumer-oriented systems.

Nevertheless, the transition toward hybrid-oriented intelligent ecosystems introduces substantial architectural and operational complexity associated with the interoperability of heterogeneous analytical paradigms, synchronization of distributed computational processes, semantic consistency of data representations, and coordination between physical and cyber components operating across multiple Smart Grid layers. These challenges indicate that future Smart Grid evolution will increasingly depend not only on the development of individual intelligent methods but also on the establishment of unified orchestration frameworks capable of integrating heterogeneous IoT infrastructures, distributed intelligence mechanisms, and adaptive decision-making models within coherent cyber–physical operational environments.

Consequently, hybrid intelligence should be viewed not merely as a combination of existing methods, but as a distinct architectural paradigm aimed at achieving interoperability among heterogeneous intelligent capabilities operating across complex cyber–physical energy infrastructures.

4.7. Generalization and Applicability of the Studied System-Level Approaches in Smart Grids

In the context of the development of IoT-based Smart Grid infrastructure, several practice-oriented approaches to the monitoring and control of decentralized energy systems have emerged. Each of these approaches is based on its own methodological framework and set of tools. In particular, data-driven, model-driven, knowledge-driven, and agent-based approaches implement the logic of a conceptual transition from data analysis to knowledge formalization and decentralized decision-making. Although each intelligent paradigm demonstrates strong applicability within specific Smart Grid operational scenarios, none of the investigated approaches independently satisfies the increasing requirements for interoperability, scalability, explainability, decentralized coordination, and adaptive resilience characterizing future cyber–physical energy systems. A generalized structural diagram of these approaches, their technological content, the tasks they address, and the characteristic issues is shown in Figure 9.

As shown in Figure 9, the system-level approaches considered do not provide a universal solution for all aspects of Smart Grid management, necessitating the development of methods for their integration and combined application. In this context, the hybrid-oriented approach represents a logical stage of development, enabling the systematic combination of the key technical and functional advantages of various approaches. This enables the implementation, within a unified IoT software-hardware architecture, of full-cycle processing of spatiotemporal data characterizing various operational states of the Smart Grid, particularly in a predictive mode that supports decision-making. Such integration improves the accuracy and reliability of system status assessment, the efficiency of energy flow optimization, and the adaptability of control under conditions of uncertainty, while placing increased demands on the computational, information, and communication resources of the IoT-based Smart Grid infrastructure.

Despite its significant potential, hybrid-oriented Smart Grid architectures are not without limitations. The integration of heterogeneous intelligent paradigms substantially increases architectural complexity and system design requirements. Such environments often require sophisticated orchestration mechanisms that synchronize heterogeneous data models, decision-making processes, communication protocols, and control objectives across multiple infrastructure layers. Furthermore, the deployment of hybrid intelligence introduces additional challenges, including interoperability, computational overhead, model consistency, and lifecycle management. Maintaining coherence between continuously evolving AI models, DTs, knowledge bases, and distributed control agents remains a nontrivial task, particularly in large-scale Smart Grid ecosystems characterized by dynamic operational conditions and high data heterogeneity. Consequently, although the hybrid-oriented approach demonstrates superior adaptability and functional completeness, its practical implementation requires advanced integration frameworks, standardized interfaces, and scalable orchestration mechanisms to fully realize its expected benefits.

To provide a system-level comparative interpretation of intelligent Smart Grid paradigms, Figure 10 presents a challenge-oriented mapping between key Smart Grid operational problems, the most suitable intelligent approaches, their associated technological limitations, and the resulting functional outcomes. The proposed synthesis highlights the complementary capabilities of data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented approaches within modern IoT-based Smart Grid organizations. Furthermore, the proposed diagram is complemented by a comparative paradigm assessment that summarizes the relative strengths and limitations of the reviewed intelligent approaches across several informative operational criteria.

While the reviewed approaches demonstrate substantial potential to improve Smart Grid intelligence, none individually addresses the full spectrum of operational challenges associated with modern energy ecosystems. Therefore, a critical assessment of their applicability, limitations, and complementary capabilities is required to support the development of hybrid-oriented, intelligent infrastructures.

Figure 10 visualizes the relationships between major Smart Grid challenges, the intelligent paradigms most frequently associated with addressing these challenges in the reviewed literature, and the corresponding operational outcomes. The visualization is intended to highlight conceptual associations and complementarities among approaches rather than quantitative performance differences.

As illustrated in Figure 10, different Smart Grid challenges require distinct intelligent paradigms characterized by specific functional strengths and architectural constraints. The comparative synthesis demonstrates that a hybrid-oriented intelligent approach offers the greatest potential for integrating adaptability, explainability, interoperability, decentralized coordination, and real-time decision-making capabilities required for future scalable Smart Grid infrastructures.

5. Primary Challenges Facing Smart Grids and Development Areas of IoT-Based Infrastructure for Their Mitigation

5.1. System-Level Operational Challenges and IoT-Enabled Mitigation Frameworks in Smart Grids

Modern Smart Grids are characterized by high complexity, dynamism, and the integration of a large number of DER, which pose practical challenges for the implementation of intelligent control of energy flows and processes [138,139,140]. Despite the practice-oriented development of IoT architectures and modern control approaches, several fundamental challenges remain to be addressed. These include the uncertainty and stochasticity of energy processes, data heterogeneity, and coordination and cybersecurity issues. In this context, the development of IoT infrastructure is particularly significant as one of the key technological factors capable of mitigating the impact of these issues and ensuring more reliable and efficient operation of Smart Grid systems. The proposed layered diagram of the classification of the primary challenges facing Smart Grids and development areas of IoT-based infrastructure for their mitigation is shown in Figure 11. The proposed diagram is organized according to a layered analytical structure. The left side summarizes the primary operational and technological challenges of modern Smart Grids, while the right side groups the corresponding IoT-enabled development areas that can mitigate these challenges. The diagram emphasizes systemic interdependencies rather than one-to-one technological solutions.

5.2. Uncertainty and Stochasticity in Electrical and Energy Processes

One of the most pressing challenges facing Smart Grids is the high degree of uncertainty caused by the stochastic nature of RES. This leads to significant short-term fluctuations in generation. An additional source of uncertainty is introduced by the complex and unpredictable nature of demand, linked to prosumer behavior and the integration of EVs into the grid, and characterized by both temporal and spatial variability. Furthermore, market dynamics directly influence electricity generation and consumption profiles, complicating the tasks of prediction, balancing, and optimal energy system management.

To mitigate uncertainty and stochasticity, the IoT infrastructure of the Smart Grid should evolve towards greater adaptability, scalability, and measurement precision, as well as the implementation of intelligent data processing at the network edge. The development of real-time prediction and analytical services, the integration of streaming data from multiple sources, and the use of standardized protocols to ensure interoperability are critical. This approach enables improved prediction accuracy, reduced decision-making latency, and more stable and adaptive energy system management.

5.3. Challenges in the Integration and Control of V2G

The integration of EVs into Smart Grids, based on the V2G concept, introduces additional control complexity due to bidirectional energy flows between vehicles and the grid. A key technical factor requiring continuous monitoring is battery degradation. Another key challenge in this context is the coordination of a large number of EVs while accounting for time constraints, tariffs, and user behavior. Furthermore, uncontrolled or unsynchronized charging processes can lead to overloads, voltage fluctuations, and reduced power grid stability.

To mitigate these destabilizing factors, the IoT-based Smart Grid infrastructure should enable intelligent control of charging infrastructure by integrating smart charging stations, bidirectional meters, and real-time battery monitoring systems. The use of edge computing technologies enables local charging coordination that accounts for grid conditions and user behavior, thereby helping to reduce peak loads. It is important to implement EV–grid communication standards and data aggregation platforms to ensure coordinated control of a large number of EVs as a synchronized, flexible resource, thereby increasing the stability and efficiency of the power system.

5.4. Data Interoperability and Heterogeneity

One of the main challenges to the efficient operation of computing, information, and communication processes in Smart Grids is data heterogeneity. This challenge stems from the use of diverse information and communication standards and protocols, as well as the variety of data and information sources. The lack of a unified standardized data representation format and semantic incompatibility between models leads to information loss, complicates system integration, and limits the possibilities for automated real-time decision-making.

To overcome these limitations, the IoT infrastructure should evolve toward the standardization and semantic unification of data through the use of ontologies and extended common information models. The implementation of middleware solutions and API-oriented platforms, along with support for open data exchange protocols, enables the integration of distributed data sources into a unified information environment, thereby enhancing interoperability and improving the efficiency of energy system management.

5.5. Computational Load and Scalability

As the Smart Grid expands, the number of IoT devices increases significantly, necessitating the processing of large volumes of data in real-time. Systems implementing a hybrid approach that integrates AI, MPC, and DT pose particular challenges, as they require the simultaneous execution of prediction, simulation, control, and optimization procedures. This creates a significant computational load and complicates the system’s real-time response to events.

To mitigate these limitations, the IoT infrastructure should evolve toward distributed and parallel computing, utilizing approaches that efficiently distribute computational tasks across the edge, fog, and cloud layers. This allows part of the processing to be moved closer to the data sources, reducing data queues and latency. It is important to implement effective adaptive hybrid solutions that balance accuracy and performance.

5.6. Coordination in Decentralized and Multi-Agent Systems

Decentralized and multi-agent approaches in Smart Grids provide flexibility and scalability; however, they also increase the complexity of agent coordination. In particular, conflicts arise between the local objectives of individual agents and the global objectives of the system, leading to inefficient or unstable operating modes. Additional constraints include communication delays and limited network bandwidth, which complicate real-time synchronization of actions.

To mitigate these issues, the IoT infrastructure should support reliable, low-latency communication channels and provide mechanisms for decision-making coordination, such as coordination protocols, consensus algorithms, and hierarchical control schemes. The large-scale deployment of edge computing technologies enables local autonomy of agents alongside global coordination, while the implementation of standards for interaction and data exchange facilitates decision-making and enhances the operational efficiency of decentralized energy systems.

5.7. Integration of GenAI

The integration of GenAI models into Smart Grids opens a wide range of possibilities for the development and implementation of highly efficient decision-support systems, particularly in analytics, prediction, and user interaction. At the same time, such systems have several limitations, including the risk of producing incorrect or unverified decisions, which is critical in energy applications. An additional challenge is the complexity of integrating LLMs with physically grounded models without conducting thorough theoretical and applied validation, which can lead to misalignment with real-world processes.

To mitigate these risks, the IoT infrastructure should support hybrid architectures that complement LLMs with proven physical models and decision-validation mechanisms. This will ensure a more reliable and effective deployment of GenAI in Smart Grids.

5.8. Cybersecurity and Privacy

An important challenge that emerges across all reviewed intelligent paradigms is cybersecurity. The increasing integration of IoT devices, distributed energy resources, edge computing platforms, AI-driven decision-making systems, and cloud-based analytical services significantly expands the cyberattack surface of modern Smart Grids. Unlike conventional power systems, intelligent Smart Grids must simultaneously protect physical infrastructure, communication networks, data streams, control mechanisms, and autonomous decision-making processes. Consequently, cybersecurity can no longer be treated as an isolated technical component but should be considered a fundamental architectural requirement spanning all layers of Smart Grid intelligence. This challenge becomes particularly critical in hybrid-oriented ecosystems, where vulnerabilities may propagate across interconnected AI models, DTs, distributed agents, and IoT infrastructures, creating new forms of systemic cyber–physical risk.

To minimize these risks, the IoT infrastructure should adopt a comprehensive information security approach, including data encryption, device authentication, secure communication protocols, and mechanisms for real-time detection and mitigation of attacks. It is important to implement data protection and user privacy measures, particularly in prosumer-oriented systems, and to apply zero-trust principles and distributed security mechanisms, thereby enhancing Smart Grid resilience to modern cyber threats.

5.9. Practical Implementation, Harmonization and Standardization of Solutions

Given the rapid development of intelligent approaches to monitoring and controlling energy generation processes and facilities, their practical implementation in Smart Grids is hindered by the lack of standardized solutions and unified standards. This leads to fragmentation of technologies and limited interoperability between systems. An additional problem is the integration of new solutions with existing infrastructure, which often has limited modernization potential.

To overcome these limitations, the IoT infrastructure should evolve toward standardized interfaces and data exchange protocols, while ensuring compatibility with existing systems, particularly through the use of middleware technologies. It is also important to take into account economic and regulatory factors, including the promotion of investment, the development of a regulatory framework, and the adoption of open standards, which will facilitate the large-scale and effective implementation of solutions for the intelligentization and digitalization of electricity systems.

6. Discussions and Suggestions for Future Research

6.1. Analysis of Results in the Context of the Review Motivation and Objectives

The obtained results indicate that contemporary Smart Grid environments increasingly rely on coordinated interactions among sensing infrastructures, communication networks, distributed computing environments, intelligent analytics, and decentralized control mechanisms, which encourage a broader system-level interpretation of their operation and evolution.

In this context, the results directly support the primary aim of this review, which was to establish a unified system-level perspective integrating IoT infrastructures with intelligent monitoring, prediction, optimization, and orchestration mechanisms. The conducted synthesis demonstrates that the hybrid-oriented intelligent approach provides the most effective conceptual foundation for future IoT-enabled Smart Grid development, as it enables the coordinated integration of AI, ML, DT, semantic reasoning, MPC, and decentralized multi-agent coordination within unified cyber–physical infrastructures.

Another important outcome of the study is the developed challenge-to-approach mapping framework, which establishes relationships among Smart Grid operational problems, architectural layers, enabling technologies, and intelligent analytical paradigms. Unlike fragmented analyses focused on individual technologies, the proposed framework enables a more systematic interpretation of interoperability constraints, scalability issues, cybersecurity challenges, and adaptive orchestration requirements within decentralized energy systems.

Beyond the conceptual and comparative insights obtained in this review, the analyzed intelligent paradigms have already demonstrated practical applicability across multiple Smart Grid domains. Data-driven approaches are widely employed for demand forecasting, predictive maintenance, and anomaly detection, while model-driven solutions and DTs increasingly support operational optimization and infrastructure monitoring. Agent-based architectures have shown effectiveness in distributed energy resource coordination and microgrid management, whereas knowledge-driven mechanisms facilitate semantic interoperability across heterogeneous energy environments. Importantly, both industrial deployments and pilot projects increasingly rely on the integration of multiple intelligent mechanisms rather than isolated analytical approaches, providing practical evidence for the growing relevance of hybrid-oriented Smart Grid architectures.

Furthermore, the obtained results highlight the growing importance of edge/fog computing, distributed intelligence, XAI, V2G interaction, and semantic interoperability as enabling mechanisms for the next generation of adaptive Smart Grid ecosystems. The study also demonstrates that future intelligent energy infrastructures will increasingly rely on integrated hybrid intelligence ecosystems rather than on the dominance of a single analytical paradigm.

6.2. Analysis of Results in the Context of Existing Review Studies

The results generally align with recent review studies that emphasize the growing importance of AI, IoT, and decentralized energy management in Smart Grid systems. Existing reviews, such as [1], primarily focus on intelligent management and optimization strategies for smart microgrids, highlighting the importance of AI-driven control and distributed energy resource coordination. Similarly, the study [2] investigates IoT-based smart microgrid management, with an emphasis on energy monitoring and operational efficiency. However, these studies make valuable and high-quality contributions to the field, particularly in intelligent energy management, optimization, and IoT-based Smart Grid applications. At the same time, their primary focus remains centered on specific technological domains and application scenarios rather than on an integrated system-level interpretation of Smart Grid operational architectures.

In comparison with these studies, the present review extends the analytical perspective by simultaneously integrating architectural decomposition, intelligent operational paradigms, interoperability analysis, and challenge-oriented synthesis within a unified conceptual framework. Unlike conventional technology-centric surveys, this article establishes explicit relationships between IoT architectural layers and higher-level intelligent control mechanisms, including data-driven, model-driven, knowledge-driven, agent-based, and hybrid-oriented approaches.

A number of recent research articles and reviews also investigate hybrid intelligent control strategies and AI-based optimization methods for Smart Grid and microgrid applications. For example, studies discussed in the context of hybrid AI–MPC and reinforcement learning approaches [127,128,129,130] demonstrate the growing practical relevance of integrating predictive analytics with physically grounded control mechanisms. Similarly, recent investigations into hybrid data-driven and model-driven frameworks [131,132,133,134,135,136,137] confirm the increasing transition toward integrated cyber–physical intelligence ecosystems.

Nevertheless, most existing review studies remain focused either on specific analytical technologies, such as ML, DT, or MPC, or on particular combinations of these approaches for solving applied Smart Grid problems, including load forecasting, energy optimization, or microgrid coordination. In contrast, the present study emphasizes the systemic interdependencies among architectural layers, operational challenges, distributed intelligence mechanisms, and orchestration processes across the entire IoT-enabled Smart Grid infrastructure.

Another important distinction of this review lies in the proposed challenge-oriented mapping between operational Smart Grid problems and corresponding intelligent paradigms. While previous reviews generally classify technologies by application domain or algorithmic characteristics, the present study introduces a functional, system-level interpretation that links operational requirements, architectural constraints, and intelligent orchestration mechanisms. This enables a more comprehensive understanding of how different analytical approaches complement one another within adaptive, decentralized Smart Grid ecosystems.

Furthermore, unlike many existing reviews that analyze distributed energy resources, renewable energy integration, or electric vehicles mainly as isolated technological components, this work treats them as dynamic entities that influence both energy and information flows within IoT-enabled cyber–physical environments. Such an interpretation broadens the conceptual understanding of Smart Grid intelligence and strengthens the case for transitioning to interoperable, hybrid-oriented ecosystems.

6.3. Limitations and Future Research Directions

Despite the integrative and system-level nature of the conducted review, several limitations should be acknowledged. First, the proposed synthesis is primarily conceptual and analytical in nature. Although the study systematically compares intelligent paradigms and architectural approaches and generalizes findings from the reviewed literature, it does not focus on the experimental evaluation or quantitative benchmarking of specific Smart Grid implementations. Therefore, the presented results should be interpreted mainly as an integrative methodological and conceptual framework intended to support the analysis, classification, and further development of intelligent IoT-enabled Smart Grid ecosystems.

Second, the analyzed approaches were primarily evaluated from the perspectives of their operational logic, architectural interoperability, and functional capabilities in the context of IoT-enabled Smart Grid ecosystems. As the study was oriented toward a system-level comparative synthesis, detailed technological aspects related to specific hardware platforms, communication standards, regional energy infrastructures, and regulatory environments were considered only to the extent necessary for the conceptual analysis. Consequently, the practical implementation and applicability of certain discussed approaches may depend on the characteristics of particular Smart Grid deployments, including infrastructure scale, digitalization maturity, and local operational conditions.

Another limitation is the rapid evolution of enabling technologies, particularly in AI, GenAI, DT, edge intelligence and distributed cyber–physical systems. As a result, some technological trends and implementation strategies may evolve considerably in the near future, requiring continuous refinement of existing methodological frameworks.

Although direct quantitative benchmarking across the reviewed studies remains challenging due to differences in datasets, operational objectives, evaluation metrics, and Smart Grid application domains, a comparative synthesis of the literature enables a semi-quantitative assessment of the principal intelligent paradigms. The analysis indicates that data-driven approaches demonstrate superior adaptability and scalability; model-driven approaches provide strong explainability and physical consistency; knowledge-driven systems offer semantic transparency; agent-based architectures support decentralized coordination; and hybrid-oriented paradigms exhibit the highest overall capability for addressing multidimensional Smart Grid challenges.

Future research should therefore focus on developing comprehensive hybrid orchestration frameworks that integrate data-driven analytics, physics-informed modeling, semantic reasoning, and decentralized multi-agent coordination within unified IoT infrastructures. Particular attention should be devoted to explainable and physics-informed AI approaches, which can improve the transparency, reliability, and operational safety of intelligent energy systems.

More specifically, several research challenges remain insufficiently addressed and require targeted investigation. These include the development of standardized interoperability frameworks for coordinating heterogeneous intelligent paradigms, formal methods for the dynamic orchestration of AI, DTs, and multi-agent systems, and scalable mechanisms for real-time semantic integration across distributed Smart Grid infrastructures. Furthermore, there is a need for unified evaluation methodologies capable of quantitatively assessing the performance, explainability, resilience, and operational efficiency of hybrid intelligent ecosystems under realistic Smart Grid conditions. Another important research direction concerns the establishment of benchmark platforms and digital testbeds that would enable reproducible validation of integrated IoT-enabled Smart Grid solutions across different deployment scenarios.

Another promising direction concerns the integration of electric vehicles, distributed energy resources, and prosumer-oriented infrastructures into adaptive Smart Grid ecosystems. Further studies are required to formalize the influence of these entities on information flows, distributed decision-making processes, and real-time energy orchestration.

In addition, future investigations should address interoperability and cybersecurity challenges associated with large-scale, decentralised Smart Grid operation, including semantic integration, standardised communication protocols, distributed trust architectures, and resilient edge/fog intelligence mechanisms. The integration of LLMs, RAG, and advanced DT ecosystems also represents a promising research direction for intelligent decision support and adaptive cyber–physical coordination.

From a practical perspective, future studies should investigate how emerging intelligent technologies can be operationalized within real-world energy infrastructures. Priority areas include the deployment of physics-informed and explainable AI for critical decision-making, the integration of DTs for predictive grid management, the use of LLMs for operator support and knowledge management, and the development of autonomous edge intelligence capable of supporting self-adaptive energy orchestration. Particular attention should also be devoted to validating these technologies through large-scale pilot implementations involving DERs, RES integration, EVs, and prosumer-centric energy communities.

An important observation emerging from this review is that the superiority of hybrid-oriented paradigms does not imply the elimination of technological and organizational barriers. On the contrary, the increasing convergence of AI, IoT, DTs, semantic technologies, and distributed control mechanisms introduces new integration challenges that may become a critical bottleneck for future Smart Grid deployment. Therefore, future research should focus not only on improving individual intelligent components but also on developing interoperable orchestration frameworks capable of managing the growing complexity of hybrid cyber–physical energy ecosystems.

Overall, the results indicate that the future evolution of Smart Grids will depend not on the dominance of a single analytical paradigm but on the ability to establish adaptive, interoperable, scalable, and explainable hybrid intelligence ecosystems that integrate heterogeneous technologies and decentralized orchestration mechanisms within unified IoT-enabled energy infrastructures.

7. Conclusions and Future Research Perspectives

This study presents a system-level integrative review and comparative synthesis of intelligent operational and architectural foundations of IoT-enabled Smart Grids. Unlike conventional technology-centric surveys, the conducted analysis establishes a unified analytical perspective that links IoT architectural infrastructures with higher-level intelligent monitoring, prediction, optimization, and control paradigms operating within modern cyber–physical energy systems.

The results demonstrate that the effectiveness of future intelligent energy systems increasingly depends on coordinated interactions among sensing infrastructures, communication mechanisms, distributed computational environments, analytical models, and decentralized decision-making architectures. In this context, the study substantiates the need to transition from fragmented technological implementations to integrated, hybrid-oriented intelligent ecosystems.

The cross-layer architectural decomposition and challenge-oriented synthesis performed in this study revealed that each intelligent paradigm considered in this study possesses both significant functional advantages and important operational limitations. Data-driven approach provides adaptive predictive capabilities and efficient processing of large-scale spatiotemporal data, but often suffers from limited explainability and insufficient physical interpretability. Model-driven approach ensures physically grounded optimization and operational reliability. However, it remains computationally expensive and difficult to scale within highly distributed environments. Knowledge-driven approach improves semantic interoperability and explainable reasoning but depends on the formalization and maintenance of domain knowledge. Agent-based approach facilitates decentralized coordination and autonomous decision-making but introduces synchronization and orchestration challenges in heterogeneous Smart Grid ecosystems.

Based on the performed challenge-oriented mapping and architectural decomposition, the study demonstrates that hybrid-oriented intelligent paradigms represent the most prospective solution for future Smart Grid development. Such an approach enables the integration of predictive analytics, semantic reasoning, physical system modeling, and autonomous, decentralized coordination within unified IoT-enabled cyber–physical infrastructures. This creates the foundation for adaptive, scalable, interoperable, explainable, and resilient Smart Grid ecosystems capable of operating under conditions of uncertainty, stochasticity, and high operational complexity.

At the same time, the conducted review indicates that the large-scale deployment of intelligent Smart Grid technologies remains associated with several unresolved challenges. Among the most critical limitations are interoperability barriers between heterogeneous platforms and communication protocols, cybersecurity and privacy risks arising from highly interconnected IoT infrastructures, data quality and semantic consistency issues, computational and communication overhead in distributed environments, and the growing complexity of coordinating multiple intelligent components operating across different architectural layers. These challenges highlight that future Smart Grid evolution requires not only advances in individual technologies but also the development of unified integration and orchestration mechanisms capable of ensuring reliable system-wide operation.

Furthermore, the conducted synthesis provides a generalized methodological perspective for understanding the development of future intelligent energy systems integrating artificial intelligence, digital twins, edge/fog computing, big data analytics, vehicle-to-grid interaction, distributed energy resources, and decentralized orchestration mechanisms. The presented results provide analytical insights that may support the design of next-generation Smart Grid architectures characterized by real-time responsiveness, adaptive optimization, cybersecurity resilience, and cross-layer interoperability.

To sum up, the results indicate that future Smart Grid evolution will increasingly depend on the establishment of interoperable, intelligent orchestration ecosystems that integrate distributed analytical paradigms, adaptive cyber–physical coordination, semantic interoperability, and decentralized operational intelligence within unified IoT-enabled infrastructures. However, realizing this vision will require overcoming persistent challenges related to interoperability, scalability, cybersecurity, explainability, and the integration of heterogeneous intelligent technologies within complex cyber–physical energy environments.

Future research directions should focus on the development of comprehensive hybrid orchestration frameworks, explainable and physics-informed AI models, distributed multi-agent coordination mechanisms, interoperable semantic architectures, and resource-efficient edge/fog intelligence solutions capable of supporting large-scale, decentralised Smart Grid environments.