A Review of Multi-Microgrids Operation and Control from a Cyber-Physical Systems Perspective

Abstract

Developing multi-microgrid (MMG) systems provides a new paradigm for power distribution systems with a higher degree of resilience, flexibility, and sustainability. The inclusion of communication networks as part of MMG is critical for coordinating distributed energy resources (DERs) in real time and deploying energy management systems (EMS) efficiently. However, the communication quality of service (QoS) parameters such as latency, jitter, packet loss, and throughput play an essential role in MMG control and stability, especially in highly dynamic and high-traffic situations. This paper presents a focused review of MMG systems from a cyber-physical viewpoint, particularly concerning the challenges and implications of communication network performance of energy management. The literature on MMG systems includes control strategies, models of communication infrastructure, cybersecurity challenges, and co-simulation platforms. We have identified research gaps, including, but not limited to, the need for scalable, real-time cyber-physical systems; limited research examining communication QoS under realistic conditions/traffic; and integrated cybersecurity strategies for MMGs. We suggest future research opportunities considering these research gaps to enhance the resiliency, adaptability, and sustainability of modern cyber-physical MMGs.

1. Introduction

The integration of renewable energy sources and the growth of smart grid technologies have revolutionized traditional power distribution networks, paving the way for introducing multi-microgrid (MMG) systems as a promising next-generation energy infrastructure option [1,2]. Microgrids face numerous challenges, including the stochastic nature of renewable energy generation, disturbances in transmission lines, disruptions in connections between microgrids (MGs) and the distribution network, and the possibility of unplanned power outages due to specific load demands; pulsed power loads (PPLs) might jeopardize a microgrid’s stability [3]. The heightened short-duration current behavior requires higher-rated power components, which may cause voltage and frequency fluctuations across the microgrid [3,4]. These vulnerabilities are especially problematic in islanded microgrids because they lack the stabilizing impact of an infinite bus (grid), which is generally used as a reference point for voltage and frequency management.

The concept of multi-microgrid systems offers a range of benefits and solutions, emphasizing improvements in the resilience and dependability of the overall power infrastructure [5]. Networking several microgrids is an emerging approach for increasing power system resilience and reliability. This strategy facilitates the integration of various distributed energy resources (DERs), leading to a robust electric infrastructure that leverages cost-effective and environmentally friendly power generation [5,6,7].

A profound shift toward decentralized energy generation is transforming power system operating paradigms. Unlike traditional centralized grids, MMG systems consist of interconnected microgrids. These microgrids can operate autonomously or cooperatively and afford MMGs greater operational flexibility and robustness. The distributed nature of MMGs enables a variety of DERs to be interconnected, including photovoltaic (PV), wind energy generation, and energy storage systems. By combining local generation and smart collaborative action, MMGs can proactively reduce transmission losses, improve energy efficiency, and improve power quality. Additionally, prolonged interconnections can lead to mutual support and improved outage reliability, compared to isolated microgrids. All these benefits also come with increased complexity for coordination and control, requiring a robust communication infrastructure and advanced energy management systems (EMS) to facilitate seamless operation between multiple layers and entities.

Advances in control strategies and communication technologies facilitate the coordinated operation of numerous microgrids, allowing the distribution system to efficiently meet growing daily energy demands [5]. By networking MGs in MMG systems, we provide a decentralized energy generation, distribution, and administration approach that allows for greater resilience, flexibility, and sustainability in addressing diverse communities’ and sectors’ changing energy demands. However, the EMS is required to properly control the dispersed energy resources inside MMG systems and maintain optimal performance. EMS plays a vital role in orchestrating the operation of individual MGs within multi-MG systems, maximizing energy output, and ensuring the stability of the MMG system across various circumstances [4,5,6]. It serves as the central control hub in MMG systems, coordinating the multiple DERs, including photovoltaics, wind turbines, batteries, and ultracapacitors, within each MG to ensure efficient energy production that meets load requirements.

Energy management in MMGs requires multi-level control architectures that can address local microgrid behavior while achieving system-wide objectives like minimizing cost, minimizing emissions, or maximizing reliability of service. Due to the inherent uncertainty in renewable generation and load variability, the EMS must continuously monitor the environment and apply adaptive control processes to direct microgrid operation in real-time. As larger systems are designed and constructed, many MMG developments prefer decentralized or distributed EMS designs to reduce the communication burden on the microgrid and improve resilience against operating failures. While decentralized EMS provides other significant advantages, the communication system must continuously communicate instructions and receive data from the microgrids. This makes the communication network performance a key enabler of energy management efficiency.

While the MMG system is a network of linked MGs, efficient communication between the EMS and the individual MGs, each with its local controllers and distributed energy supplies, is critical for successful energy management and system optimization. This communication network enables real-time data sharing, allowing the EMS to monitor DER performance and status, adapt energy management techniques, and maintain system stability in response to changing system circumstances and load demands [5,6,7]. Nevertheless, as MMG systems become more complicated and interconnected, the communication network becomes a significant source of vulnerability. MMG systems become increasingly vulnerable to cyberattacks and malicious intrusions as the communication surface expands, potentially compromising system security and disrupting operations [8,9]. As a result, MMG systems must be analyzed as cyber-physical power systems (CPPS), considering the intricate interactions between physical components, communication infrastructure, and control systems, as demonstrated in Figure 1.

Modern MMG architecture has an inherent cyber-physical nature, with physical electrical assets and the cyber layer or communication and control layered deeply. The communication network adds to a layer of threat now that the network’s reliability is not only supporting monitoring and control but is also a possible attack surface for cyber-based threats such as data tampering, denial-of-service (DoS) attacks, and false data injections (FDI). Cyber-attacks can lead to improper control actions, instability, and potential physical damage. This means MMGs need to be redesigned to integrate cybersecurity; this can be achieved in an integrated way that considers the power system and the network as one comprehensive cyber-physical system. The emergence of industry standards and secure communications will continue to converge to highlight the need for reliable communication infrastructures in MMGs.

The quality of service (QoS) in the communication network directly impacts the performance and stability of the MMG system. Communication issues, such as delays, changes in data rate, and communication failures, can significantly affect the effectiveness of EMS operations within MMG systems [5,6,7,8]. Delays in data transmission may lead to suboptimal decision-making by the EMS, resulting in inefficient allocation of energy resources and poorer system performance. Similarly, changes in data rate can alter the timing of control signals and system responses, leading to energy supply interruptions and grid instability [10,11,12]. Furthermore, communication failures can be highly disruptive to the functioning of MMG systems. For example, if communication fails while managing a pulsed load, the EMS may incorrectly schedule energy, resulting in voltage swings or power outages.

Impairments to communications networks (e.g., latency, jitter, bandwidth fluctuation, and packet loss) pose a serious issue for MMG control. Timely and reliable data exchange and communications between sensors, controllers, and DER units are the foundation of real-time EMS decisions and operations. Minor delays can extend and lead to loss of synchronization of control actions, missed load adjustments, or delay in detecting faults. Also, packet losses may cause decision makers and actors to receive incomplete or stale information at critical decision-making moments, which can deteriorate the system’s responsiveness and even lead to emergency states. Given all these considerations, accurately framing, modeling, and mitigating the impact of communications QoS, including physical-layer attributes and potentially higher-layer network attributes (which may impact latency and reliability in varying ways), is essential to developing a robust and resilient MMG.

Consequently, addressing communication issues and studying the performance of communication networks and their impact on MMG system operation is essential for optimizing the resilience of cyber-physical MMG systems under real-world operational conditions. To address these problems and enhance MMG system performance, researchers use simulators to examine the interacting dynamics of MMG systems and their communication networks. However, simulators focus primarily on power or communication networks, necessitating additional capabilities to mimic both networks simultaneously as a CPPS. Recognizing this constraint, the emerging strategy of co-simulating power and communication real-time simulators becomes a reliable approach for investigating the performance of cyber-physical power systems [13,14,15].

Real-time co-simulation infrastructures that combine power system simulators with network simulators have recently gained prominence as effective methods for examining cyber-physical aspects of MMGs. The bidirectional nature of electric behavior with communication delays makes them particularly useful for testing realistic EMS algorithms and control behaviors, particularly when there are complicated network conditions, e.g., congestion and cyber-attack events. This integration involves coupling tools, such as OPAL-RT or RTDs (for hardware-in-the-loop power system simulation) and Network Simulator 3 (NS-3), Network Simulator (Netsim), or Optimized Network Engineering Tool (OPNET) (for communications network emulation). These cyber-physical platforms enable detailed analysis of cyber-physical interdependence of system stability, control robustness, and resilience, providing insights that are not achievable through power-only or network-only simulations.

1.1. Paper Contributions

While MMG systems can be analyzed from different perspectives, this review adopts a cyber-physical systems (CPS) perspective. From a power system perspective, prior studies often focus on aspects of DER integration, voltage and frequency concerns, and stabilization issues [1,2,3,4]. From a communication networks perspective, prior studies have focused on communication network design and configuration, with significant emphasis on quality of service (QoS) [7,11,12], without considering implications for stability in power systems. However, such isolated viewpoints do not consider cyber-physical interdependence in modern MMG systems, where the quality of communication services and the overall cybersecurity shape physical performance.

The CPS perspective is particularly applied to MMGs because it considers the physical layer (DER dynamics, load behaviors, and stability requirements), cyber layer (communication networks, data exchange, and protocols), and interaction of the systems (effects of QoS on energy management systems, cybersecurity resiliency, and co-simulation realizations) simultaneously. The inclusion of CPS is increasingly recognized as critical in the recent literature [13,14,15] since it allows researchers to examine emergent vulnerabilities and operational risks associated with MMG scenarios, which may not be observed in isolation from both layers.

Accordingly, this focused review contributes to the body of knowledge on MMG systems by bringing together perspectives from control, communication, cybersecurity, and co-simulation and examining them through the lens of cyber-physical integration. While literature studies have often focused on isolated aspects of microgrids—such as hierarchical control or communication modeling, this work emphasizes the interdependence that shapes real-world MMG performance, particularly under diverse traffic and operational conditions. By systematically identifying the gaps in existing studies, the paper highlights how communication QoS and cybersecurity considerations are central to ensuring resilient and adaptive MMG operations.

Furthermore, the review critically discusses why these research gaps persist, including limitations in scalable co-simulation platforms and the need for integrating communication QoS with control design. Thereby offering a more analytical perspective rather than a purely descriptive one. Moreover, it provides future research directions incorporating emerging technologies, such as software-defined networking (SDN), digital twins (DTs), and physics-informed approaches, for the coordinated control, security, and resiliency of modern energy systems, as depicted in Figure 2. The key contributions of this focused review paper can be summarized as follows:

• Examining multi-microgrids operation and control strategies focusing on their cyber-physical characteristics, notably control approaches, and how these schemes interact with supporting communication networks.
• Review of communication network emulation and cyber-physical power systems analysis, emphasizing how power and communication simulation tools are beginning to be integrated to represent modernly complex cyber-physical interdependence in energy systems studies.
• Systematic identification of current research gaps and issues, primarily related to the limited exploration of QoS metrics (latency, throughput, packet loss) and the need for scalable, real-time cyber-physical frameworks capable of handling high traffic and heterogeneous communication infrastructures.
• Investigating cybersecurity issues in MMGs from a CPS perspective, specifically around detection and mitigation methods, and constructing integrated, adaptive, physics-informed security approaches with both a cyber and physical context.
• Future research opportunities include software-defined networking (SDN), digital twins (DTs), machine learning–enabled adaptive controllers, and CPS-centric frameworks for combined cybersecurity and QoS management of modern cyber-physical power systems.

1.2. Paper Organization

The remainder of the paper is organized as follows. Section 2 provides a cyber-physical perspective of multi-microgrids, which will be presented regarding the physical layer, communication components, and their interactions. Section 3 provides a thorough literature review on MMG operation and control, which discusses control strategies, energy management systems, communication networks, and quality of service considerations. Section 4 identifies the research gaps and challenges with a primary focus on limiting criteria, particularly the MMG studies examined, which considered ideal conditions concerning communication assumptions, limited cyber-physical interdependence assessment, and cybersecurity vulnerabilities. Section 5 discusses some promising future research directions, including SDN, DTs, physics-informed cybersecurity schemes, artificial intelligence-based adaptive control schemes, and scalable cyber-physical co-simulation platforms for an integrative evaluation of MMG. Lastly, Section 6 concludes the review paper by summarizing the key findings and highlighting the vital need for interdisciplinary approaches to realize more resilient and adaptive cyber-physical power systems.

2. Cyber-Physical System Perspective on Multi-Microgrids

2.1. Physical Layer Dynamics

Multi-microgrid (MMG) systems utilize distributed energy resources (DERs): photovoltaic arrays, wind turbines, battery storage systems, and pulsed demand loads, all interconnected through medium or low-voltage networks. Each microgrid has to be able to reactively and responsively manage frequency and voltage. The microgrid must operate in a highly dynamic environment, especially when islanded, as there is no grid inertia to absorb disturbances. PPLs—including industrial motors and electric vehicles (EV) rapid chargers—can introduce instantaneous and transient peaks in load demand on DER systems [3,4].

When this occurs, energy storage and fast-responding DERs must be dispatched immediately, and they rely on coordinated EMS algorithms within their control loops. Therefore, without reliable communication, these decisions could be too late or not communicated to the supplier to mitigate the instantaneous demand, resulting in the system being out of balance or an irreversible event.

Beyond local dynamics, the interconnected MGs present challenges in power sharing, synchronizing power across MGs, and congestion management. Specifically, operator tie-line power flows must be actively managed to avoid overloading, and load variations in one microgrid will disappointingly creep across and affect the neighboring microgrid. Thus, the EMS that operates over the MMG cluster needs to dynamically allocate energy across the MGs and control the MGs across this sub-system. The EMS must coordinate the reallocation of resources and energy to achieve balance in the face of uncertainty [1,2]. These management processes depend on common and effective communication-aware EMS logic and possibly more robust physical infrastructures.

2.2. Cyber Layer Components

Wired (Ethernet/IEC 61850, Modbus TCP) and wireless (Wi-Fi, LoRa, 5G/NR) segments make up communication networks in MMGs. There are latency, packet loss, throughput, and dependability trade-offs for every protocol and medium type. IEC 61850 GOOSE messages, for instance, provide quick tripping operations but rely on local area network (LAN) environments that closely control packet traffic. In contrast, wireless links are susceptible to congestion or interference and have variable delays and packet losses. For EMS state estimation, all these features are connected to the control loop timing and synchronization with time synchronization [11,12].

Sensor data flows and status updates are essential for local, secondary, and tertiary controllers for optimal operation. Beyond control thresholds, latency or jitter can make it difficult for a controller to follow time frames for frequency or voltage regulation or oscillate since feedback is useless, and missed packets can provide blind spots for internal decision-making. EMS designs must, therefore, account for communication risks by implementing fallback operational modes or utilizing robust delay-tolerant control algorithms [5,7].

To improve practice guidance, it is essential to associate communication technologies with the different needs of MMG applications, while also considering QoS and cybersecurity aspects. The various layers of MMG operation include multiple levels of control, ranging from primary to tertiary, each requiring differing latency, reliability, and security guarantees.

Table 1. Communication Protocols and Scenarios Adaptation [7,10,11,12].

Protocol/Medium	Typical Application Scenarios in MMGs	Suitability & QoS Characteristics	Trade-Offs/Limitations
IEC 61850 (GOOSE)	Protection schemes, substation automation, and intra-MG coordination	Low latency, high reliability, and determination within the LAN	Requires a well-managed wired LAN Wired links are less susceptible to external attacks but vulnerable to insider threats
Modbus TCP (Ethernet)	Supervisory control, device status monitoring, and local control data exchange	Simple, widely supported, and with moderate latency that is acceptable for non-critical commands	Limited support for complex event-driven control. scalability issues and minimal built-in security
Wi-Fi	Non-critical monitoring signals and human–machine interfaces	Flexible, low-cost, moderate bandwidth	Vulnerable to interference/congestion, and variable latency Not ideal for protection and susceptible to eavesdropping, spoofing, and DoS attacks
LoRa/LPWAN	Wide-area sensor data (DER availability, environment, metering)	Long range and ultra-low power	Very low data rate, high latency, and unsuitable for control/protection Limited encryption standards may expose data to interception
5G/NR	Wide-area MMG coordination, distributed secondary/tertiary control, V2G applications	Reliable, low-latency, high bandwidth, and strong mobility support	Deployment cost and performance depend on operator coverage Enhanced security features, but centralization may create single points of failure
Optical Fiber	Backbone links between MGs and central EMS coordination	Extremely high bandwidth, ultra-low latency, high immunity to interference	High installation and maintenance costs Physically secure, but insider attacks remain possible

summarizes the commonly used communication protocols in MMG applications [7,10,11,12]. It provides recommendations for the protocol-scenario combinations, while also considering the assumed trade-offs and limitations from a security perspective. From the MMG perspective, the protocol choice involves finding an equilibrated balance between QoS requirements, operating limitations, and cybersecurity implications.

For example, wired protocols (i.e., IEC 61850) are generally deterministic, low-latency, and ideal for MMG operation due to limited attack vectors. In contrast, wireless protocols (e.g., Wi-Fi/WiMAX/LoRa) have much simpler protocols and longer ranges, but added layers of security (i.e., encryption/authentication) may also need to be included to mitigate the impact of potentially successful cybersecurity threats. Emerging 5G/NR recommendations include ultra-low latency and improved security implementations. Still, they also involve centralized control planes, which can be likened to a single point of failure when considering the broader aspects of resilience in MMG. These guidelines offer practical insights into selecting communication protocols, considering operational performance and security trade-offs.

2.3. Cyber-Physical Interaction Dynamics

The cyber and physical layers are intricately connected; failures or disturbances in communication may cause control loops to operate based on stale or lost information, while disturbances in the power system, such as PPL surges or DER outages, create high communication loads (e.g., alarms and status updates), causing a network to congest further. Controllable decisions must be made on a strict schedule; delays or message losses can lead to out-of-phase control actions, incomplete state coordination, and reactive imbalances within microgrids.

Also, cyber-attacks, such as FDI or DoS, may induce malicious latency or inaccuracy into state estimation processes. If state estimation is corrupted or delayed, it may trigger inappropriate control actions, or the EMS in a protection scheme may unintentionally trigger undesired control actions [5,6,7,8]. These types of threats can be studied only if we conduct analysis based on real-time co-simulation, which allows for examining power levels and communications almost in parallel to explore and assess the power structure and the consequences on the integrity of power and communications.

Thus, from a practical implementation perspective, simply using a standalone simulation of either a power system or a communication system cannot reveal emergent or unusual behaviors that will stem from interdependence between the two structures. If you fail to understand timing error propagations or the beginning of control degradations, or if failure occurs, the consequences can lead to a failure propagation. The only way to do this is to co-simulate EMS performance and quality impairments in real-time, ensuring that the designers of MMG considered a robust, resilient, and safe structure of operation.

3. Literature Review of Multi-Microgrid Systems Operation and Control

The literature included in this section was selected based on a defined three-part criterion. First, peer-reviewed literature and conference papers strictly related to operation, control, communication, or cyber-physical attributes of MGs and MMG systems were included. At the same time, studies that were, for instance, economic in nature or did not relate to technical control/communication issues were excluded. Second, for the reviewed works, the timeframe primarily included recent literature, where the field of MMG systems and cyber-physical integration has accelerated. However, a few earlier foundational contributions were included when appropriate.

Finally, they would be preferred if the studies included real-time simulations, hardware-in-the-loop validations, and/or detailed modeling frameworks. Ensuring a minimum quality threshold versus more theoretical or conceptual research aspects was present. These criteria allowed this focused review to contain representative and technically rigorous contributions and appropriately filter out off-topic or non-validated works.

The last set of references is thematically organized into four identifiable classes: (i) control strategies and energy management inside MMGs, (ii) communication networks modeling and QoS in MGs, (iii) communication-aware control and impacts studies, and (iv) cyber-physical frameworks and co-simulation studies. Systematically analyzing and synthesizing the references for the review provides a strong basis for identifying existing research gaps and future research directions, as detailed in the following sections.

3.1. Control Strategies and Energy Management in Multi-Microgrid Systems

Numerous research and experimental studies have investigated the control, energy management, security, and communication aspects to advance the resiliency and reliability of multi-microgrid (MMG) system operation. In [16], researchers proposed a distributed control scheme based on an experimental study consisting of two control layers for MMG systems, with two levels of control: MG-control level and NMG-control level. The proper power sharing between the DERs and frequency/voltage reference tracking is the function of the MG-control layer, while regulating the power flow at the common coupling point (tertiary control) in each MG is ensured through the NMG-control layer.

The authors in [17] developed a new resilient distributed control algorithm to address the frequency synchronization issue within the NMG system under denial-of-service (DoS) attacks. They propose a unified notion of Persistence-of-Dataflow (PoDF) to quantify the attack effect on data unavailability and introduce an edge-based control framework. All the controller parameters are adapted through an online adaptation scheme. In [18], the proposed research focuses on developing an event-triggered distributed control algorithm equipped with predictive compensation for AC/DC networked microgrids under DoS attack. The aim is to enhance the control performance and the MMG stability under DoS attacks by estimating the control variables. The investigations were implemented through a hardware-in-loop (HIL) simulation using an RT-LAB simulator, MATLAB, and a digital controller. Nevertheless, they did not include the communication network as a real-time simulation; instead, they modeled it using MATLAB/Simulink.

The investigations in [19] focused on multi-MG voltage regulation by proposing a volt/var control (VVC) strategy based on data-driven control. They optimize the reactive power output from each MG based on a conventional neural network (CNN). Then, they replace the time-consuming iteration with a reactive power adjustment algorithm to streamline the real-time application of the proposed process. A centralized protection scheme for MMG is presented in [20] through real-time simulations; the proposed method employed an adaptive relaying scheme alongside a network theory-based zone selection method. By leveraging the wide range of IEC 61869-9 [20] sampled measurements and IEC 61850 GOOSE messages, the proposed centralized approach can identify the protection zone and select the appropriate protection scheme in various circumstances. However, they did not address the adverse impact of the communication network QoS on the performance of the proposed protection scheme.

Authors in [21] have proposed a multi-source trading method that operates in a distributed manner to optimize the operation of interconnected microgrids, enabling diverse multi-energy production/load demands. This has been performed by dividing the trading of MMGs with multi-energy and communication into the social allocation of multi-resource and payoff allocation as a subproblem. The proposed approach was tested through numerical analysis of three interconnected MGs, noting that they did not include the impact of unreliable communications on the MMG system performance.

Focusing on MMG’s energy management, researchers in [22] have introduced EMS within the MMG system based on model predictive control (MPC). The operation of each MG is managed individually through an MPC scheme, and a coordinating algorithm is proposed to ensure cooperative operation between the MGs within the networked MG system, addressing the energy management scalability problem. Evaluation is performed through a co-simulation environment without analyzing the communication network performance within the proposed MMG system. In [23], the authors present an EMS in a multi-MG system based on a data-driven algorithm. The study employs deep neural networks for MMG simulation under variations in retail price signals, without requiring local generation or consumption information. In addition, the price can be optimized using a model-free Monte Carlo reinforcement learning method to ensure the maximum selling power profits with the minimum demand-side peak-to-average ratio.

Meanwhile, authors in [24] explore an Economic Dispatch (ED) method based on a consensus-distributed scheme within a DC microgrid EMS. It introduced a resilient, collaborative, distributed EMS (R-CoDEMS) framework for modeling the dynamics and cybersecurity aspects of the DCMG system. It also proposed an FDI attack detection and mitigation technique to enhance the DCMG resiliency. A numerical analysis to investigate the effectiveness of the proposed work has been presented using MATLAB simulation.

In [25], the authors proposed an EMS to coordinate the operation of an excessive number of MMGs, employing game theory to address the energy cost allocation problem. The developed model involves real-time data regarding electricity prices. They present a numerical analysis of the proposed EMS; however, it has not been tested in real-time operation. Researchers in [26] have presented an EMS for the MMG system, aiming to minimize harmful emissions and total cost. They proposed a numerical analysis to solve the management issue within the MMG system. In [27], the authors presented an energy management technique employing an adaptive optimal control mechanism for an MMG system. The Tunicate Swarm Algorithm is considered an optimization function at the tertiary control level to obtain the optimal connection models of microgrids. Using MATLAB and DigSilent, simulation results have proved the developed technique’s effective performance under several operating conditions.

Overall, the studies [16,17,18,19,20,21,22,23,24,25,26,27] illustrate significant contributions to control, protection, and EMS schemes in MMG systems. However, they notably overlook the explicit role of communication networks and their influence on the proposed schemes. Exploring the networks’ QoS is crucial when investigating the performance of microgrids, as they play a vital role in information exchange and optimal coordination among microgrids, which helps advance research in cyber-physical MMG systems. The previous analysis [16,17,18,19,20,21,22,23,24,25,26,27] identified three patterns that could be summarized as follows:

• Control and EMS designs typically assumed ideal or near-ideal communications—sensors and actuators to exchange measurements and set-points reliably and with no time delays, thus demonstrating stability and power-sharing claims with an under-specified set of real QoS impairments.
• Most validations were software-based (MATLAB/Simulink, OPAL-RT) with no packet-level network emulation, so controllers were not exercised against burst loss, re-ordering, or variable jitters such that secondary and tertiary loops would remain in practice.
• Scalability was assessed at the MG or small MMG scale, with limited attention paid to hierarchical coordination, e.g., how do tertiary EMS decisions map to constrain inter-MG communication while secondary controllers seek to maintain the power quality.

Consequently, although these contributions advance droop designs, predictive EMS, and coordination methods, their robustness has not been tested under realistic networking conditions, where decision latency and stale data become first-order effects. These contributions provide a solid foundation for energy management in MMG and its control design. However, most of them assume ideal communication conditions—control signals and measurements are considered timely and reliable. In practice, the communication channel introduces latency, jitter, and packet loss, which will render coordination across control levels more difficult. Moreover, most control frameworks were assessed in MATLAB or OPAL-RT, and no modeled behavior associated with the real network, nor did they rely on communication network simulators when validating the control framework.

Thus, while the control methodologies were robust, they remain unvalidated concerning impaired signals associated with practical communication with the actual network. This limitation certainly gives rise to a need to explore the communication networks, including their structure, models, and the associated QoS impairments. As communication QoS directly impacts the operation of power systems. The next section reviews contributions that explore the modeling of MG communication networks and their QoS assessments.

3.2. Communication Networks Modeling and QoS Considerations in Microgrids

Researchers in [5,8,28,29,30,31,32,33] have proposed various simulation models for communication infrastructures for microgrid applications. Authors in [5] have proposed a comparative analysis for Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) communication protocols under several traffic patterns. The evaluation was presented for an MMG communication network that is designed to accommodate centrally controlled multiple DC interconnected microgrids. The emulation of the communication infrastructure and the QoS performance analysis were performed by using the NS-3 emulation environment. Researchers in [8] presented a detailed communication emulation model that integrates multiple islanded inverter-based microgrids. The study focused on evaluating the communication QoS in terms of the mean delay, data rate, and percentage packet loss using the UDP communication protocol.

In [28], the authors have developed a communication infrastructure within the microgrid system using the OPNET simulation tool. Through the proposed model, they explored the impacts of communication issues, such as delays and packet losses, on the performance of the communication network, specifically in the case of MGs responding to utility grid disturbances, without investigating the impact on MG stability. The introduced research in [29] focused on modeling a communication network for a protection scheme within a distribution power system.

The research outcomes indicate that using OPNET as a communication network simulation tool enables us to analyze the QoS of communication performance. Moreover, in [30], the authors present a study on the performance of a communication network by modeling a smart microgrid communication infrastructure using NetSim software. They propose an optimal communication topology for data transfer within the MG and test performance metrics, including round-trip time delay, throughput, and packet loss percentage. The authors highlighted in [31] the development of a communication infrastructure that mimics monitoring architecture based on the IEC 61850 standard, using the OPNET simulator for large-scale PV power stations. The primary goal is to evaluate communication behavior across various circumstances. Evaluation of the proposed model was assessed in terms of end-to-end delay.

The research conducted in [32] focused on emulating the communication infrastructure in a shipboard power system (SPS) that consists of four zones to investigate the SPS as a cyber-physical system. They integrate the RTDs and Common Open Research Emulator (CORE) to emulate the shipboard power system and corresponding communication network within a co-simulation framework. However, they overlook the impact of communications performance on SPS performance. The proposed work in [33] explores the performance and scalability of a communication network based on the IEC 61850 standard within an MG that facilitates a peer-to-peer (P2P) energy trading architecture. The performance of the communication network, in terms of delays and communication bandwidth, was analyzed through simulations conducted using OPNET.

Although studies [28,29,30,31,32,33] targeted single-MG contexts, they remain directly relevant to MMGs, as each MG forms a fundamental building block within an MMG. By understanding how QoS affects local MG control, these works provide the foundation for scaling up to MMG scenarios, as investigated in [5,8].

Work proposed in [5,8,28,29,30,31,32,33] comprehensively explored the modeling of communication networks and their performance metrics using different simulators. Nevertheless, they focused more on communication QoS without investigating their critical influence on the microgrid’s operation. The studies provided valuable insights into latency, throughput, and packet loss for different architectures, but they rarely relate these QoS metrics to control performance. From an MMG perspective, this is a vital shortcoming as localized QoS degradations at the MG level may propagate to destabilize MMG coordination. Accordingly, it affects tie-line power flows, shared reserves, and overall system resiliency.

Comprehensively characterizing latency, throughput, and loss in protocols and topologies is performed by earlier works [5,8,28,29,30,31,32,33]. But it is rare for QoS metrics to be mapped to power system performance variables. Extending this analysis from MGs to MMGs requires explicit mapping of communication impairments to interdependent operational challenges at the system level. For an MMG operation, a useful transition to map network side indicators to control side performance indicators is the following:

• Delays and their influence on secondary control performance and power quality.
• Jitter impact on inter-MG tie-line oscillations and EMS convergence frequency.
• Packet loss ratio probability and how it will affect the control system performance.

These interdependence investigations are needed to convert QoS outcomes from descriptors into operational quantities. Without this type of investigation, we are left with a descriptive QoS result and an open question on how to relate the time delay distribution or burst loss profile to control design, EMS scheduling horizons, or protection margins. Addressing this gap requires research that articulately connects QoS impairments to contexts for control performance. This change in focus produces the next section, which reviews communication-aware control approaches and research that tries to analyze how delays, packet loss, and other network performance degradations translate into outcomes of stability, frequency regulation, and power-sharing.

3.3. Communication-Aware Control and Impacts Studies

Focusing on the impact of communication on the performance of MG, as the fundamental element within the MMGs system-level, researchers in [34] have analyzed the effect of communication delay on controlling frequency and load sharing within an AC microgrid. They developed a control algorithm in a distributed manner to restore frequency deviation and achieve accurate load demand sharing among the DERs, considering communication latency. Authors in [35] also addressed the delay in a wireless communication scenario involving the DC microgrid bus voltage during simulation analysis. Study [36] proposed an intelligent distributed control scheme to enhance the performance of DC microgrids under communication delays and interconnectivity failures. Using MATLAB simulation, various scenarios were run on an updated version of the IEEE 34-bus system to study the effectiveness of the proposed scheme in ensuring seamless communication.

The problem of delays in communication is also addressed in [37], where the authors conduct a simulation study aimed at controlling DCMG in a centralized manner, while considering the influence of delays. They presented the DCMG with the delay using a mathematical model and subsequently proposed a control scheme to mitigate the impact of communication delay and ensure MG stability. In [38], the researchers conducted a study focusing on load frequency control (LFC) of an islanded microgrid with a high penetration of renewable energy resources. They developed an optimal virtual inertia control scheme to ensure MG stability under different operating conditions, including the occurrence of communication delays between the LFC center and the primary control loop of the MG.

Real-time simulation studies using OPAL-RT as a power systems real-time simulator were examined in [39,40,41], targeting the control of microgrids. Predictive controllers’ role in enhancing AC microgrids’ performance under communication delay is investigated in [39]. The authors developed a control scheme based on predictive control algorithms to restore voltage variations for individual DERs within the MG, compensating for the impact of communication delays. Meanwhile, the authors in [40] explored the problem of loss of transmitted packets in a DC microgrid system that includes energy storage systems (ESS). The packet loss issue can be addressed through the proposed multi-agent dynamic-tracking consensus protocol (DCP), which enables the estimation of information within the studied MG. In addition, they propose an adaptive control strategy to regulate power sharing between the ESS, ensuring a balance in their state of charge (SoC) during operation. Relevant studies [34,35,36,37,38,39,40] demonstrated the importance of single MG scenarios in the MMG context, as an MMG is essentially a coordinated system of interconnected MGs. Insights gained from [34,35,36,37,38,39,40] provide the necessary foundation for understanding and addressing challenges that emerge once multiple MGs are integrated, as reflected in [41,42].

At the MMG-system level, the impact of networks’ QoS on latency and failure is explored in [41]. The researchers developed a leader-follower control scheme in a distributed manner for controlling the frequency in an MMG system with multiple DERs. By employing a nonlinear fractional-order PID controller adapted through an adaptive hybrid chaotic atom search-particle swarm optimization (ACAS-PSO) algorithm, the proposed control strategy can enhance the frequency performance of the MMG system across various real-time simulation scenarios. Authors in [42] investigated the impact of communication loss. They explored methods to improve the performance of the MMG system, which includes two interconnected microgrids, addressing this communication issue using PSCAD/EMTDC simulations. The proposed strategy is based on switching the networked MGs’ operation into two phases: division and unification. As a result, they can ensure flexible and resilient operation under severe conditions without requiring additional communication to switch between different operational phases.

These examples demonstrate why discussing single MG scenarios is still relevant in the MMG context: an MMG is essentially a coordinated system of interconnected MGs. Insights gained from MG-focused studies [34,35,36,37,38,39,40] provide the necessary foundation for understanding and addressing challenges that emerge once multiple MGs are integrated, as reflected in [41,42].

These communication-aware control studies demonstrate how to employ strategies that can mitigate the effects of latency and packet loss using prediction control, consensus protocols, and adaptive algorithms. However, these investigations are based on power system simulators, in which delays are generally modeled, rather than at the packet level with actual communication impairments. Thus, it is unclear how these strategies perform in the presence of real disturbances over the network.

The reviewed studies [34,35,36,37,38,39,40,41,42] investigate the influence of communication delays, packet loss, and failures on the performance of MG and MMG systems, targeting various aspects, including power sharing, frequency control, and voltage stability. They illustrated that predictive, consensus, and adaptive controllers can tolerate bounded delay and moderate loss of packets. However, their disturbance models are usually time-invariant, not correlated with real-time communication network performance degradation caused by congestion, contention for capacity in communication channels, and cyberattacks. Two conclusions are as follows:

• The stability margins that are reported may over-approximate resilience when the performance degradation takes the form of jitter bursts that switch the controllers on and off synchronously and intermittently.
• Estimation and consensus schemes that assume symmetric availability of packets may be biased when there is asymmetrical packet loss across MGs, resulting in asymmetric power sharing.

Hence, while these strategies [34,35,36,37,38,39,40,41,42] are promising, their guarantees remain contingent on traffic conditions that are gentler than those observed in MMG communications or under attack. As these research examinations were conducted through a power systems-based simulation environment, this introduced a notable research gap because the communication infrastructure within the MG or between the MMGs was not emulated. Consequently, these research studies evaluate power system performance without investigating the microgrids from the cyber-physical power system operation perspective, incorporating power and communication network QoS in a unified framework.

To overcome this limitation, researchers are transitioning to cyber-physical frameworks that incorporate both the power and communication layers within co-simulation environments, capturing more realistic packet-level network dynamics coupled with physical power system behavior. Such frameworks allow for representations closer to reality for latency, jitter, congestion, and cyber-attack scenarios. The following section explores the cyber-physical and co-simulation studies, demonstrating how they cover the gap identified and contribute to the resilient operation of a cyber-physical power system.

3.4. Cyber-Physical Frameworks and Co-Simulation Studies

Building on the need for unified cyber-physical approaches, the reviewed studies [43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61] employed co-simulation frameworks combining power system simulators with network emulators. These relevant research studies comprehensively examine the operation and control of microgrids as cyber-physical power systems. Authors in [43] have implemented a co-simulation system that integrates Grid LAB-D and network simulator ns-3 to emulate the MG power system and its communication infrastructure, respectively. The work aims to study the impact of communication on the dynamic stability of an islanded AC microgrid, considering the integration of wired and wireless communication networks to ensure the resilient operation of the control system.

The role of integrating the power and communication networks for the resilient control of MG has been explored in [44]. While addressing the delay changes, data-driven control of MG is presented and validated through a cyber-physical framework that combines Grid LAB-D and NS-3 alongside the proposed central control running in Python. Experimental tests demonstrate the effectiveness of the control scheme in enhancing the MG’s dynamic response in the IEEE-123 model, particularly when subjected to communication delays.

Study [45] focused on controlling the DCMG voltage under delays and information loss in communication by using a PI-predictive control algorithm. The robustness of the proposed technique has been tested through a hardware setup that integrates PV panels with a battery to form a DCMG, with the implementation of the proposed controller in OPAL-RT, allowing it to run in real-time. In [46], researchers assessed the frequency control performance of an AC inverter-based MG in an autonomous mode of operation. The performance of the MG was analyzed as a CPPS under various cyber scenarios, including communication delays and distributed DoS (DDoS) attacks. They implemented an experimental study using real-time simulation of ACMG and the communication model, combining OPAL-RT and ns-3 in a co-simulation platform.

The research proposed in [47] explores the protection of multi-MG systems problems by developing an adaptive central protection strategy based on artificial neural networks (ANN) and communication networks. The introduced strategy comprises two phases of operation: identifying and isolating the fault within the MG through the central protection agent and activating the local protection units under islanding operation. This can be performed using the IEC 61850 communication standard to monitor the overall operation of the MMG system and transfer the tripping actions. The proposed MMG protection study was evaluated using a framework that integrates MATLAB, NetSim, and an IEC 61850 emulator. A Power-HIL (PHIL) framework was presented in [48] to assess strategies for service restoration in networked MG systems. To emulate the outage scenarios, the authors have integrated a real-time EMT feed model, an advanced optimization tool, a grid-forming inverter, and a Modbus-based DER dispatch. The results demonstrate the effectiveness of PHIL-based approaches in enhancing the resilience of MMG systems following service restoration after outages.

The cyber-physical interdependence analysis was considered in [49,50] by investigating the correlation between the control system performance and communication QoS for an islanded AC microgrid. In [49], real-time simulation analysis was considered for evaluating the communication QoS for the developed ns3-based communication emulation that facilitates a communication-based secondary control scheme responsible for the voltage and frequency regulation of the microgrid. In [50], the authors presented a Cyber-HIL MG framework for frequency regulation in an islanded low-inertia MG. The study investigated the frequency performance under several practical challenges, considering the high penetration of renewable energy resources, communication delays, and cyberattacks.

In [51,52,53,54,55,56,57,58,59,60,61], cyber-physical co-simulation testbeds were presented to enable the analysis of cyber vulnerabilities in modern energy systems. Authors in [51] have integrated the physical layer, represented by OPAL-RT and Typhoon digital simulators, with the cyber layer utilizing EXata network simulator. An extensive analysis was conducted to assess the performance of MGs during cyberattacks. In [52], the proposed testbed incorporated OPAL-RT for physical layer representation and the utilization of open-source tools, such as NS-3 and Docker containers, for representing the cyber layer. The experimental validations were conducted considering the interdependence between the control performance and communication QoS for an islanded autonomous MG and shipboard power system.

Researchers in [53] developed a co-simulation platform that integrates PowerWorld Simulator and CORE to emulate the physical and cyber layers. Several attack scenarios were presented to evaluate their impact on the stability and resiliency of a microgrid system. The work presented in [54] proposes a cybersecurity scheme that utilizes SDN for a controllable communication surface under cyberattacks. The experiments were conducted within a cyber-physical framework that accommodates the IEEE 34-bus test feeder for the physical layer and Cisco Modeling Lab for the cyber layer representation. Study [55] developed a CPS-based MG simulation platform, enabling real-time interactions between the information and energy domains. A finite state machine-based state transition scheme was introduced for fault optimization scenarios. The scheme was experimentally validated via an HIL system that includes a rapid control prototype; however, there are still limitations regarding the packet-level network QoS and cyberattack influence on the MG performance.

A hybrid multi-model Co-simulation framework was designed in [56], integrating heterogeneous simulation tools, including software simulators (e.g., MATLAB/Simulink, Modelica, EnergyPlus, Python models), digital real-time simulators (OPAL-RT), and physical devices (smart meters). A middleware layer ensures real-time data exchange and synchronization using IEEE 1588 PTP. Several scenarios were implemented, including voltage regulation service at the distribution level within building EMS, PV, batteries, and physical smart meters interconnected with a grid model in OPAL-RT. Researchers in [57] presented a co-simulation and controller-HIL (CHIL) platform, bridging modeling and hardware validations in the MG research domain. The platform incorporates RTDS with Field-Programmable Gate Array (FPGA)-based control to enable interoperability testing in multi-domain subsystems. The integration between the developed CHIL and the co-simulation facilitated the resilience validation across load variations, varying in wind penetrations, and fault scenarios, demonstrating the framework’s feasibility under practical challenges.

Advances towards co-simulation platforms targeting the MMG operation were demonstrated in [58] by developing a platform that integrates GridLAB-D and HELICS. The platform was introduced to train multi-agent controllers against cyberattacks that target the grid-forming inverters, while ensuring data privacy across the MMG infrastructure. A federated-based algorithm was developed to facilitate the adaptive control operation with HIL experimental validations. Results demonstrated the effectiveness of the developed mechanism in mitigating adversarial disruptions. Study [59] presented an innovative cyber-physical co-simulation framework that integrates HIL validation with machine learning-based resilience improvements for wide-area damping control (WADC). The study highlights challenges in terms of communication modeling and limited attack scenarios that limit its generality for several cyber-physical power systems applications.

A cyber-physical testbed that integrates power system real-time digital simulators (OPAL-RT, RTDS) with EXata network emulation software was proposed in [60]. The platform targets assessing cyber vulnerabilities in large-scale power systems and supports hardware-in-the-loop and human-in-the-loop evaluations. The developed testbed’s effectiveness was validated using the New York State power grid model in RSCAD, showing its applicability and scalability for cybersecurity analysis in modern energy systems. Authors in [61] presented a co-simulation framework targeting communication networks performance evaluation for smart grids. The platform includes power system simulators, such as OPAL-RT, RTDS, and Typhoon HIL, with communication network emulator software (NetSim). The results focus more on the communication QoS aspect under several operating conditions. However, there is a persistent need for more exploration regarding cybersecurity assessment and power system performance analysis.

Table 2. Summary of Cyber-Physical and Co-Simulation Literature Studies.

Ref.	Framework/Tools	Focus and Application	Cyber-Physical Integration	System Level	Validation/Test Case	Key Contributions	Limitations
[43]	GridLAB-D + NS-3	Impact of communication on stability of MG	Wired & wireless communication integrated with the MG	Single MG	Co-Simulation system	Resilient control operation under mixed communications	Limited to ACMG; no MMG scope
[44]	GridLAB-D + NS-3 + Python	Data-driven central control under delays	CPPS with integrated control for MG	Single MG	Co-Simulation with delay variations	Enhanced dynamic response with data-driven methods	Delay-only focus; lacks scalability to MMG
[45]	OPAL-RT + PV + Battery hardware	Voltage control in DCMG under comm. delays/losses	Real-time HIL linking power (PV + battery)	Single MG	HIL	PI-predictive control robust under comm. impairments	Limited to DC single MG
[46]	OPAL-RT + NS-3	Frequency control of ACMG under cyber scenarios	Real-time CPPS with DDoS + delay emulation	Single MG	Real-time co-simulation	Frequency control robustness tested against cyberattacks	Limited attack spectrum; no MMG
[47]	MATLAB + NetSim + IEC 61850 emulator	Adaptive ANN-based central protection for MMG	Central protection agent + comm.-based IEC 61850	MMG	Real-time co-simulation of MMG	Fault identification and isolation; coordinated local/central protection	Protection-only focus; lacks EMS/control depth
[48]	EMT real-time + OPAL-RT + PHIL + Modbus	Service restoration post-outage in MMG	Power-HIL with inverter dispatch + comm.	MMG	PHIL-based outage scenarios	Restoration strategies validated; resilient service recovery	Limited to outage recovery; not ongoing MMG dynamics
[49]	OPAL-RT + NS-3	Secondary control (V/f) under QoS constraints	Comm. QoS impacts on sec. control	Single MG	HIL	Evaluates QoS effect on control performance	Single MG only; no MMG
[50]	OPAL-RT + NS-3 + Docker	Frequency regulation in MG	Renewables + delays + cyberattack scenarios	Single MG	CHIL	Low-inertia MG stability tested under practical cyber challenges	No MMG-level scalability
[51]	OPAL-RT + Typhoon + EXata	Cyberattack resilience assessment	Multi-simulator (physical + comm.) integration	Single MG	HIL	Demonstrates MG under coordinated cyber threats	Lacks MMG integration
[52]	OPAL-RT + NS-3 + Docker	QoS–control interdependence in MG + shipboard	Containerized control with NS-3	Single MG	HIL	Highlights cyber-physical interdependence with modular emulation	Mostly single-MG; shipboard extension only
[53]	PowerWorld + CORE	Stability under cyberattack scenarios	Cyberattack injection in CPPS	Single MG	Co-Simulation	Attack impact on MG stability	No real-time validation; lacks HIL
[54]	IEEE 34-bus + Cisco Modeling Lab	SDN-based cybersecurity for MG	Programmable SDN comm. under attack	Single MG	Co-Simulation	Controllable comm. surface for cyber resilience	Small-scale feeder; limited generality
[55]	CPS-based MG sim. + HIL	Fault optimization with FSM transitions	Real-time energy–info domain coupling	MMG	HIL with control prototyping	Novel FSM-based optimization validated in HIL	Weak on packet-level QoS; no MMG
[56]	Modelica+ EnergyPlus+ OPAL-RT+ smart meters	Voltage regulation + EMS at the distribution level	Multi-model sync via IEEE 1588 PTP	MMG	Heterogeneous platform	Comprehensive EMS validation across domains	Very complex; computationally heavy; no MMG
[57]	RTDS + FPGA + CHIL	Resilience validation under load/fault scenarios	Real-time CPPS + controller-HIL	Single MG	Practical CHIL + co-simulation	High level of fidelity with FPGA-based control	Focus on single MG; limited MMG scalability
[58]	GridLAB-D + HELICS + HIL	Adaptive MMG control against cyberattacks	Federated learning + multi-agent control + privacy	Single MG	IEEE test feeder + HIL	First federated-based CPS control with HIL validation	Computational intensity; limited attack exploration
[59]	ML-based WADC + HIL	Wide-area damping control resilience	HIL integration with ML resilience	Single MG	HIL	Machine learning improves damping control under cyber stress	Limited comm. modeling; not MMG
[60]	OPAL-RT/RTDS + EXata	Cyber vulnerability assessment	Large-scale grid CPPS with HIL	Single MG	RSCAD-based real-time validation	Scalable cyber-physical testbed	Grid-level, not MMG
[61]	OPAL-RT/RTDS/Typhoon + NetSim	Comm. QoS in smart grids	Comm. emulation + power RT simulators	Single MG	HIL	Strong focus on comm. QoS	Weak cybersecurity integration; no MMG

summarizes the relevant co-simulation studies and cyber-physical interdependence analysis for power systems [43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61]. Despite the relevant research presented in evaluating power systems operation from a cyber-physical perspective, study [47] is the only one that targets MMGs and their correlated communication QoS. Moreover, ref. [48] evaluated the restoration services after outages with the MMG framework with PHIL simulation; however, the study is limited to the recovery performance after outages rather than the ongoing MMG cyber-physical dynamics. Additionally, work presented in [58] addressed the CPS perspective of operation in MMG frameworks, making a relevant contribution to this research area. However, challenges persist regarding the computational intensity of the developed federated training mechanism and the limited exploration of communication vulnerabilities. Notably, the relevant cyber-physical analysis studies presented focused on the single microgrid level of operation without tackling the interconnectivity between microgrid systems. Accordingly, the cyber-physical implementations in [43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61] demonstrated the potential for co-simulation between power systems and communication modeling environments. However, there are limitations, as most are still single-MG environments, and systematic studies of the impacts of QoS on MMG-tier EMS operations are lacking. The following structural limitations persist:

• Scope of research investigations, as studies focus more on single-MG dynamics, and MMG-level phenomena (e.g., tie-line congestion management and competing EMS updates) remain in need of more exploration.
• Synchronization and interfaces of the cyber and physical layers emulators can hide transient bursts and scheduling jitter that may affect control loops.
• Reproducibility and the reliance on open-source tools must be considered, as an open, packet-accurate, real-time co-simulation stack for MMGs can extend to EMS optimization, frequency regulation, and protection needs more investigation.

Overall, the literature discussed in Section 3 indicates significant advances in multi-microgrid operation and control, particularly in EMS designs, control strategies, and cyber-physical networked frameworks that integrate power and communication networks. However, most studies either focus on a single MG or examine communication and control aspects in isolation. Only a limited number of works explore MMG-level integration with a cyber-physical perspective, and even these are often restricted to specific applications like protection schemes. Moreover, most studies do not provide a comprehensive treatment of the simultaneous impacts of communication impairments in real-time, hierarchical coordination, or the simultaneous impacts of communication QoS on control performance.

This focused review highlights the need to address the remaining research gaps in MMG operation. Particularly in areas such as scalable co-simulation, integrated control and communication analysis, and real-time cyber-physical validation, which will be discussed in the following section.

4. Research Gaps and Challenges

Despite notable progress in the operation and control of MG and MMG, including systems that consider communication, many critical gaps remain open, which are preventing more widespread deployment and the resilience of real-world cyber-physical power systems. To clearly show how the studies presented here address these gaps, we developed a Venn diagram (Figure 3) to compare and situate each study according to its communication, control, and real-time implementation approach. Figure 3 provides a multi-dimensional view of the breadth and depth of previous work, and most notably indicates the lack of overlap between control strategies and the real-time network emulation in an MMG context.

One of the crucial points is that most previous efforts focus on only one or two facets of a cyber-physical system. For instance, while research on control and energy management approaches has been extensive in both the MG and MMG examples [16,17,18,19,20,21,22,23,24,25,26,27], most studies have relied on idealized communication assumptions. Conversely, work [5,8,28,29,30,31,32,33] provided detailed models of communication and its effects on MG operations. Still, nearly all overlooked the detailed dynamics of any power system aspects or control feedback loops, which limit their effectiveness in fully integrated systems. Additionally, the intersection of control and communication in a real-time implementation, specifically for MMG-level cyber-physical power systems.

The existing literature is summarized in

Table 3. Classification of Literature Studies Based on Control and Communication Focus.

Ref.	Work Categories	Control and EMS Strategies	QoS Modeling	Cyber-Physical Analysis	MMG Focus	Design Trade-Off/Limitation
[16,17,18,19,20,21,22,23,24,25,26,27]	EMS and Control Strategies	✓	✗	✗	✓	Assumes ideal communications High fidelity control, but ignores network effects
[5,8,28,29,30,31,32,33]	Communication Modeling	✗	✓	✗	✗	Accurate network evaluation, but neglects the impact on control performance
[34,35,36,37,38,39,40,41,42]	Communication-based Control Schemes	✓	✓	Partial	Partial	Simulating control and network, but with limited real-time validation and scalability
[43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61]	Cyber-Physical Interdependence Analysis	✓	✓	✓	Partial	Comprehensive, but often computationally complex Limited MMG scalability

based on four distinct research directions:

• Control and EMS strategy integration
• Communication quality-of-service (QoS) modeling
• Cyber-physical interdependence analysis
• Specific focus on MMG operation and control topologies.

The classification demonstrates an evident gap in the literature evaluating cyber-physical MMG architecture using real-time experimentation or higher-fidelity co-simulation approaches. For example, ref. [41] models MMG control schemes, but only performs the analysis via software experiments that do not effectively emulate the real-time aspects of protocol disturbances such as jitter or burst communication dynamics. Similarly, refs. [30,33] both analyze the performance of the communication aspect in microgrids with common networking simulations, but do not provide an extensive treatment of control performance elements. Even the presented work in [5,8] has tackled the communication networks’ QoS for MMG infrastructures, but overlooked its impact on the power system operation.

Study [47] presents one of the few attempts at developing an MMG communication implementation under the cyber-physical domain. Nevertheless, it is limited in scope of application to protection use cases and lacks an open, extensible platform, such as for EMS optimization or frequency regulation. Additionally, study [48] has tackled the networked MGs operation with PHIL investigations; nevertheless, the developed optimization-based restoration of services solution did not consider the communication QoS or cyber-resilience under operation, leaving an area of research in terms of real-time cyber resiliency under MMG operation. Moreover, the introduced co-simulation MMG framework presented in [58] poses some challenges regarding computational scalability and practical data privacy concerns in real-world implementations that warrant further exploration.

Table 4. Key Research Challenges in Cyber-Physical MMG Systems.

Challenge Area	Description	Current Research Status	Research Gaps	Implications/Trade-Offs
Communication QoS Evaluation	Impact of latency, jitter, and packet loss on MMG control and stability	Mainly idealized models or isolated simulations	Lack of real-time, integrated communication-power co-simulation with dynamic traffic patterns	Simplicity of models reduces operational relevance Realism requires complex co-simulation
Realistic Control-Communication Integration	Co-simulation frameworks combining both power and communication networks	Few cyber-physical testbeds; limited studies on hierarchical MMG control	Need scalable frameworks for MMG systems with diverse DERs and communication types	Trade-off: fidelity vs. scalability Complex integration limits adoption
Cybersecurity	Resilience against DoS, FDI, and spoofing attacks	Initial proposals, mostly simulation-based	Comprehensive real-time attack detection and mitigation with adaptive control has not been fully explored	Simplified models may overlook stealthy attacks Full implementation is costly
Network Topology and Traffic Impact	Effect of network architecture, communication protocols, and traffic variations on control	Basic studies on network delay, packet loss	Limited focus on varied traffic loads, multi-LAN MMG communication impact analysis	Static topologies simplify experiments but ignore realistic traffic interactions.
Standardized Evaluation Metrics	Common KPIs and benchmarks for communication-enabled control in MMGs	No widely accepted standards	Industry-wide standards for performance, latency margins, and resilience	Without standardization, results are hard to compare or generalize

provides an overview of key unresolved issues regarding integrating communications networks into the operation and control of MMGs. One of the significant issues is assessing the impacts of communication QoS on latency, jitter, and packet loss, which all have the potential to affect system stability and the response of control loops. Most current models are so simplistic or idealized that they presume constant delay or that communication is error-free. However, real-world networks—especially under dynamic traffic conditions—do not experience continuous latency and rarely experience congestion challenges, where congestion can impact time-critical control actions. Also, PI controllers typically perform secondary voltage and frequency regulation roles susceptible to delays. If these regulators are not appropriately modeled, the actual performance of such a regulator may be overestimated by simulation.

Despite research on co-simulation frameworks, few studies have successfully realized synchronized real-time co-simulators. Scalability remains a fundamental limitation, particularly for MMG systems with hierarchical communication architectures involving MMG data flows. Most current simulation platforms are permissioned for single MGs; therefore, the expandability to multi-domain, layered MMGs remains to be further explored. Cybersecurity is also a crucial area of concern. While a few studies have already assessed the impact of data injection attacks or DoS scenarios on power system stability, these studies are primarily carried out in isolated domains within control simulations or in simplified forms of networking models. The lack of real-time cyber-physical testbeds to carry out coordinated cyberattacks and assess possible integrated protection measures at both the control and communications levels renders their assessment theoretically.

For example, adaptive EMS algorithms that utilize both packet inspection (network level) and anomalies in the power system’s state (system level) to derive conclusions on data anomalies are only theoretical ideas. In addition, the lack of defined performance metrics and measurement approaches leads to difficulties reproducing and comparing the findings of various studies. While some rely on basic metrics such as delay, packets received, and packet loss, researchers’ methods to produce and measure these metrics, their time frames, and the types of underlying networks of each study vary substantially.

Table 5. Strengths, Limitations, and Trade-Offs of Current Modeling Approaches.

Approach	Strengths	Limitations	Key Trade-Offs/Design Considerations
Power System Simulations (e.g., MATLAB, DIgSILENT)	High fidelity power system modeling; control validation	Communication modeled simplistically Lacks realistic traffic behavior	Fidelity in control vs. realism of network behavior
Communication Network Simulations (e.g., NS-3, NetSim, CORE)	Detailed network performance metrics; protocol-level analysis	Power system effects are not modeled It ignores dynamic control interactions	Accurate QoS vs. ignoring power dynamics
Co-simulation Platforms	Combines both domains; enables cyber-physical analysis	Complexity in synchronization Limited number of MGs supported Limited open-source tools for multi-layer integration	Completeness vs complexity; scalability vs. fidelity

offers an overview of modeling approaches categorized by type, summarizing the benefits and shortcomings. It compares modeling approaches, revealing necessary trade-offs in design choices. Power simulation provides inherently high-fidelity control and load modeling capability; however, it makes many assumptions about communication behavior. On the other hand, network simulation provides an exhaustive measurement of QoS. Still, it does not assess the impact of the interactions between control systems and physical state, including voltage, frequency, or load distribution.

Hardware-in-the-loop (HIL) systems are a good example of bridging this divide. HIL systems inherit a complex real process by including hardware and software systems. However, these systems are typically expensive and difficult to scale when many MMGs are operated. Although they are essential to the functioning of MMGs, network topology and traffic changes are often neglected. The actual physical structure of MMG systems, with many routers, switches, and LANs connected in hierarchical or meshed systems, is frequently not fully represented in point-to-point or even static topologies commonly used in many studies. Work to model such behaviors using NetSim in [47], but not as a whole feedback mechanism with control layers.

Moreover, researchers in the literature studies on co-simulation platforms often work with a limited number of MGs, making it difficult to gain insight into model hierarchical control and communication interactions in MMG systems. The computational burden also increases in platforms as the number of MGs increases. Finally, open-source simulation tools representing full multi-layer integration are often limited, highlighting practical challenges with full-scale MMG modeling.

Figure 4 highlights the key research gaps and challenges in the cyber-physical investigations for MMG systems. Additionally, Figure 3 and Table 3, Table 4 and Table 5 provide the next research priorities in cyber-physical MMG operation, as follows:

• To work with different MMG designs and real network emulation, real-time co-simulation frameworks should emphasize open access tools based on scalable, modular structures.
• Communication-aware EMS and control algorithms should contain less sensitive procedures to time-varying and QoS variations to decrease delay-induced oscillations or data loss. Suggestions for consideration should be adaptive gain tuning and fault-tolerant logic.

Introducing security-by-design strategies incorporating cybersecurity procedures in the control systems and communication networks will allow governments to manage human-induced variants alongside renewables as attacks evolve.

4.1. Challenges in MMG Co-Simulation

Although MMGs are expected to improve resiliency and flexibility, extending the current single MG co-simulation platforms to MMGs might bring technical challenges. Managing the cyber-physical MMG system that integrates multiple domains, as summarized in Figure 5. From the review sections and the highlighting of research gaps, the following challenges are identified:

A.

Hierarchical Coordination and Scalability:

• Existing co-simulation platforms have been developed mainly to support a single MG, so these are not scalable for multiple interconnected MGs.
• Hierarchical control schemes rely on synchronized updates across multiple MGs, which is challenging under the constraints imposed by real-time communication.

B.

Real-Time Synchronization Between Power and Communication Layers:

• Synchronizing network emulators with power system simulators can be a challenging task for MMGs, where multiple control layers operate concurrently
• Simulations that are not synchronized can mask fast transient events, jitter bursts, and packet loss that can have significant implications for hierarchical control loops.

C.

Communication QoS and Dynamic Traffic Loads:

• MMG co-simulation must consider various latency, jitter, and packet loss profiles across multiple LANs or communication paths.
• Static traffic assumptions or idealized networks may cause controllers’ performance to be overestimated.

D.

Integration of Cybersecurity Aspects:

• Co-simulation methods in MMG should include cyber-attack scenarios.
• Incorporating adaptive detection and mitigation strategies across multiple MGs in real-time introduces an additional layer of complexity.

E.

Resources and Computational Considerations:

• Conducting standardized co-simulation of multiple MGs generates enormous computational overhead to capture power dynamics, communication, and security monitoring in real time.

F.

Standardization and Reproducibility:

• There is a need for evaluative metrics for MMG co-simulation that have foundations from accepted standards.
• Methodical differences in control models, communication capabilities, and simulation capabilities can introduce challenges in comparability across studies.

Tackling these technical challenges is vital for advancing the research in MMG co-simulation. Future work should focus on developing co-simulation frameworks that can scale, incorporate hierarchical control, packet-level communication dynamics, cybersecurity strategies, and real-time operating performance validation to create continuity between single-MG simulations and the performance of practical, resilient MMG systems.

4.2. Synthesis and Design Trade-Offs

Overall, power-only simulation often overestimates the robustness of control. However, in-network-only simulation often overestimates the sufficiency of QoS, and co-simulation, which is not real-time or packet accurate, risks smoothing away the very transients that ultimately destabilize secondary control. Accordingly, closing these loops will require interdisciplinary collaborations for designing integrated frameworks and platforms that consider security, communication, and power.

Furthermore, modeling more realistically how the MMG system will behave in the real world, including consideration of cyber-physical interaction, necessitates more accurate representations of network dynamics, traffic patterns, and attack scenarios. Taken together, Table 3, Table 4 and Table 5 highlight the following important patterns:

• The studies emphasizing control/communication often did not address either domain explicitly.
• The assumptions of realism relevant to QoS modeling can strongly dictate the relevance of the reported results.
• Simulations incur trade-offs between depth, scale, and performance fidelity.

The breadth of design trade-offs underscores our critical literature analysis and presents opportunities to identify open challenges that serve as direct motivation for future research, as outlined in Section 5.

5. Future Research Directions

The focused review presented in this paper illustrates the current advancements in the MMG systems regarding their operation and control. However, it highlights critical research gaps and challenges regarding cyber-physical interdependence analysis for the MMG infrastructures. Accordingly, future work should address the identified research gaps and challenges to develop more resilient, adaptive, and scalable MMG infrastructures. In this section, we highlight the perspectives on some potentially promising research directions.

5.1. Software-Defined Networking for Controllable Communication

Traditional communication structures present static network architectures in MMGs. However, static architecture cannot support dynamic traffic patterns, diverse QoS, and fault recovery. Software-Defined Networking (SDN) enables centralized control with a programmable structure, allowing for flexibility in response to the dynamics of MMGs by dynamically allocating bandwidth, rerouting traffic in the event of failures, and prioritizing control signals when appropriate, over less critical data.

Thinking about MMGs from a cyber-physical perspective, SDN embedded in MMG control systems enables communication-aware energy management, where the network and control actions evolve together in real-time. For example, SDN controllers can reroute packets communicating control commands during congestion, keeping devices in sync and avoiding destabilizing the power system. Additionally, SDN-enabled monitoring provides visibility into flow-level activity, allowing for anomalies to be identified and automated responses to stop cyberattacks before they degrade the quality of service.

Regarding feasibility, SDN integration into MMGs is a near-term direction, as programmable switches, controllers, and protocols are already reasonably mature in networking research. However, careful consideration of single points of failure in the MMG must be accomplished to help ensure reliable adoption.

5.2. Digital Twins for Cyber-Physical MMGs

Digital twins (DTs) are a virtual representation of physical microgrid elements, including real-time measurements from sensors, communications, and other relevant data. In multi-microgrid contexts, DTs can be used as live testbeds to evaluate control policies, predict failures caused by the network, and schedule delivery of resources. The ability to exchange physical states (voltage, frequency, load) for cyber states (delay, jitter, packet loss) is where the actual value of DTs lies.

For example, a DT could run simulations of how communication delay propagates from network congestion to a delayed secondary control command and possibly verge on microgrid instability. If we can forecast these incidents, we could act well before they occur. Furthermore, utilizing DTs and AI optimization features allows operators to test new configurations or control policies in the twin before committing to real-time operations, ultimately lessening risk.

Although there are DT frameworks for industrial IoT and for single MGs, the use of DTs with MMGs means incurring computational overhead and the risks of model drift, so we set DTs as a medium-term research objective requiring advances in synchronization methods and computing platforms appropriate to scaling before we can consider full use in large MMG infrastructures.

5.3. Physics-Informed Cybersecurity Schemes

Cybersecurity remains one of the most crucial aspects in the operation of cyber-physical MMG systems, with the increase in reliance on communication infrastructures for facilitating the operation, monitoring, protection, and control of these smart energy systems. Future research should shift its focus away from conventional intrusion detection and develop physics-informed methods. Physics-informed detection methods use knowledge of physical laws and grid dynamics, and even control loops, in the detection algorithms, such that stealthy cyberattacks that mimic normal traffic patterns would still be detectable as no longer obeying the physical behavior.

For instance, a physics-informed detection system could detect malicious data injections if the proposed control actions conflicted with the droop characteristics or violated fundamental power balance equations. These methods could be even more potent by augmenting them with data-driven approaches, such as anomaly detection using machine learning, thereby creating hybrid detection methods that balance robustness and adaptability.

Recent works have provided practical frameworks to implement such schemes in microgrids and related cyber-physical systems. Ref. [62] proposed a Hybrid Physics-Informed Neural Network (HPINN) that combines linear Kalman Filters with neural networks to detect and defend against false data injection attacks (FDIs) in DC microgrids. The proposed scheme enhances their state estimation and ensures the control remains stable, even when the sensors use compromised states. The results were verified through online experiments.

In addition, ref. [63] created a Physics-Informed LSTM (PI-LSTM) with federated learning to identify cyberattacks on Virtual Synchronous Machines (VSM). The proposed PI-LSTM detected anomalies with few false positives, as it incorporated physical swing dynamics and control constraints into the proposed LSTM architecture, enabling online inferences. Furthermore, ref. [64] developed HPINN for sensor attack detection in DC-DC converters, providing a state estimator and NN classifier in a hybrid form. The hybrid procedure allows real-time detection and accommodation for control measures on low-latency embedded hardware for sensor-targeted FDI attacks. These applications show that data-driven models enable robust and practical cyberattack detection.

Future work may involve pushing these physics-informed detection schemes into MMG systems, using real-time cyber-physical co-simulation with federated learning that can implement scalable and distributed cyberattack detection for interconnected microgrids. Cybersecurity based on physics is essential, but it is considered a long-term research priority. There are established proof-of-concept solutions, but for large MMGs, operational deployment will require rigorous validation. In addition to the need for standardized benchmarks, this is why it is currently not very feasible in the short term. Nonetheless, its importance in resilience means it is a strategic future research direction.

5.4. Integration of Artificial Intelligence and Adaptive Control

There is a vital need for intelligent control strategies that can adapt to real-time changes in traffic load, renewable generation variability, and cyber threats within the MMG infrastructures. Artificial intelligence (AI)-based adaptive control utilizing adaptive control methods based on reinforcement learning or federated learning could allow for adaptive control of MMG controller parameters like droop coefficients or secondary PI gains, in real time, based on changes in quality of service.

Furthermore, concerning federated learning, MMGs could allow for collaborative training of models with federated learning across distributed datasets, without compromising sensitive local datasets, enhancing accuracy and privacy.

Accordingly, adaptive control driven by AI can be regarded as a medium-term goal. There are significant advancements in reinforcement learning and federated learning architectures. However, issues with system stability, explainability, and operator trust still need to be addressed before large-scale MMG deployment. Its feasibility depends on creating hybrid solutions integrating AI with physics-based safety constraints.

5.5. Cyber-Physical Frameworks and Co-Simulation Platforms

Finally, advancing MMGs requires scalable co-simulation platforms integrating real-time power system simulators, network simulators, and cybersecurity modules. The platforms that allow simulated MMGs should be modular to accommodate a wide range of architectures, facilitate large-scale studies of MMGs, and allow fault injections to fully stress test the MMG’s response under realistic cyber-physical attack conditions.

Furthermore, utilizing open-source software and tools to develop these cyber-physical power systems’ platforms. This could provide more opportunities for enhancing this research direction and improving the reproducibility of state-of-the-art research. Creating a co-simulation infrastructure on cloud platforms could further democratize access by removing the burden of hardware on individual labs and enabling cooperative experiments across institutions.

Given the capabilities of the developed cyber-physical frameworks [43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61], resilient co-simulation platforms are a higher priority in near-term, tangible projects. Co-simulation platforms provide direct access for researchers to assess MMG resilience, test control strategies, and superimpose threats involving cyber-physical attacks. Aside from cyber-physical attacks, the primary feasibility concerns are overcoming the complexity of integration and balancing fidelity against scalability.

5.6. Prioritization and Feasibility Roadmap for Future Research Directions

By synthesizing the points raised in Section 5, it becomes evident that future research on cyber-physical MMG systems would need to look beyond bounded control or communication modeling. Rather, evolving towards comprehensive, cross-layered approaches that interconnect power system dynamics, communication network dynamics, and security dynamics.

Using concepts like SDN and DTs enables greater flexibility, increased visibility, and real-time organization of increasingly diverse network behaviors. Similarly, including physics-informed methods in detection-and-mitigation methods bridges cyber data analytics with physical system knowledge, building on resilience.

All these research directions are significant, but their effectiveness will ultimately depend on feasibility and timing. Because they utilize existing tools and infrastructure, SDN-enabled MMGs and robust co-simulation platforms will be the most demonstrable areas in the near future. The viability of MMG intelligence will be advanced in the mid-term by DTs and AI-enabled adaptive control; nevertheless, this will necessitate technological breakthroughs in explainability, stability guarantees, and synchronization. Though a long-term research subject needing substantial validation, benchmarking, and hardware support before being widely deployed, physics-informed cybersecurity is equally significant. This layered perspective on future research transforms the proposed futures section into a reframed roadmap.

Table 6. Summary of Future Research Directions for Cyber-Physical MMGs.

Research Direction	Key Contributions	Technical Challenges	Trade-Offs/Design Considerations
SDN-enabled MMGs	Dynamic bandwidth allocation, traffic rerouting, QoS-aware control	Controller reliability, scalability of SDN for large MMGs	Flexibility vs added control-plane complexity Centralized design may create single points of failure
Digital Twins	Real-time prediction, safe testing of new control policies, and security for cyber-physical systems	Data synchronization, computational overhead	High visibility vs. resource demands Strong predictive capacity, but risk of model drift if twins are not updated continuously
Physics-Informed Cybersecurity	Robust detection of stealthy attacks, hybrid physical-data defense	Model accuracy, false positives	Strong resilience vs. potential increase in false alarms Physics rules improve detection but limit adaptability
AI-driven Adaptive Control	Real-time parameter tuning, federated learning for privacy	Stability guarantees, explainability	Adaptability vs interpretability Performance gains may come at the cost of transparency and operator trust
Resilient Co-Simulation Platforms	Modular testing environments, cloud-based access	Integration complexity, resource demands	Comprehensive analysis vs. high implementation effort Scalability often conflicts with real-time fidelity

summarizes the future research directions by mapping the challenges to solutions or potential tools for consideration based on the literature review. This summary emphasizes the variety of new modalities currently being investigated and the need for a single framework that integrates modularity, adaptability, and robustness for MMG systems.

6. Conclusions

The evolution of multi-microgrid (MMG) systems will fundamentally change modern power systems’ design, operation, and resiliency. MMGs use communication networks to operate and control, becoming cyber-physical systems. The resilience and reliability of the physical power infrastructure are related to the resilience and QoS of the supporting communication infrastructure. This paper presents a focused review of MMG systems from a cyber-physical viewpoint that addresses control mechanisms, communication network emulation, co-simulation environments, and cybersecurity considerations. The review highlights that while hierarchical, distributed, and decentralized control mechanisms have been thoroughly studied, insufficient attention has been paid to performance in realistic communication dynamics such as latency, packet loss, or throughput constraints. Furthermore, while the number of studies employing co-simulation for microgrids is increasing, relatively few studies consider communication QoS metrics in evaluating operation and control, or the connectivity between microgrids. Cybersecurity challenges like those seen with coordinated cyber-physical attacks, especially in real-time high-traffic contexts related to resilience, are still explored. Still, the critical area of cybersecurity and resilience needs further exploration.

The previously identified research gaps highlight the need for adaptable and scalable frameworks that can simultaneously analyze the control, cyber, and physical dynamics in real-time. The future research directions outlined in this paper, including physics-based detection and mitigation of coordinated attacks, DTs for prediction-based real-time operations, and SDN for adaptive communication, represent practical approaches to addressing these challenges. Additionally, integrating AI-driven adaptive control and resilient co-simulation platforms presents a vital opportunity for developing MMGs that adapt to changing circumstances and threats. Most importantly, these proposed developments reinforce the potential for MMGs to evolve into increasingly intelligent, resilient, and adaptable systems, enabling the transition towards decentralized, resilient, and sustainable energy systems. Accordingly, this review is a roadmap for designing next-generation cyber-physical MMG infrastructures.