Real-Time Energy System Optimization and

Full text

Turn on search term navigation

1. Introduction

1.1. The Context of the Energy Transition and the Role of PV–BESS Microgrids

The surging demand for electricity, the large-scale integration of renewable energy sources (RESs), and the increasing pressure of decarbonization policies are profoundly transforming the operational paradigm of energy systems. Low-voltage distribution networks, originally designed for unidirectional power flow, are currently challenged to integrate distributed generation, flexible loads, and storage systems, while simultaneously maintaining power stability and quality [1].

In this context, microgrids powered predominantly by renewable sources—particularly hybrid configurations of Photovoltaics and Battery Energy Storage Systems (PV–BESSs)—are emerging as pivotal components of resilient energy grids. Their capability to operate in both grid-connected and islanded modes grants them a crucial role in preventing blackouts, enhancing local energy security, and ensuring supply continuity for critical loads (e.g., critical infrastructure, public buildings, and residential neighborhoods).

Nevertheless, the high penetration of PV generation, the load variability at the low-voltage level, and the complex interaction with storage systems render PV–BESS microgrids more vulnerable to rapid disturbances. These disturbances stem from both the physical domain (faults, voltage fluctuations, phase imbalances) and the digital domain (communication failures, configuration errors, cyber-attacks). Under these conditions, intelligent monitoring and real-time assessment of stability and resilience become as critical as the proper sizing of power components [2].

1.2. Specific Challenges in Monitoring Low-Voltage PV–BESS Microgrids

Low-voltage networks, particularly in urban and peri-urban areas, are characterized by:

Complex topologies featuring multiple branches;

Unbalanced and time-varying loads (households, prosumers, and electric vehicle (EV) charging stations);

A high density of potential measurement points, subject to strict cost and communication constraints.

In this environment, stability and blackout resilience cannot be assessed solely based on centralized measurements at the substation or Point of Common Coupling (PCC). To effectively address these challenges, the following are required:

Distributed sensors installed in close proximity to critical nodes (PV systems, BESS, coupling points, and feeder endpoints);

The capability for early detection of voltage and frequency deviations, as well as the identification of pre-fault patterns;

Lightweight predictive models running at the edge (at the IoT device level) to anticipate instability risks and potential Energy Not Supplied (ENS) events.

Classical monitoring approaches, relying on SCADA systems or data concentrators with aggregation intervals of 15–60 min, are insufficient to capture the fast dynamics inherent to microgrids integrating PV and BESS. Furthermore, transmitting all raw data to a central server introduces significant challenges regarding scalability, latency, and—crucially—cybersecurity.

To bridge this gap, the proposed architecture introduces a distributed, deterministic Intelligent Edge Processing (IEP) layer. Figure 1 illustrates a comparative timeline of the response latencies between the proposed solution and existing alternatives. While cloud-based analytics offer high processing power, they function as “black-box” systems with non-deterministic latencies. In contrast, our deterministic edge logic ensures a superior trade-off by providing:

Latency: Sub-millisecond local processing compared to the multi-second round-trip times of cloud or SCADA cycles, enabling a sub-300 ms global response time;

Explainability: Transparent, rule-based decision-making based on physical network limits and EN 50160 standards [3], allowing operators to identify the exact cause of a control action;

Robustness: Continued operation during network partitions or communication failures, as the decision logic resides locally at the IoT node level.

Consequently, there is an emerging need for intelligent IoT systems capable of combining:

Local measurements of voltage, current, power, and frequency;

Edge processing (encompassing feature extraction, local VDI/ENS indicators, Rate of Change of Frequency (RoCoF), and stability margins);

Selective and robust communication with a monitoring center or an Energy Management System (EMS).

1.3. The Role of IoT Systems and Intelligent Edge Processing

The evolution of embedded platforms (e.g., ESP32, ARM Cortex-M, Single Board Computers), combined with modern communication protocols (MQTT, OPC UA, and IEC 61850-based profiles), enables the deployment of IoT sensor networks capable of functioning as a “distributed intelligence” layer overlaying the electrical infrastructure, through the concept of Intelligent Edge Processing (IEP) [4].

Recent advancements in communication frameworks, such as Time-Sensitive Networking (TSN) and 5G Ultra-Reliable Low-Latency Communications (URLLC), offer promising high-performance capabilities for smart grid control. However, their large-scale implementation in low-voltage residential networks is often hindered by high deployment costs and infrastructure complexity. In this context, recent studies emphasize the shift toward decentralized energy management to mitigate communication latencies [5], particularly when integrating battery storage systems [6].

Consequently, the proposed architecture utilizes a combination of MQTT and Modbus protocols. While these are traditional IoT protocols, when coupled with the edge-centric decision-making of the IEP layer, they provide a robust and cost-effective alternative that ensures sub-300 ms response times. This approach minimizes the critical dependency on high-bandwidth backbone connectivity and maintains system stability even during network congestion or partial communication failures.

In this vision:

Autonomous Virtual Instrumentation: Each IoT node becomes an autonomous virtual instrument capable of locally calculating critical grid stability indicators (e.g., dynamic voltage variation, RoCoF, V–Q margins) and operational resilience metrics (local recovery time, frequency of undervoltage/overvoltage events);

Data Optimization: Data is not transmitted in its raw state; instead, it is sent as pre-processed indicators, hierarchical alerts, and risk scores, thereby optimizing data flow and bandwidth usage;

Low-Latency Execution: Real-time evaluation algorithms and adaptive threshold logic can run directly on resource-constrained devices, ensuring low latency and eliminating the critical dependency on centralized computing power.

This IoT-centric approach complements classical power system analysis methods. While the physical microgrid model (based on the CIGRE LV network or other benchmarks) provides the electrical reference context, the Intelligent Edge Processing (IEP) layer ensures proactive monitoring (“intelligent monitoring”), rapid anomaly detection (Fault Detection and Diagnostics—FDD), and immediate decision support for blackout prevention.

1.4. Scope and Contributions of the Paper

Given the challenges associated with the massive integration of distributed renewable energy sources into low-voltage networks—particularly the high variability of photovoltaic generation and the risk of pre-instability situations—the primary aim of this work is to propose and evaluate a distributed IoT architecture capable of ensuring real-time monitoring and assessment of the stability and resilience of PV–BESS microgrids [5]. The study employs a CIGRE European LV benchmark reference model, adapted to reflect the fast dynamics of modern networks.

In this context, the paper addresses the following research questions:

How can a distributed IoT layer be designed to provide granular visibility over stability indicators in a low-voltage microgrid, overcoming the limitations of classical SCADA systems?

To what extent can deterministic local processing (Intelligent Edge Processing) detect dynamic anomalies (RoCoF, VDI) at an early stage and contribute to preventing local blackouts, compared to strictly centralized or “black-box” approaches?

The main contributions of this work are:

Explainable and Deterministic Edge Intelligence: A deterministic, rule-based Intelligent Edge Processing (IEP) logic is introduced and validated on resource-constrained ESP32 nodes. Unlike cloud-dependent or “black-box” machine learning models, this approach offers a robust and explainable alternative for safety-critical infrastructure, ensuring predictable response times in real-time grid control.

Integrated Simulation-to-Edge Validation Pipeline: The development of a holistic framework that combines a high-fidelity CIGRE LV electro-thermal model with realistic PV (IncCond MPPT) and BESS (P–f/Q–V) control strategies. This is seamlessly integrated with a full IoT communication stack (Modbus, MQTT, and Node-RED) to validate monitoring algorithms in high-fidelity scenarios.

Quantitative Resilience Performance: The study provides a quantitative demonstration of the architecture’s capacity to achieve stability recovery with response times of sub-300 ms. This directly links edge processing latency to operational metrics and EN 50160 compliance, proving the efficacy of edge computing for grid modernization.

Methodology for Real-Time Resilience Assessment: Specific quantitative indicators (VDI, RoCoF, and a normalized Resilience Index) are defined and implemented at the edge level. This enables the microgrid to numerically assess its capacity to absorb disturbances and return to an equilibrium state autonomously.

Validation under Realistic Stress Scenarios: The performance of the architecture is proven through extensive simulations of major disturbances (solar cloud transients, sudden overloads, and severe grid faults), highlighting its ability to prevent local blackouts through coordinated local action.

1.5. Organization of the Paper

The remainder of this paper is organized as follows:

Section 2 provides the theoretical background and reviews related work concerning PV–BESS microgrids, stability and resilience indicators in low-voltage networks, and the application of IoT systems for intelligent monitoring.

Section 3 describes the system model, including the PV–BESS microgrid based on the CIGRE European LV benchmark network, the load and generation profiles, and the proposed architecture for IoT sensors and edge processing.

Section 4 details the materials and methods employed, covering the defined stability and resilience indicators, the evaluation methodology, and the simulation scenarios.

Section 5 presents the numerical results and discussion, focusing on the interpretation of VDI, RoCoF, and RI indicators, and on the capability of the IoT–Edge–BESS architecture to achieve early detection of pre-instability conditions. Furthermore, practical implications for distribution operators are analyzed, and the limitations of the current study are discussed.

Section 6 summarizes the main conclusions of the paper, reaffirming the validity of the proposed approach and outlining major directions for future research, such as experimental validation and Hardware-in-the-Loop (HIL) testing.

2. System Modeling

2.1. The CIGRÉ LV Network—Structure, Parameters and Topology

The CIGRÉ Low Voltage (LV) network represents one of the most widely utilized benchmark models for evaluating microgrid performance, particularly in the context of stability analysis, renewable energy integration, and distributed monitoring via IoT systems. Developed by the CIGRÉ C6 Working Group, the model provides a realistic representation of a European urban low-voltage network, incorporating both structural characteristics and the typical dynamics of residential loads.

The CIGRÉ LV model serves as a standard for analyzing European LV distribution grids and is extensively used in studies dedicated to stability, power quality, and the integration of distributed renewable sources. The network features a radial structure and is supplied by a 20/0.4 kV MV/LV transformer with an ONAN (Oil Natural Air Natural) cooling type and a rated power of 400 kVA, reflecting the typical configuration of urban distribution infrastructure in Central Europe.

On the low-voltage side, the model includes:

A main feeder equipped with eight secondary branches (sub-feeders), configured to reproduce real-world network topologies;

18 LV nodes, distributed along underground cable segments, imparting the model with the actual characteristics of modern urban grids;

Varied residential loads, distributed across both single-phase and three-phase connections to reproduce natural demand imbalances;

Predefined connection points for Photovoltaic (PV) system integration, aligned with real European renewable energy penetration scenarios;

Optional Electric Vehicle (EV) charging points, included in the CIGRÉ European LV Benchmark Grid—2014 version, allowing for the analysis of emerging network stress conditions.

This structure enables the simultaneous analysis of multiple operational phenomena:

Dynamic power flows in the presence of rapid load and generation variations;

Voltage deviations under uneven loading conditions;

Phase imbalances caused by the asynchronous distribution of single-phase loads;

PV–BESS–Load interactions impacting local stability;

Effects of distributed storage on power quality;

System behavior under disturbances, including local voltage dips, branch congestion, and transformer overloads.

Due to these features, the CIGRÉ LV network becomes an ideal framework for testing advanced IoT monitoring algorithms, specifically:

RoCoF (Rate of Change of Frequency) analysis;

Evaluation of Volt/Var and P–f stability margins;

Rapid detection of nodal deviations;

Implementation of deterministic intelligent edge processing mechanisms based on electrical indicators and operational thresholds, deployed on embedded platforms such as ESP32, STM32, and ARM Cortex-M.

Furthermore, the use of a European reference network, as opposed to North American-inspired IEEE models, allows for algorithm validation within a framework consistent with the requirements of the EN 50160 standard [3]. This compatibility enhances the practical relevance of the results for Distribution System Operators (DSOs) and the energy infrastructure within the European Union.

The CIGRÉ LV model is defined by a rigorous set of electrical parameters that faithfully reflect the configuration of a European urban distribution network. These parameters include the characteristics of the MV/LV transformer, the low-voltage lines, the nodes with diversified loads, and the integration points for distributed sources. The parameter values utilized in the simulations for this paper are synthesized in Table 1.

Each element of the network—from cable impedances to transformer losses and dynamic consumption profiles—is standardized within the CIGRÉ documentation, ensuring model reproducibility and the comparability of results with other experimental and analytical studies.

The following section presents the most relevant technical characteristics of the grid, which constitute the basis for the simulations performed in this work.

Network Infrastructure. The coupling point with the distribution system is established via a 400 kVA MV/LV transformer (20/0.4 kV, Dyn11), modeled with a short-circuit impedance of 4%, no-load losses of 0.7 kW and load losses of 5.1 kW. The low-voltage network, spanning a total length of 900 m, exclusively employs NAYY 4 × 150 mm² underground cables (aluminum conductor). These are characterized by the parameters R = 0.206 $Ω$ /km and X = 0.080 Ω/km, standardized values that allow for a faithful analysis of voltage drops and the influence of impedances on dynamic power flows.

Load Modeling and Consumption Profiles. Residential consumers connected to the 18 nodes are modeled using a static ZIP characteristic (constant Impedance–Current–Power) to capture the dependency of consumption on voltage variations. The temporal profiles used correspond to the European “Evening Peak” curve, with maximum demand occurring between 17:00 and 22:00. A critical aspect of the model is the asymmetrical distribution of single-phase and three-phase loads, which generates natural phase imbalances (voltage and current). This particularity provides a realistic environment for testing the sensitivity of IoT algorithms to non-ideal operating conditions, in contrast to simplified, perfectly balanced models.

Integration of Distributed Energy Resources (PV and BESS). To analyze scenarios specific to active networks, the standard CIGRÉ model was extended by integrating sources in Zone C (downstream), a critical point regarding voltage stability. The configuration includes:

Photovoltaic Unit (PV): Single-phase (P_nom < 5 kW), connected at Node 15, operating with variations dependent on irradiance and temperature.

Battery Energy Storage System (BESS): Three-phase (P_nom = ±2.5 kW, C = 30 kWh), integrated at Node 16. The positioning adjacent to the PV source facilitates the damping of local fluctuations and ensures voltage support (Volt–Var control) in the most vulnerable area of the network.

Network Topology

The adopted topology of the CIGRÉ LV network is illustrated in Figure 2 and reflects the typical radial configuration of urban low-voltage distribution.

The main feeder is structured into three distinct operational sectors, defined according to load typology and the location of distributed sources:

The first sector (Zone A), corresponding to nodes 1–4, is characterized by the predominant presence of single-phase loads, resulting in a higher degree of phase imbalance.

The second sector (Zone B), corresponding to nodes 8–12, includes mostly three-phase loads and represents the zone with the best load distribution.

The downstream sector (Zone C), comprising nodes 14–18, is the most suitable for integrating distributed sources; in this work, it is used for connecting the PV unit at Node 15 and the BESS at Node 16, due to its strategic position in the network.

This segmented structure allows for a precise analysis of the influence of asymmetrical load distribution on voltage stability and power flows, offering an appropriate framework for evaluating the performance of an IoT system dedicated to monitoring the resilience of PV–BESS microgrids.

2.2. Photovoltaic (PV) Unit Model

Modeling photovoltaic sources within low-voltage microgrids is a fundamental aspect of analyzing system stability and dynamic behavior.

In this work, the PV units integrated into the CIGRE LV network are modeled based on a realistic single-stage architecture featuring a single-phase inverter. This configuration complies with the EN 50530 and IEC 61727 standards [7] regarding the quality of power injected into European grids.

The photovoltaic unit is represented by a high-fidelity model that includes the following subsystems:

The Solar Array: Modeled based on the non-linear Current–Voltage (I–V) characteristic;

MPPT Control: Utilizes the Incremental Conductance algorithm for Maximum Power Point Tracking;

DC–AC Converter: A single-phase inverter synchronized with the grid via a Phase-Locked Loop (PLL);

Control Loops: Dedicated cascading loops for regulating grid current and voltage.

2.2.1. Photovoltaic Generator

The photovoltaic generator is modeled using the single-diode model, widely adopted in microgrid literature due to its favorable trade-off between physical accuracy and computational complexity. This model enables a precise description of the dependency of the PV module’s electrical characteristics on environmental conditions, particularly solar irradiance and cell temperature. The model parameters are scaled such that the total rated power of the PV unit is approximately 3.2 kW under Standard Test Conditions (STCs) (1000 W/m², 25 °C). This value is representative of typical European residential PV installations connected to low-voltage networks and is consistently applied across all simulation scenarios analyzed in this work. The electrical behavior of the PV generator is characterized by non-linear I–V and P–V curves, which highlight the existence of a Maximum Power Point (MPP). The position of the MPP varies dynamically as a function of environmental conditions, directly influencing the active power available for injection into the microgrid and the stability of power flows [8].

The active power generated by the PV array is computed using a standard irradiance–temperature scaling referenced to standard test conditions:

(1) $T_{c e l l} (t) = T_{a m b} + \frac{G (t)}{800} \cdot (N O C T - 20)$

where

T_cell(t) denotes the photovoltaic cell temperature at time t [°C];

T_amb represents the ambient temperature [°C];

G(t) is the incident solar irradiance at time t [W/m²];

NOCT (Nominal Operating Cell Temperature) is the nominal operating temperature of the cell (typically °C), while the constant 800 [W/m²] represents the reference irradiance used for NOCT determination.

This relationship captures the natural heating of the PV modules under real-world operating conditions. An increase in temperature leads to a reduction in the MPP voltage and, consequently, to a degradation of the available output power—a phenomenon that is critical for the dynamic stability analysis of the microgrid. The single-diode model is employed to:

Define the characteristic I–V and P–V curves;

Demonstrate the influence of irradiance and temperature on the Maximum Power Point (MPP);

Calibrate the maximum level of active power injected by the photovoltaic source into the microgrid.

In microgrid-level dynamic simulations, the temporal variation in the photovoltaic power, P_PV(t), reproduces the effects of irradiance and temperature fluctuations on the MPP without explicitly incorporating the complex electro-physical equations of the PV cell into the inverter control loops. This approach is widely adopted in stability and resilience studies, facilitating the efficient analysis of PV–BESS–grid interactions.

The electrical parameters of the photovoltaic generator employed in the simulations are summarized in Table 2.

Figure 3 depicts the I–V curve (blue) and the P–V curve (red), obtained for a standard 400 W module under Standard Test Conditions (STC). The Maximum Power Point (MPP) is clearly observable, corresponding to the point where the current–voltage product is maximized and the module operates at optimal efficiency.

To capture the actual dynamics of the photovoltaic source, realistic temporal profiles for solar irradiance and cell temperature were adopted. These are depicted in Figure 4 and Figure 5.

Figure 4 illustrates the daily solar irradiance profile, G(t), expressed in W/m². The typical “bell-shaped” curve characteristic of a clear-sky day is observable, exhibiting peak values around noon (12:00).

Figure 5 illustrates the evolution of the photovoltaic cell temperature, T_cell(t), which is influenced by both the ambient temperature and solar irradiance. The rise in temperature negatively impacts the terminal voltage, resulting in a slight shift in the Maximum Power Point (MPP).

The I–V and P–V curves, correlated with the temporal evolution of G(t) and T(t), represent more than mere static results; they constitute a crucial element for understanding the behavior of the PV source within a dynamic system. This model facilitates the assessment of the impact of rapid solar resource fluctuations on microgrid stability, including the system’s response to abrupt variations in irradiance or temperature—aspects that are essential to the study of energy resilience.

2.2.2. MPPT Algorithm and the Selection of the Incremental Conductance Method

To maximize the energy efficiency of a photovoltaic generator, it is essential that the operating point aligns as precisely as possible with the Maximum Power Point (MPP), which varies continuously depending on atmospheric conditions. In this context, modern power conversion systems employ Maximum Power Point Tracking (MPPT) algorithms capable of a rapid response to irradiance and temperature variations. Within the proposed model, the inverter connected to the photovoltaic generator employs the Incremental Conductance (IncCond) algorithm, selected for its superior performance compared to conventional methods such as Perturb and Observe (P&O). Unlike P&O, which exhibits steady-state oscillations around the MPP and may incur power losses under rapidly changing irradiance conditions, IncCond utilizes the derivative dP/dV of the P–V curve to precisely determine the MPP position, as described by Equation (2):

(2) $\frac{d P}{d V} = \frac{d (I V)}{d V} = I + V \frac{d I}{d V}$

The MPP condition is met when the derivative is zero:

(3) $\frac{d P}{d V} = 0 \Rightarrow \frac{d I}{d V} = - \frac{I}{V}$

This relationship (3) enables the algorithm to determine whether the system is operating on the left or right side of the Maximum Power Point (MPP) and to adjust the voltage accordingly, without introducing unnecessary oscillations. Maximum Power Point Tracking (MPPT) algorithms are employed to maintain the operation of the photovoltaic generator in the vicinity of this MPP. By mitigating active power oscillations from the photovoltaic source, the IncCond algorithm directly contributes to the reduction in voltage fluctuations within low-voltage microgrids. In contrast to the Perturb-and-Observe (P&O) method, the IncCond algorithm provides a faster and more stable response to irradiance fluctuations, effectively avoiding undesirable oscillations around the MPP. This behavior is illustrated in Figure 6, which compares the performance of the two algorithms under a scenario characterized by rapid power variations.

2.2.3. Performance Parameters of the MPPT Algorithm

For the applications analyzed in this study, the performance of the Incremental Conductance (IncCond) MPPT algorithm is characterized by the following key parameters:

Update Frequency: 1–2 kHz, enabling a rapid response to dynamic atmospheric conditions and fluctuations in the P–V curve. This frequency is sufficiently high to track irradiance variations without introducing significant delays in the inverter’s response.

MPPT Efficiency: >99%, consistent with the technical specifications of the employed inverter and aligned with values reported in the literature for IncCond-type algorithms. The efficiency level is further verified through MATLAB/Simulink R2023b simulations by comparing the effectively extracted power against the theoretically available maximum power.

Response Time to Sudden Irradiance Drops: <20 ms, ensuring the stability of the PV source and a relatively constant active power flow to the grid. A reduced response time limits the amplitude of voltage variations at the Low Voltage (LV) nodes and contributes to preventing the triggering of protections sensitive to voltage sags.

These parameters are relevant not only at the converter level but also within the broader context of the PV–BESS microgrid, as they define the dynamics with which the photovoltaic source responds to disturbances. In the subsequent sections, these characteristics are correlated with the stability and resilience indicators monitored via the proposed IoT system.

2.2.4. Relevance of the Photovoltaic System for IoT Monitoring

The implementation of an advanced Incremental Conductance (IncCond) algorithm within the photovoltaic inverter yields significant benefits for the monitoring and control of modern microgrids, particularly when integrated into an IoT (Internet of Things) architecture. Within such a framework, the PV inverter functions not merely as a power conversion element but also as a relevant data source for assessing the grid state.

Distributed IoT systems facilitate real-time data collection and analysis, with direct implications for reliability and energy efficiency. In this context, the utilization of the IncCond algorithm offers the following advantages:

Real-time monitoring of voltage and current variations caused by rapid changes in solar irradiance and PV cell temperature, contributing to the identification of deviations from the optimal operating regime of the PV panels;

Access to dynamic characteristics: It provides access to the dynamic I–V and P–V curves of the PV source, which are essential parameters for performance diagnostics, the detection of degradation phenomena (e.g., aging, partial shading), and stability assessment under transient regimes;

Power loss quantification: It allows for the determination of power losses associated with delays in tracking the Maximum Power Point (MPP). This information is critical for optimizing both local inverter control and microgrid energy management strategies (e.g., coordination with BESS or flexible loads).

By integrating this data into a distributed IoT infrastructure—based on smart sensors, MQTT communication protocols, and processing platforms such as Node-RED—real-time information acquisition, transmission, and preprocessing are ensured. At the edge level, this data can be utilized by Edge AI algorithms for rapid microgrid state estimation, early anomaly detection, and the anticipation of potentially critical situations. In parallel, aggregated data is transmitted to the cloud layer for storage, historical analysis, and the generation of energy reports.

This architecture enables the development of advanced functionalities, such as:

Preventive fault identification at the panel or inverter level;

Real-time alert generation in the event of deviations from normal behavior;

Adaptive microgrid operation based on meteorological conditions and load profiles;

PV production forecasting based on historical data series and the current behavior of the source.

Overall, the tight integration between MPPT, IoT infrastructure, and distributed analysis mechanisms contributes directly to enhancing microgrid resilience and supporting automated decision-making processes within smart distribution grids.

2.2.5. Photovoltaic (PV) Inverter Model

The photovoltaic inverter employed in this study is a single-phase, full-bridge topology, responsible for converting the Direct Current (DC) electricity supplied by the PV generator into Alternating Current (AC) compatible with the low-voltage grid [9,10]. The inverter serves as the interface element between the renewable source and the grid, playing a crucial role in both controlling the injected power and maintaining local voltage stability. The control architecture of the inverter is organized into multiple functional layers, the most critical being the grid synchronization system (PLL) and the regulation structure based on Proportional-Integral (PI) loops for current and DC-link voltage.

(a) PLL—Grid Synchronization

The Phase-Locked Loop (PLL) synchronization system ensures the locking of the inverter onto the phase and frequency of the voltage measured at the Point of Common Coupling (PCC). The PLL provides real-time estimates of the grid phase angle and frequency, information that is subsequently utilized in the injected current control strategy.

The primary functions of the PLL are:

Maintaining synchronism between the inverter and the grid under steady-state conditions;

Ensuring correct current injection in phase with the grid voltage;

Tracking dynamic frequency variations in the event of disturbances.

The role of the PLL becomes essential in the instability scenarios analyzed in Section 4, where rapid frequency variations and voltage imbalances occur.

(b) Current and DC-link Voltage Control

The control structure of the photovoltaic inverter is implemented based on two Proportional-Integral (PI) regulation loops:

The fast inner current loop, which ensures the tracking of current references and limits overcurrents during transient regimes;

The outer DC-link voltage regulation loop, which maintains the energy balance between the photovoltaic source and the AC side of the inverter.

This regulation structure enables the achievement of the following functional performance metrics:

Low Total Harmonic Distortion (THD) of the injected current (THD < 3%), in compliance with power quality requirements;

Controlled active power injection, based on the reference provided by the MPPT algorithm;

Capability for limited reactive power injection or absorption, to support voltage at the connection node (Volt–Var support function).

Through these functions, the PV inverter ceases to be merely a conversion element and becomes an active device for supporting voltage stability and power quality within low-voltage microgrids.

2.2.6. PV Power and Microgrid Operation Mode

Throughout the simulations, the photovoltaic unit connected to Node 15 of the CIGRE LV network (Zone C) operates in a power-controlled mode. This implies that the inverter injects the active power available at the MPPT algorithm output into the grid, based on instantaneous solar irradiance and cell temperature conditions. In this regime, the injected active power is dictated by the primary source (the PV generator), while the inverter is tasked with precisely tracking this reference while simultaneously ensuring compliance with grid-imposed voltage and current limits.

The power-controlled strategy is representative of the normal operating regime for prosumers connected to low-voltage networks, where the priority is to maximize locally generated energy and inject it safely into the grid. Within this framework, rapid variations in solar irradiance translate directly into fluctuations in injected active power; consequently, local voltage stability is directly dependent on the dynamics of the photovoltaic source.

Specifically, in the absence of supplementary damping mechanisms, sudden spikes in PV power can lead to local overvoltages, while abrupt drops can generate voltage sags or transient disturbances in power flows. These effects are particularly accentuated given the PV unit’s location in the downstream area of the network (Zone C), where line impedances are higher and voltage sensitivity to power variations is significant.

To prevent the degradation of local voltage stability, the inverter’s operation is complemented by automatic voltage limitation mechanisms at the connection node, as well as by Volt–Var support capabilities, as described in Section 2.2.5. Thus, under overvoltage conditions, the inverter can reduce active power injection or absorb reactive power, thereby contributing to the stabilization of the voltage profile.

In the disturbance scenarios analyzed in this paper—which include voltage sags, rapid frequency variations (high RoCoF), and transient load disturbances—the PV unit switches from the power-controlled regime to a power-hold regime. In this operating mode, the inverter temporarily maintains a near-constant level of injected power or gradually reduces injection, avoiding instantaneous disconnection from the grid. This behavior is essential for:

Preventing the loss of synchronism between distributed sources;

Avoiding cascading trips of protection systems within the microgrid;

Maintaining a minimum energy contribution during the critical phases of disturbances.

Maintaining the local coupling of the photovoltaic source in power-hold mode is coordinated with the Battery Energy Storage System (BESS) connected at Node 16, which can rapidly supply or absorb power to compensate for temporary imbalances between generation and demand. This PV–BESS interaction plays a fundamental role in enhancing microgrid resilience, allowing for the damping of disturbances without interrupting the supply to sensitive loads.

The dynamic response of the photovoltaic unit in both power-controlled and power-hold modes is illustrated in Figure 7, which presents the evolution of the active power injected by the PV source and the voltage at Node 15 during a sudden drop in solar irradiance. In this simulation, the PV cell temperature is maintained constant, as thermal effects manifest over much larger time intervals (tens of seconds to minutes), whereas the analysis focuses exclusively on the source’s rapid dynamic response to abrupt irradiance variations.

At time t ≈ s, the reduction in irradiance triggers a rapid decrease in photovoltaic power from approximately 5 kW to 2 kW. Consequently, the voltage at the connection node records a controlled deviation, temporarily dropping towards 220 V. Due to the inverter’s action and the voltage support provided by the BESS, no dangerous oscillations or instabilities occur.

Following the recovery of irradiance (t > 2 s), the MPPT algorithm returns the unit to its nominal operating regime, and the nodal voltage is rapidly restored to its initial value. This behavior confirms the capacity of the PV–BESS microgrid to respond in a stable and resilient manner to rapid generation disturbances.

Following the clearance of the disturbance and the restoration of grid parameters within nominal limits, the PV unit undergoes a controlled transition back to the power-controlled mode. This is achieved through a process of gradual resynchronization, supervised by the PLL and the inverter’s regulation loops. This controlled transition prevents the occurrence of power surges and transient overcurrents, thereby contributing to the short-term stability of the microgrid.

From the perspective of IoT-based monitoring, the two operating regimes (power-controlled and power-hold) are directly reflected in:

Rapid variations in active power, P;

Voltage dynamics at node 15;

Changes in local stability indicators (VDI, RoCoF, voltage margins).

These parameters are subsequently utilized in Section 4 to evaluate the capability of the proposed IoT architecture to achieve early detection of pre-instability situations and to support strategies for local blackout prevention.

2.3. Battery Energy Storage System (BESS) Model

The integration of Battery Energy Storage Systems (BESS) represents a pivotal element for enhancing the flexibility and resilience of microgrids supplied by renewable energy sources. In low-voltage microgrids characterized by a high penetration of photovoltaic generation, the BESS serves a fundamental role in buffering rapid power variations, supporting local voltage profiles, and ensuring supply continuity under disturbed operating conditions [11,12].

In this study, the BESS is integrated into the CIGRE LV network at node 16 (Zone C), in the immediate vicinity of the photovoltaic unit connected at node 15. This strategic positioning facilitates:

Rapid compensation of imbalances between generation and consumption;

Voltage support in the downstream area of the network, which is characterized by higher line impedances;

Direct cooperation with the photovoltaic inverter during power-hold and power-controlled transient regimes.

2.3.1. General Architecture of the BESS

In this work, the Battery Energy Storage System (BESS) is modeled as an integrated power assembly comprising the following main subsystems:

The electrochemical battery itself;

The bidirectional DC–AC electronic converter (storage inverter);

The Battery Management System (BMS);

Control loops for active power, reactive power, voltage, and State of Charge (SOC).

The BESS is connected to the low-voltage grid via a three-phase connection through a bidirectional inverter, enabling operation in both fundamental operational modes:

Discharge mode, wherein the system injects active and/or reactive power into the grid to support local consumption;

Charge mode, wherein the system absorbs surplus energy from the grid, particularly during periods of high photovoltaic production.

In this study, the BESS is assumed to be based on Li-ion technology, which represents the dominant solution in microgrid and distributed storage applications due to its high energy density, excellent cycling efficiency, long lifespan, and rapid response times for both charging and discharging.

The BESS is connected to the low-voltage network via a bidirectional inverter, and the analysis performed in this work is conducted at the AC equivalent level at the microgrid connection node. Detailed modeling of the DC-link bus and the converter’s internal power electronics is not explicitly detailed, as these aspects are implicitly incorporated into the inverter’s dynamic behavior and the operational limits for active and reactive power variation [13,14].

The Battery Management System (BMS) ensures continuous monitoring of the battery pack’s internal parameters, including cell voltages, currents, and temperatures, while also implementing protections against overvoltage, undervoltage, overcurrent, and overtemperature. Simultaneously, the BMS provides real-time critical information regarding the State of Charge (SOC), which is utilized by the BESS control algorithms and the microgrid energy management architecture.

Regarding network placement, the BESS is connected at Node 16 of the CIGRE LV network (Zone C), in the immediate vicinity of the photovoltaic unit connected at Node 15. This positioning is strategically selected to facilitate the rapid compensation of power fluctuations generated by the PV source, reduce transient power flows along distribution lines, and limit electrical stress on the MV/LV transformer.

The nominal power of the BESS is sized at approximately ±2.5 kW, a value comparable to that of the photovoltaic source connected at the adjacent node and representative of individual residential applications or low-voltage microgrids with distributed storage. This sizing allows for a realistic evaluation of the BESS’s capacity to damp local power disturbances and support voltage and frequency under transient conditions.

Through this architecture, the BESS becomes an active element of the PV–BESS microgrid, playing an essential role in energy balancing, voltage stabilization, and enhancing resilience under disturbed operating conditions.

The main parameters of the battery storage system used in the simulations are summarized in Table 3. The selected values are representative of low-voltage microgrid applications and ensure an appropriate trade-off between energy support capacity, dynamic response, and the physical limitations of the power converter and battery. These parameters are maintained constant across all analyzed test scenarios, enabling a comparative evaluation of the PV–BESS microgrid behavior exclusively under the influence of applied disturbances rather than changes in the storage configuration.

2.3.2. Dynamic Battery Model

From a mathematical modeling perspective, the battery is represented using a Thevenin equivalent dynamic model, which allows for the accurate capture of both steady-state behavior and the transient effects inherent to charging and discharging processes. The model comprises:

A voltage source dependent on the battery’s State of Charge (SOC);

A series internal resistance;

An equivalent RC circuit, employed to model the transient phenomena associated with electrochemical polarization.

The battery terminal voltage is expressed by Equation (4):

(4) $V_{b a t} = V_{o c} (S O C) - R_{i n t} \cdot i_{b a t} - V_{R C}$

where

V_OC(SOC) represents the open-circuit voltage, which is dependent on the State of Charge;

R_int is the equivalent internal series resistance of the battery;

i_bat is the charge/discharge current;

V_RC is the transient voltage component associated with the RC equivalent circuit.

The temporal evolution of the State of Charge (SOC) is described by Equation (5):

(5) $S O C (t) = S O C (t_{0}) - \frac{1}{C_{n o m}} \int_{t_{0}}^{t} i_{b a t} (τ) d τ$

where

C_nom is the nominal capacity of the battery;

i(t) represents the charge/discharge current, with a positive convention associated with discharging.

To prevent accelerated battery degradation and to ensure an extended lifespan of the storage system, the State of Charge is constrained within the following range:

(6) $S O C_{m i n} \leq S O C (t) \leq S O C_{m a x}$

with typical values:

SOC_min = 20%,

SOC_max = 90%.

In the simulations, the battery charge and discharge processes are modeled with a regime-dependent global efficiency, typically ranging between 92% and 96%. This accounts for internal cell losses as well as the conversion losses of the bidirectional inverter. The dynamic battery model, defined by the SOC evolution and by the voltage dependence on current and internal parameters, facilitates the faithful simulation of the BESS’s capability to respond rapidly to power disturbances—an aspect that is essential for the stability and resilience analysis of the PV–BESS microgrid.

To ensure the reproducibility of these dynamic responses, the parameters for the Thevenin equivalent circuit were calibrated as follows: the internal resistance was set to Rint = 0.025 Ω, while the RC network parameters were defined with a polarization resistance of R₁ = 0.01 Ω and a polarization capacitance of C₁ = 500 F. These specific values reflect the high-performance characteristics of the Lithium-Ion units used to achieve the stabilization times presented in the results.

Through this model, the following aspects can be accurately evaluated:

The actual supply and absorption limits of the storage system;

The transient effects on the voltage at the connection node;

The impact of the State of Charge on the BESS’s ability to support the microgrid during pre-blackout or operational stress regimes.

2.3.3. BESS Control Strategy Within the Microgrid

The control of the BESS is implemented based on a hierarchical structure, which enables its simultaneous participation in the following functions [15]:

Active power imbalance regulation;

Local voltage support via reactive power injection or absorption;

Frequency stabilization at the microgrid level.

Under normal operating conditions, the BESS operates in a power-balancing mode, according to Equation (7):

(7) $P_{B E S S} = P_{l o a d} - P_{P V}$

where

P_BESS represents the active power reference of the storage system [kW], taking a positive value for the discharge regime (injection into the grid) and a negative value for the charging regime (absorption from the grid);

P_load is the total active power demanded by the local loads connected within the microgrid [kW];

P_PV constitutes the active power instantaneously generated by the photovoltaic unit [kW].

Consequently:

A surplus of photovoltaic production (P_PV > P_load) results in a negative power reference (P_BESS < 0), with the energy being stored via battery charging;

A generation deficit (P_PV < P_load) results in a positive power reference (P_BESS > 0), which is compensated through the controlled discharge of the battery.

This strategy ensures the dynamic balancing of power flows within the microgrid and limits the stress on the distribution network in the presence of rapid solar resource variations.

In addition to this basic energy regime, the BESS is also equipped with dynamic frequency and voltage support mechanisms, which become active during deviations from nominal operating conditions. These are implemented through the P–f and Q–V control functions, described hereinafter.

Frequency Support via P–f Control

P–f control enables the BESS to stabilize the local frequency through the injection or absorption of active power, contingent upon the frequency deviation from its nominal value. Within an IoT architecture, this function is implemented based on real-time local measurements, ensuring a rapid response to the dynamic disturbances of the microgrid.

The control law is expressed by Equation (8):

(8) $P_{B E S S} = K_{f} \cdot (f_{n o m} - f)$

where

f_nom is the nominal grid frequency;

f is the locally measured frequency;

K_f represents the control constant (droop gain) of the P–f characteristic.

Through this strategy:

Upon a frequency drop, the BESS injects active power;

Upon a frequency rise, the BESS absorbs active power, transitioning into charging mode.

Consequently, the storage system contributes directly to reducing the Rate of Change of Frequency (RoCoF) and maintaining frequency stability during transient regimes.

Voltage Support via Q–V Control

Q–V control enables the BESS to regulate the local voltage through the controlled injection or absorption of reactive power, based on real-time voltage measurements. This function is essential within a distributed IoT architecture, where regulation decisions rely on rapid local data to maintain voltage stability.

The control law is expressed by the relationship (9):

(9) $Q_{B E S S} = K_{v} \cdot (V_{r e f} - V)$

where

V_ref is the voltage reference value;

V is the voltage measured at the connection node;

K_v is the sensitivity coefficient (droop gain) of the Q–V characteristic.

The voltage control of the BESS is implemented at the Alternating Current (AC) level, utilizing the nodal voltage expressed in per-unit (p.u.), with a reference value of V_ref = 1.0 p.u, corresponding to the nominal voltage of the low-voltage network. The selection of the 1.0 p.u. value corresponds to the nominal grid voltage and represents the operational equilibrium point around which the BESS Volt–Var control is executed.

Through this strategy:

Under overvoltage conditions, the BESS absorbs reactive power;

Under undervoltage conditions, the BESS injects reactive power, thereby contributing to the restoration of the voltage profile.

The implemented BESS control strategy integrates a dedicated current-limiting (saturation) block within the inner current control loop. This functional component is designed to emulate the physical protection of power electronic converters, ensuring that during severe disturbances—such as the 0.65 p.u. voltage sag in Scenario 4—the current injection is capped at a safe threshold (e.g., 1.2 p.u. of rated current). This saturation logic prevents the violation of the IGBTs’ Safe Operating Area (SOA) and ensures that the system’s contribution to the fault current remains within limits that do not interfere with the coordination of upstream protection devices. The BESS inverter is modeled as a balanced three-phase source. Although the network integrates single-phase loads, the control logic operates in the synchronous dq reference frame, focusing on the aggregate voltage stability of the microgrid node.

BESS Behavior during Transient Regimes

In severe transient regimes (rapid irradiance drops, sudden load steps, micro-faults, or high RoCoF variations), the BESS enters a rapid response mode, supplying or absorbing power within a timeframe of the order of milliseconds, with the aim of:

Limiting voltage deviations at the nodal level;

Damping sudden frequency variations;

Preventing the unintended disconnection of distributed sources;

Maintaining supply continuity for sensitive loads.

Through the coordinated combination of power-balancing, P–f control, and Q–V control, the BESS acts as an active stabilization element and a central pillar of the PV–BESS microgrid’s resilience, being perfectly integrable into a distributed IoT monitoring and control architecture.

2.3.4. The Role of BESS in Enhancing PV–BESS Microgrid Resilience

From the perspective of energy resilience, the Battery Energy Storage System (BESS) plays a pivotal role in maintaining the secure and stable operation of the microgrid, particularly in the presence of high variability from the photovoltaic source and disturbances within the low-voltage network. Through its capacity to rapidly supply or absorb active and reactive power, the BESS contributes directly to [5]:

Mitigating the amplitude of voltage sags and local overvoltages;

Damping power oscillations induced by the rapid variability of photovoltaic production;

Maintaining the supply to critical loads during pre-blackout regimes;

Facilitating controlled transitions between the power-controlled and power-hold regimes of the PV unit.

In the absence of a storage system, sudden variations in solar irradiance are reflected almost instantaneously in the active power injected by the PV unit, potentially leading to significant voltage profile disturbances, especially in the downstream area of the network (Zone C), where line impedances are high. Integrating the BESS at Node 16 allows for the partial decoupling of this direct dependence between photovoltaic production and the grid’s electrical behavior, offering an efficient “energy buffer” mechanism.

Through the power-balancing function, the BESS absorbs energy surplus during periods of high irradiance and rapidly compensates for generation deficits during sudden PV production drops. This mechanism reduces dynamic stress on the transformer and Low Voltage (LV) lines, while simultaneously limiting the occurrence of local congestion and accelerated voltage variations [16].

In disturbed regimes, the P–f and Q–V control mechanisms enable the BESS to actively participate in frequency and voltage stabilization. By rapidly injecting active power during frequency dips and absorbing reactive power under overvoltage conditions, the BESS contributes to reducing critical instability indicators, such as RoCoF and VDI. This capability is particularly important in pre-blackout scenarios, where the synchronization of distributed sources and the maintenance of voltage limits are essential conditions for avoiding cascading disconnections.

Another major role of the BESS in enhancing microgrid resilience is supporting energy continuity for sensitive loads. In the event of severe external grid disturbances or local micro-faults, the BESS can temporarily take over the supply of priority consumers, preventing sudden interruptions and significantly reducing Energy Not Supplied (ENS).

From the perspective of the IoT architecture proposed in this work, the BESS is not merely a passive storage element but an intelligent active actor, integrated into the sensor network and distributed analysis mechanisms. Parameters such as the active and reactive power delivered by the BESS, local voltage, current, and State of Charge (SOC) are monitored in real-time and correlated with stability and resilience indicators. This facilitates the anticipation of critical situations and the triggering of preventive control strategies before instability thresholds are reached.

Through the coordinated interaction between PV, BESS, and the IoT monitoring and decision layer, the microgrid acquires an adaptive behavior, capable of responding intelligently to disturbances, maintaining the supply to critical loads, and significantly reducing the probability of a local blackout. Thus, the BESS becomes a central element not only of energy balance but also of the functional resilience of the PV–BESS microgrid [17].

2.3.5. Relevance of BESS for IoT-Based Monitoring

In the proposed IoT architecture, the BESS functions not merely as an energy storage element but as an intelligent node responsible for the acquisition, processing, and provision of operational data critical to microgrid stability and resilience. Through its integrated sensors and control system, the BESS provides the following categories of information in real-time:

Active power absorbed from or injected into the grid;

Reactive power supplied for voltage support;

Phase voltages and currents;

State of Charge (SOC) of the battery;

The dynamics of charge/discharge/standby operating modes.

These data are transmitted via specific IoT communication protocols (such as MQTT/Modbus) to the edge monitoring layer, where they undergo preprocessing to calculate local stability and resilience indicators, including the Voltage Deviation Index (VDI), frequency variation, RoCoF, and the estimation of Energy Not Supplied (ENS).

Consequently, the integration of BESS into the IoT architecture facilitates:

Early detection of active and reactive power imbalances;

Identification of voltage stability degradation in Zone C;

Anticipation of pre-blackout states through the correlation of SOC–P–f–V parameters;

Automatic triggering of preventive control strategies (increasing power injection, shedding non-critical loads, adaptive Volt–Var support).

In this manner, the BESS becomes an active actor within the monitor–decide–act loop specific to intelligent IoT systems, contributing not only to energy balancing but also to enhancing the autonomy and adaptivity of the PV–BESS microgrid.

The data provided by the BESS are subsequently utilized in Section 4 for:

The quantitative evaluation of stability and resilience indices;

The analysis of microgrid behavior during transient regimes;

The validation of the IoT architecture’s performance in preventing local blackouts.

2.4. Real Load and Generation Profiles (Load and PV Production Profiles)

To ensure a high degree of realism in the simulations and to enable an accurate assessment of the performance of the PV–BESS microgrid and the proposed IoT architecture, this study utilizes realistic temporal profiles for load and photovoltaic generation, representative of European urban low-voltage networks.

Residential load profiles are characterized by a typical daily variation, featuring low consumption levels during the nocturnal interval (00:00–06:00), a progressive rise in the morning, a moderate regime during the day, and a pronounced evening peak between 17:00 and 22:00. This distribution pattern is specific to residential consumers and facilitates the analysis of microgrid behavior under conditions of maximum energy stress. Loads are unevenly distributed across phases, in accordance with the CIGRE LV specification, to reproduce the natural imbalances inherent to low-voltage networks. ZIP modeling (Constant Impedance–Constant Current–Constant Power) ensures an adequate representation of the load’s dependence on voltage levels.

Photovoltaic generation profiles are derived based on the temporal evolution of solar irradiance and PV cell temperature. Irradiance follows a daily “bell-shaped” profile, with peak values occurring around noon, reflecting a typical clear-sky day. Cell temperature is correlated with irradiance and directly influences the module terminal voltage and the position of the Maximum Power Point (MPP). By incorporating these dependencies, the employed PV model captures both the natural variations in production and the efficiency losses associated with module overheating.

An essential element of the analysis is the temporal correlation between load, photovoltaic production, and the operation of the BESS. During intervals of high photovoltaic production, energy surplus is temporarily stored in the batteries, while during the evening peak, the accumulated energy is utilized to compensate for the deficit between demand and generation. This strategy leads to a reduction in energy imports from the main grid, the limitation of voltage drops during peak intervals, and the damping of rapid variations in the power injected by the PV source.

The use of these realistic profiles allows for a faithful reproduction of the alternation between energy surplus and deficit regimes, pre-instability situations induced by the overlap of load peaks with declining photovoltaic production, as well as the rapid transients associated with solar resource variability. These conditions constitute the testing basis for the simulation scenarios analyzed subsequently and for the evaluation of the microgrid’s stability and resilience indicators.

Consequently, the Load–PV profiles utilized in this work represent not merely input signals, but a fundamental element for validating the dynamic behavior of the microgrid and the performance of the proposed IoT architecture; the corresponding results are analyzed in detail in Section 4.

3. Proposed IoT Architecture

The PV–BESS microgrid model presented in Section 2 details the physical structure of the system: the CIGRE LV network, the photovoltaic unit, the battery storage system, and the realistic load and generation profiles. However, to effectively evaluate and exploit the potential of this model within the context of smart grids, it is necessary to overlay a distributed IoT (Internet of Things) monitoring and analysis layer, capable of collecting, processing, and transmitting relevant information regarding the microgrid state in real-time [18].

This section defines the proposed IoT architecture for the analyzed PV–BESS microgrid, with the primary objective of transforming the physical elements—PV, BESS, load nodes, and measurement points—into intelligent, connected entities that actively contribute to stability and resilience monitoring.

The proposed IoT architecture is structured into several functional layers, as follows [19]:

The Data Acquisition Layer, represented by IoT sensors integrated at key microgrid nodes (PV, BESS, load nodes, terminal points), responsible for measuring voltage, frequency, active and reactive powers, and dynamic indicators;

The Communication Layer, based on lightweight and robust protocols such as MQTT and Modbus, which ensures efficient real-time data transmission to processing nodes;

The Intelligent Edge Processing Layer, implemented using platforms such as ESP32 and Node-RED, where data filtering, aggregation, and preliminary analysis take place;

The Upper Aggregation and Analysis Layer, implemented at the cloud or local server level, responsible for long-term storage, historical analysis, advanced visualization, and integration with energy management systems.

On this infrastructure, intelligent analysis algorithms (edge analytics) are implemented, enabling microgrid state estimation and the real-time generation of relevant stability indicators (VDI, RoCoF, and RI), which are subsequently utilized in the simulation scenarios presented in Section 4.

3.1. Integrated IoT Sensors (Voltage, Frequency, P, Q, RoCoF)

To ensure real-time monitoring of the PV–BESS microgrid state and to facilitate the dynamic assessment of stability and resilience, the proposed IoT architecture relies on a distributed network of smart sensors, integrated at key nodes of the CIGRE LV network. These sensors constitute the primary functional layer of the IoT architecture and are responsible for the continuous acquisition of fundamental electrical quantities.

Monitored Electrical Quantities

The main set of quantities acquired in real-time includes:

Nodal voltage V_node(t), monitored for the detection of overvoltages, undervoltages, and phase imbalances;

Grid frequency f(t), utilized in dynamic stability analysis and in the P–f support strategies of the BESS;

Active power P(t), measured at the level of the PV unit, the BESS, and main loads, for the characterization of energy flows;

Reactive power Q(t) essential for the analysis of Q–V control and voltage support;

Rate of Change of Frequency (RoCoF), a critical indicator for the rapid detection of power imbalances and pre-instability situations [20,21].

These quantities enable a comprehensive description of the local electrical state at each monitored point of the microgrid.

Sensor Placement within the Microgrid

IoT sensors are strategically positioned to capture both the behavior of distributed sources and the response of the grid and consumers:

At Node 15 (Zone C)—for direct monitoring of the photovoltaic unit (voltage, current, PPV, QPV);

At Node 16 (BESS)—for measuring absorbed or injected power, local voltage, and the dynamics of charge/discharge regimes;

On the load branches in Zones A and B—for tracking consumption profiles, phase imbalances, and the impact of loads on global stability;

At the MV/LV transformer level, for the voltage and frequency reference of the entire microgrid.

This arrangement allows for high system observability without requiring the instrumentation of every network node, maintaining an optimal trade-off between accuracy and cost.

Temporal Resolution and Sampling Frequency

To capture the rapid dynamics of the analyzed phenomena (voltage sags, frequency variations, PV–BESS transients), the sensors operate with:

Sampling frequencies in the 1–10 kHz range for voltage and current;

Update rates for derived indicators (P, Q, f, RoCoF) every 10–50 ms.

This resolution is sufficient for:

Early detection of disturbances;

Real-time feeding of edge-level intelligent processing algorithms;

Rapid alert generation within the IoT system.

The Role of Sensors in the Context of Stability and Resilience

IoT sensors do not merely serve a passive measuring role but represent primary information sources for distributed intelligence mechanisms. Based on the acquired data:

Local stability indicators (VDI, RoCoF) are evaluated in real-time;

Energy imbalance situations between generation and consumption are detected;

Pre-instability conditions that may lead to cascading trips are identified;

Control strategies for the BESS and the PV unit are supplied with data.

Through this integration, the microgrid is transformed from a passive system into an active cyber–physical system, capable of assessing its state and reacting adaptively to disturbances.

3.2. Node-RED/MQTT/Modbus/ESP32—Communication and Edge Processing Infrastructure

The transmission of data acquired by IoT sensors and their real-time processing are realized through a hybrid communication infrastructure, which combines established industrial protocols with modern IoT technologies and edge computing platforms. In the proposed architecture, ESP32 nodes serve as field devices, being directly connected to sensors and power equipment within the microgrid. These nodes ensure measurement acquisition, signal filtering, the local computation of derived quantities (P, Q, f, RoCoF), and data transmission to the upper processing layer.

Interfacing with PV inverters, BESS, and energy meters is achieved via the Modbus (RTU/TCP) protocol, which facilitates synchronous access to measurement registers and guarantees compatibility with existing industrial equipment. The collected data are subsequently transmitted via the MQTT protocol, selected for efficient information transport due to its low latency, minimal bandwidth consumption, and publish–subscribe mechanism. This structure allows for the flexible distribution of data streams to multiple monitoring and control applications.

To provide an overview of the physical and logical interconnection between equipment, Figure 8 illustrates the implemented hardware topology. The diagram highlights the role of the IoT node (ESP32) as a critical conversion interface between wired industrial protocols (Modbus RTU), utilized at the physical layer, and wireless data streams (MQTT) feeding the Edge Computing layer, thereby ensuring a clear separation between the energy domain and the digital domain.

The Node-RED platform is utilized as the primary processing node at the edge level, serving the role of aggregating, validating, and preprocessing data originating from the microgrid. At this layer, mechanisms for supplementary filtering, alert generation, and the preliminary calculation of local stability indicators are implemented. Executing these functions at the network edge significantly reduces decision latency and the volume of data requiring transmission to the upper analysis layer.

The information flow follows the sequence: local measurement → edge processing on ESP32 → transmission via MQTT protocol → processing and orchestration in Node-RED → storage and analysis at the cloud or local server level. This architecture efficiently distributes the computational load between the field level and the central level and establishes the necessary support for implementing deterministic intelligent edge processing mechanisms, detailed in Section 3.3.

In the context of this study, ‘Intelligence’ refers to the Edge Intelligence (EI) paradigm, where the node possesses the autonomous capability to process complex local data and execute stabilization strategies without cloud intervention. This deterministic approach is prioritized to ensure the sub-millisecond response times required for power grid resilience, where the high latency or unpredictability of complex AI models could be detrimental.

3.3. Intelligent Edge Processing for Microgrid Stability Estimation

The IoT architecture proposed in this work extends the classical monitoring paradigm by integrating a distributed Intelligent Edge Processing (IEP) layer at the network edge. This functional layer is essential for reducing decision latency and for the early detection of pre-instability situations, endowing the PV–BESS microgrid with capabilities for self-monitoring and rapid response, independent of the latencies inherent to communication with a central server. The operational flow and the processing logic implemented at the edge nodes are detailed in Figure 9, which illustrates the transformation of raw local measurements into active control decisions [22].

Although modern hardware platforms like the ESP32 facilitate the implementation of machine learning models (TinyML), this study proposes and validates a deterministic Rule-Based IEP approach, a design decision grounded in the strict safety requirements of critical energy infrastructures.

Explainability (XAI): Grid operators must be able to unequivocally identify the cause of an automated control action—a requirement difficult to achieve with “black-box” neural models.

Predictability (Hard Real-Time): Deterministic algorithms guarantee a predictable response time, eliminating the latency variations in statistical inference that could compromise stability during rapid transient regimes.

Robustness: A logic based on physical network limits and the EN 50160 standard ensures robustness against rare events (“Black Swans”), scenarios that models trained exclusively on nominal historical data risk classifying incorrectly [23].

The operationalization of this logic follows the functional structure presented in Figure 9.

The process commences with Feature Extraction, where instantaneous measurements of voltage V(t), frequency f(t), and power are processed to derive synthetic indicators such as the Voltage Deviation Index (VDI) and the Rate of Change of Frequency (RoCoF). These indicators feed into the Stability Assessment module, which compares dynamic values against adaptive thresholds to classify the microgrid state into one of three discrete regimes: Normal Operation, Alert Condition, or Pre-instability State.

The detection of a critical state instantaneously activates the Local Decision Logic, which triggers autonomous corrective measures without awaiting confirmation from the cloud level. These actions include activating frequency support (P-f) or voltage support (Q-V) via the BESS or preventively limiting photovoltaic power injection. In parallel, aggregated data and generated alerts are transmitted asynchronously to the Cloud Platform for storage and historical analysis, thereby achieving an efficient decoupling between the fast control loop (Edge) and the strategic supervision level (Cloud).

Processing is executed directly on the edge nodes (ESP32) and at the local gateway level (Node-RED), eliminating exclusive dependence on a central server and enabling rapid local decision-making, even under conditions of limited or unstable connectivity. Through this approach, the microgrid is transformed into a cyber–physical system capable of continuous self-monitoring and self-evaluation of its operational state.

The general structure of the proposed IoT–Edge architecture, as well as the data and decision flows between the physical, edge, and cloud levels, are illustrated in Figure 8.

Figure 9 illustrates the proposed IoT–Edge architecture for stability monitoring and control coordination within a PV–BESS microgrid. Local measurements are acquired at the CIGRE LV microgrid level and processed in real-time at the edge level (ESP32), where stability indicators (VDI, RoCoF, ΔP) are extracted and operational states (normal, alert, pre-instability) are identified. Based on the local evaluation, control actions are generated for the BESS and PV, while aggregated data are transmitted to the cloud platform for offline analysis and visualization.

The stability of a low-voltage microgrid is a dynamic phenomenon, influenced simultaneously by the variability of photovoltaic production, residential load fluctuations, BESS interventions, and the occurrence of local disturbances such as voltage sags, overloads, or micro-faults. In this context, the intelligent edge processing layer serves the role of correlating real-time local measurements of voltage, frequency, active power, reactive power, and RoCoF to assess the instantaneous state of the microgrid and identify conditions associated with stability degradation.

Although the hardware platforms employed technically support the implementation of machine learning models, this work does not utilize a trained machine learning model, but rather a deterministic intelligent processing logic. This represents a design decision grounded in the strict safety requirements of critical energy infrastructures.

Firstly, the necessity for Explainability (XAI) mandates that grid operators be able to unequivocally identify the cause of an automated control action—a requirement that is difficult to fulfill with “black-box” neural models.

Secondly, deterministic algorithms guarantee a predictable response time (Hard Real-Time), eliminating the latency variations associated with statistical inference that could compromise stability during rapid transient regimes.

Finally, a logic based on physical network limits and the EN 50160 standard ensures robustness against rare events (“Black Swans”), scenarios that models trained exclusively on nominal historical data risk classifying incorrectly.

Thus, the selection of this architecture ensures transparency and reproducibility, while being perfectly adapted to the limited resources of embedded IoT platforms.

The main set of variables used as inputs for the intelligent processing module includes:

Nodal voltage V_node(t);

Local frequency f(t);

Active power P(t);

Reactive power Q(t);

Rate of Change of Frequency (RoCoF);

Battery State of Charge SOC(t).

Based on these quantities, the following are calculated and monitored in real-time:

Local voltage stability indicators (Voltage Deviation Index—VDI);

Local disturbance severity indicators;

Discrete operational states (normal regime, alert regime, critical regime);

The necessity of activating support from the BESS or limiting the power injected by the PV source.

Concretely, the implementation of this strategy is realized through explicit if–else rules, which correlate instantaneous deviations of voltage, frequency, and RoCoF with predetermined safety thresholds. This logical structure allows for the rapid identification of pre-instability situations and the immediate triggering of control measures, translating theoretical robustness principles into an efficient execution algorithm.

To operationalize this deterministic approach, the Stability Assessment module (shown in Figure 9) utilizes specific adaptive thresholds calibrated against the EN 50160 standard. These thresholds categorize the microgrid state as follows:

Normal Operation: Defined by a Voltage Deviation Index |VDI| ≤ 0.05 and a |RoCoF| ≤ 0.1 Hz/s. In this state, as seen in Scenario 1, the BESS maintains a baseline power exchange to balance minor residential load fluctuations.

Alert Condition: Triggered when 0.05 < |VDI| ≤ 0.10 or frequency deviates beyond nominal limits. This state is exemplified in Scenario 2 (irradiance drops) and Scenario 3 (local overload), where the voltage drops to approximately 0.94–0.95 p.u., initiating moderate reactive power support (Q-V control) from the BESS.

Pre-instability State: Activated by severe disturbances where |VDI| > 0.10 or |RoCoF| > 0.2 Hz/s. This condition is clearly evidenced in Scenario 4, where a grid fault causes an instantaneous voltage drop to 0.65 p.u. (VDI = 0.35). In this regime, the logic triggers maximum active power injection (P-f support) to prevent a local blackout.

The decision logic also monitors the State of Charge (SOC); control actions are only permitted if SOC > 20%, ensuring the BESS is not depleted during emergency support. By using these physical network limits, the architecture achieves a response time under 300 ms, significantly outperforming traditional polling-based SCADA systems.

The systematic execution of this rule-based approach is illustrated in the flowchart in Figure 10. The logic follows a continuous loop, starting with the acquisition of nodal measurements and the calculation of the VDI and RoCoF stability indicators. By correlating these physical parameters with the battery’s SOC, the algorithm ensures a graduated response, transitioning from nominal balancing to active voltage or frequency support only when predefined safety margins are exceeded. This deterministic flow guarantees that every control action is explainable and executed within the strict real-time constraints required for microgrid resilience.

The intelligent processing module is implemented either at the Node-RED gateway level or directly on ESP32 nodes with sufficient processing capacity. The results of the local evaluation are used for the rapid triggering of control actions, such as activating active and reactive power support from the BESS, and are transmitted to the cloud level in the form of synthetic indicators, while also being stored for the historical analysis of microgrid stability and resilience [24].

Through this organization, a clear separation is achieved between:

The local rapid detection and reaction level (edge);

The advanced analysis, aggregation, and reporting level (cloud).

Although the proposed architecture is compatible with the future integration of advanced machine learning techniques, the present study focuses exclusively on deterministic intelligent processing at the edge level, aiming to ensure a robust, explainable implementation suitable for real-world applications in low-voltage microgrids.

In conclusion, the synthetic indicators derived through this intelligent processing layer—particularly VDI, RoCoF, and stability margins—constitute the fundamental input data for system performance evaluation. These are utilized extensively in Section 4, where they are formally defined and analyzed within the simulation scenarios. Thus, the Intelligent Edge Processing architecture moves beyond the conceptual stage, being integrated directly as a critical operational component in the validation methodology of the PV–BESS microgrid stability and resilience.

3.4. Complete Data Flow: “Measurement → Edge Processing → Cloud Logging”

The efficient operation of the proposed IoT architecture relies on a coherent and hierarchical data flow that transforms raw electrical measurements into relevant indicators for assessing stability, resilience, and preventing local blackouts. This flow is structured into three main levels: measurement, edge processing, and cloud logging, each having a well-defined role in the monitoring and decision chain.

At the lower level of the architecture (The Acquisition Level—Measurement), distributed IoT sensors are placed at key nodes of the PV–BESS microgrid, with an emphasis on Zone C, in the vicinity of the photovoltaic unit connected at Node 15 and the BESS at Node 16. These sensors perform real-time acquisition of the main electrical quantities, including voltage, current, frequency, active and reactive power, as well as the Rate of Change of Frequency (RoCoF).

Data are sampled at sufficiently high frequencies to capture the rapid dynamics of transient phenomena and are transmitted to the edge level via lightweight and robust communication protocols, suitable for embedded platforms.

At the intermediate level of the architecture (The Local Processing Level—Edge Processing), raw data undergo filtering, normalization, and preprocessing stages, after which they are analyzed through the deterministic Intelligent Edge Processing mechanisms described in Section 3.3. In this stage, local stability and resilience indicators—such as voltage deviations, frequency variations, RoCoF, local disturbance severity, and the microgrid operational state (normal, alert, or critical)—are calculated in real-time [25]. Based on these indicators, local decisions are generated regarding the necessity of activating support mechanisms from the BESS or limiting the power injected by the photovoltaic source.

By performing processing directly at the edge level:

Decision latency is significantly reduced;

The volume of data transmitted to the upper level is diminished;

System operation continuity is ensured even under partially degraded connectivity conditions.

Decisions resulting from this stage can directly trigger control actions, such as activating frequency (P–f) or voltage (Q-V) support from the BESS.

At the upper level of the architecture (The Aggregation and Storage Level—Cloud Logging), only synthetic indicators, alerts, and significant events are transmitted to a local server or cloud infrastructure. This level is responsible for:

Long-term storage of operational data;

Historical analysis of the microgrid’s stability and resilience evolution;

Temporal correlation of events;

Report generation and support for strategic and operational decisions.

The clear separation between local processing and centralized analysis allows for the scaling of the IoT architecture without significant increases in data traffic and without overloading the communication infrastructure.

Through this “measurement → edge processing → cloud logging” chain, the proposed architecture combines the advantages of distributed monitoring with those of centralized analysis. Early detection of pre-instability regimes at the edge level, correlated with historical analysis at the cloud level, enables both rapid reaction to disturbances and the adaptation of control and local blackout prevention strategies [26]. In this configuration, the PV–BESS microgrid acquires an adaptive and robust behavior, reinforcing the primary objective of this work: enhancing energy resilience through intelligent IoT monitoring and distributed processing at the network edge.

A critical advantage of this hybrid Edge–Cloud architecture is its communication fault tolerance. Since the critical decision logic for stability (calculation of VDI, RoCoF indicators, and BESS activation) is executed locally at the Edge node level (ESP32 and Gateway), the system maintains its operational functionality even in the scenario of a total loss of internet connectivity. In this digitally islanded operating regime, data are temporarily stored in local buffers, and control actions for blackout prevention continue to be executed autonomously. The Cloud level intervenes exclusively for long-term storage, historical analysis, and complex event correlation, not being a critical link in the real-time control loop. This decoupling ensures superior robustness compared to centralized SCADA systems, which are dependent on the continuous availability of communication channels.

4. Methodology and Stability Indicators

The evaluation of the stability and resilience of the PV–BESS microgrid analyzed in this work is based on a coherent set of electrical and energetic indicators, calculated in real-time via the proposed IoT architecture presented in Section 3. This architecture integrates distributed sensors, hybrid communication protocols (MQTT/Modbus), intelligent edge processing nodes, and an upper layer for aggregation and analysis. This configuration facilitates not only the parametric monitoring of the network but also the dynamic interpretation of its behavior under both normal and disturbed operating conditions.

The adopted methodology aims to correlate local electrical measurements—voltage, frequency, active and reactive power, RoCoF, and the Battery Energy Storage System State of Charge (SOC)—with synthetic stability and resilience indicators capable of describing both the severity of disturbances and the microgrid’s capacity to return to a safe operating regime. To this end, indicators established in the scientific literature are employed, adapted for low-voltage applications and for implementation within distributed IoT systems with constrained resources.

The indicators analyzed in this section address:

Nodal voltage stability;

Dynamic frequency behavior and associated rapid variations;

The microgrid’s capacity to maintain load supply under operational stress conditions;

The coordinated response of the photovoltaic source and the Battery Energy Storage System (BESS).

Based on these indicators, several test scenarios representative of modern microgrid operation are subsequently defined and analyzed, including rapid variations in photovoltaic production, local overloads, and the occurrence of short-duration transient faults. These scenarios allow for a comparative evaluation of the microgrid’s performance under nominal and disturbed conditions, while also highlighting the benefits derived from distributed monitoring and intelligent edge-level processing.

In the following subsections, the employed indicators are formally defined, the corresponding mathematical relationships are presented, and the simulation scenarios underpinning the numerical results analysis and discussions in Section 5 are introduced.

4.1. Voltage Deviation Index (VDI)

Voltage stability represents one of the most critical indicators of safe operation in low-voltage microgrids, particularly in the presence of photovoltaic sources and battery energy storage systems. Rapid variations in active power injected by PV units, as well as local load dynamics, can lead to overvoltages or undervoltages that affect both power quality and the safety of connected equipment [27].

To quantify the level of voltage stability at the nodal level, this work utilizes the Voltage Deviation Index (VDI), defined by Equation (10):

(10) $V D I (t) = \frac{∣ V (t) - V_{n o m} ∣}{V_{n o m}}$

where

V(t) is the instantaneously measured voltage at the analyzed node;

V_nom is the nominal voltage of the low-voltage network (400 V line-to-line, or 230 V phase-to-neutral) [28].

The VDI expresses the relative deviation of the voltage from its nominal value and allows for a direct assessment of the severity of voltage disturbances. In compliance with the requirements of the EN 50160 standard, normal operation of the low-voltage network requires maintaining the voltage within the following range (11):

(11) $0.9 \cdot V_{n o m} \leq V (t) \leq 1.1 \cdot V_{n o m}$

Values of the VDI exceeding this threshold indicate the occurrence of abnormal operating conditions (overvoltages, undervoltages), with a potential negative impact on consumers and distributed sources.

Within the proposed IoT architecture, the VDI is calculated in real-time based on voltage values acquired by distributed sensors located at critical nodes of the microgrid (specifically at Node 15—the connection point of the PV unit—and Node 16—the connection point of the BESS) [29,30].

The voltage values are:

Measured locally by sensors;

Transmitted via the MQTT protocol to edge nodes;

Processed in real-time for VDI calculation;

Utilized by intelligent edge processing modules to estimate the local stability level.

Consequently, the VDI becomes not merely a passive monitoring indicator but an active decision-making parameter, utilized for:

Triggering voltage support via the BESS Q–V control;

Dynamically limiting the power injected by the PV unit;

Generating IoT alerts in the event of critical threshold violations.

In this work, the VDI is employed as the primary indicator for evaluating voltage stability across the following scenarios:

Rapid variations in photovoltaic production;

The occurrence of local overloads;

Transient fault events in the low-voltage network.

For each analyzed scenario, the following are evaluated:

The maximum value of the VDI;

The duration of the permissible threshold violation;

The voltage recovery speed to the nominal range.

These parameters allow for the assessment of disturbance severity and the capacity of the IoT-monitored PV–BESS microgrid to maintain voltage stability.

4.2. Resilience Index (RI)

Electric microgrid resilience represents the capacity to anticipate, absorb, withstand, and rapidly recover from disturbances, while simultaneously maintaining an acceptable level of supply to consumers. In the context of low-voltage PV–BESS microgrids, resilience is strongly influenced by:

The variability of the photovoltaic source;

The dynamics of residential loads;

The rapid response capability of the BESS;

The efficiency of IoT-based monitoring and control mechanisms.

To quantify this dynamic behavior, this work introduces the Resilience Index (RI), which expresses the microgrid’s capacity to return to a safe operating regime following the occurrence of a disturbance [31].

The Resilience Index is defined based on the voltage and/or frequency recovery time relative to the disturbance duration, according to Equation (12):

(12) $R I = \frac{T_{r e c}}{T_{d i s t}}$

where

T_dist is the duration of the disturbance (the time interval during which the monitored parameter remains outside admissible limits);

T_rec is the system recovery time to nominal limits following the clearance of the disturbance.

For a unified interpretation, this work also utilizes the normalized form of the Resilience Index (13):

(13) $R I_{n o r m} = 1 - \frac{T_{r e c}}{T_{r e c, m a x}}$

where T_rec,max represents the maximum admissible recovery time, established based on the operational requirements of the microgrid [32].

Thus:

RI_norm ≈ 1 indicates very good resilience (rapid recovery);

RI_norm ≈ 0 indicates low resilience (slow recovery);

Intermediate values reflect partial levels of resilience.

Regarding the calculation of the Resilience Index (RI) defined in Equation (13), it is important to specify that for all investigated cases, and particularly for Scenario 2 characterized by continuous irradiance variations, the term T_dist in the denominator represents a predefined observation window of 10 s post-disturbance. This fixed timeframe was selected to fully capture the transient response and the subsequent stabilization phase, ensuring a consistent and normalized basis for comparing the microgrid’s recovery performance across the different disturbance types.

From a physical perspective, the Resilience Index expresses the speed at which the microgrid manages to restore energy balance and voltage stability following a disturbance. A system with high resilience is characterized by [33]:

Rapid voltage recovery within the admissible range (EN 50160);

Frequency stabilization following transient variations;

Maintenance of supply to critical loads;

Avoidance of successive protection trips (cascading trips).

In the analyzed microgrid, the value of the resilience index is strongly correlated with:

The response capacity of the BESS via P–f and Q–V control;

The power-hold operating regime of the PV unit;

The speed of decision-making at the IoT architecture level.

Within the proposed IoT architecture, the resilience index is calculated automatically at the edge processing node level, based on real-time measured data describing the electrical and energetic state of the microgrid, namely:

Nodal voltage V(t);

Local frequency f(t);

Active and reactive powers;

The State of Charge (SOC) of the BESS.

Based on these quantities, the following are identified:

The moment of disturbance occurrence;

The duration of the interval during which parameters exceed admissible limits;

The moment of system return to a stable regime.

Intelligent edge processing utilizes this information to estimate the microgrid’s recovery dynamics and to evaluate, in real-time, the level of local resilience. Depending on the value of the resilience index, the following actions can be triggered:

Supplementary support strategies from the BESS;

Temporary limitations of PV power;

IoT alerts to the upper monitoring level.

The resilience index is utilized in this work for a comparative evaluation of the microgrid within the four analyzed operating scenarios:

Scenario 1—Reference operation (baseline);

Scenario 2—Rapid variations in photovoltaic production;

Scenario 3—Local overloads;

Scenario 4—Transient fault events.

Within these scenarios, the analysis also includes the impact of the photovoltaic inverter switching between power-controlled and power-hold regimes on the microgrid’s stability and recovery capacity.

In the specific case of Scenario 2, the disturbance induced by rapid fluctuations in photovoltaic production does not possess a well-defined discrete duration, being characterized instead by a continuous energetic variation associated with changes in irradiance and temperature. In this situation, the disturbance duration T_dist cannot be strictly determined for the application of Equation (12).

Consequently, for the evaluation of resilience in the context of PV fluctuations, the normalized form defined previously in Equation (13) is utilized directly. This approach eliminates the dependence on the uncertain duration of the disturbance and assesses performance exclusively based on the recovery time (T_rec) required for the voltage and frequency to return to admissible ranges following a solar cloud transient event.

Thus, for all analyzed scenarios, the parameters monitored for quantifying resilience are:

The disturbance duration (where applicable);

The voltage and frequency recovery time (T_rec);

The value of the normalized resilience index (RI_norm), calculated both before and after BESS intervention.

These parameters allow for a unified assessment of the efficiency of the proposed IoT architecture in enhancing PV–BESS microgrid resilience, regardless of the differing nature of the disturbances (energetic or electrical).

4.3. RoCoF (Rate of Change of Frequency)

Rapid frequency variation represents one of the most sensitive indicators of the imbalance between active power generation and demand within an electrical system. In low-voltage microgrids characterized by a high penetration of photovoltaic sources and battery energy storage systems, frequency can exhibit more pronounced variations compared to conventional grids, particularly during transient regimes or islanded operation.

To quantify frequency dynamics and facilitate the early identification of pre-instability conditions, this study employs the RoCoF (Rate of Change of Frequency) indicator. This parameter expresses the temporal rate of frequency variation and constitutes an essential element in modern blackout prevention strategies [34,35].

RoCoF is defined as the time derivative of the local frequency, according to Equation (14):

(14) $R o C o F (t) = \frac{d f (t)}{d t}$

where

f(t) is the locally measured frequency;

df(t)/dt represents the rate of frequency variation.

In the numerical implementation, RoCoF is calculated discretely based on finite differences:

(15) $R o C o F (k) = \frac{f (k) - f (k - 1)}{Δ t}$

where

f(k) represents the frequency measured at the current time step;

f(k−1) represents the frequency measured at the previous time step;

Δt is the sampling interval of the IoT measurement system.

From a physical perspective, RoCoF reflects the instantaneous imbalance between active power generation and consumption [36]. Thus:

High positive RoCoF values indicate a sudden generation surplus;

High negative RoCoF values indicate a rapid active power deficit;

Small values, close to zero, correspond to a nearly balanced regime.

In low-inertia microgrids (such as those dominated by power electronic converters, PV, and BESS), RoCoF is significantly more sensitive than in conventional grids based on synchronous generators. For this reason, RoCoF is utilized as a critical indicator for the rapid detection of frequency instabilities and for triggering protection and control mechanisms.

In practical applications, RoCoF thresholds are used to detect dangerous situations, with typical values ranging between: ∣RoCoF∣ ≤ 0.5“–” 1“ Hz/s”

Exceeding these thresholds may lead to:

Triggering of frequency protections;

Disconnection of distributed sources;

Separation of the microgrid into islanded mode.

Therefore, real-time RoCoF monitoring is essential to prevent the occurrence of severe instability phenomena and to avoid cascading protection trips [37].

In the proposed IoT architecture, frequency is measured at critical microgrid nodes via digital sensors, while RoCoF is calculated in real-time at edge processing nodes. The values obtained are:

Filtered to eliminate measurement noise;

Utilized by intelligent edge processing modules to classify operating regimes;

Transmitted to the cloud level in the form of synthetic stability indicators.

Intelligent edge processing correlates RoCoF with quantities such as voltage, powers (P and Q), and the BESS State of Charge, enabling:

Early detection of energy imbalances;

Estimation of the probability of entering a pre-instability regime;

Preventive triggering of energy support strategies.

In the analyzed PV–BESS microgrid, RoCoF plays an essential role in the dynamic coordination of distributed sources, particularly for:

Rapid activation of the BESS P–f control;

Temporary limitation of the power injected by the PV unit;

Frequency stabilization during severe transient regimes.

In the event of a rapid frequency drop (high negative RoCoF), the BESS is commanded to inject active power, compensating for the energy deficit. Conversely, the BESS absorbs surplus power, preventing uncontrolled frequency rise [38].

Thus, RoCoF becomes a key indicator for local blackout prevention strategies, allowing for rapid interventions before classical frequency limits are exceeded.

In this work, RoCoF is utilized for the analysis of:

Sudden variations in photovoltaic production;

Rapid load switching;

The occurrence of transient faults;

Transitions between microgrid operating regimes.

For each scenario in Section 4.4, the following are evaluated:

The maximum RoCoF value;

The duration of critical threshold violations;

The efficiency of BESS intervention in damping frequency oscillations.

These indicators constitute the basis for interpreting the performance of the IoT architecture and PV–BESS coordination in Section 5.

4.4. Test Scenarios (Faults, PV Variation, Overloads)

To evaluate the performance of the PV–BESS microgrid and the efficiency of the proposed IoT architecture in monitoring stability and resilience, several test scenarios representative of the real-world operation of low-voltage networks have been defined. These scenarios are implemented within the MATLAB/Simulink simulation environment and facilitate the analysis of the microgrid’s dynamic behavior in the presence of both energetic and electrical disturbances.

In all scenarios, the following parameters are monitored:

Voltage at Node 15 (PV connection point);

Voltage at Node 16 (BESS connection point);

Active and reactive powers of the PV and BESS units;

Local frequency and the RoCoF indicator;

State of Charge (SOC) of the battery;

VDI indices and the Resilience Index.

Furthermore, the effect of PV cell temperature is explicitly included in the simulations through the variation in the MPP voltage as a function of the daily thermal profile, in accordance with standard equations associated with the single-diode model. This approach allows for the realistic capture of PV performance degradation under high-temperature conditions.

The simulations were conducted in MATLAB/Simulink R2023b, utilizing a detailed model of the CIGRE European LV microgrid, extended to include the photovoltaic source (PV), the Battery Energy Storage System (BESS), and IoT analysis modules. The integration time step employed is 1 ms, which is sufficient for capturing the rapid dynamic phenomena associated with irradiance variations, power oscillations, and frequency transitions (RoCoF).

Scenario 1—Reference Operation (Baseline)

This scenario represents the normal operating regime of the PV–BESS microgrid, in the absence of any external disturbances. Photovoltaic production follows a slow, sinusoidal-type variation, used here as a parametric approximation of an irradiance profile characteristic of a typical summer day in Romania. PV cell temperature evolves gradually, in accordance with the diurnal thermal profile, determining corresponding variations in the MPP voltage.

The operational characteristics of this scenario include:

Residential load with an “Evening Peak” type profile;

Operation of the PV unit in power-controlled mode, injecting the available power determined by the MPP;

Operation of the BESS in power-balancing mode, with moderate absorption/injection to compensate for the slow variation in PV power;

The absence of local faults, overloads, or sudden frequency variations.

The main objective of the Baseline scenario is to establish the reference values for the indicators subsequently used in the microgrid stability and resilience analysis, namely:

The Voltage Deviation Index (VDI);

Frequency and RoCoF indicators;

The disturbance resilience index;

The typical amplitude of voltage and power variations under nominal conditions.

These values allow for the calibration of IoT thresholds and the definition of the nominal operating regime, an essential reference for comparison with disturbance scenarios.

The time series obtained for P_PV, P_BESS, nodal voltage V_node₁₅(t), and frequency f(t) indicate a stable regime, with slow variations induced by the PV profile and counterbalanced by the gradual response of the BESS. The nodal voltage is maintained around 230 V with minor deviations, while the frequency varies insignificantly around the nominal value of 50 Hz, confirming the absence of power imbalances.

Under these conditions, the stability indicator values calculated based on the time series confirm the microgrid’s nominal operating regime. The Voltage Deviation Index (VDI) remains well below the admissible threshold, RoCoF exhibits values close to zero, and the resilience index (RI) possesses a value close to unity, indicating a high capacity for maintaining a stable regime [39]. These results define the reference state subsequently utilized for the comparative evaluation of the disturbance scenarios.

Figure 11 illustrates the microgrid dynamics within the Baseline scenario: the moderate variability of PV power is compensated by the adaptive behavior of the BESS, resulting in a stable nodal voltage profile and confirming the correct operation of the PV–BESS–IoT model.

This state represents the absolute reference point for the evaluation of all disturbance scenarios analyzed subsequently.

Scenario 2—Rapid variations in photovoltaic production (Irradiance and temperature fluctuations)

Unlike Scenario 1 (Baseline), where variations are slow and quasi-stationary, Scenario 2 introduces a rapid, high-amplitude disturbance that explicitly tests the dynamic response capacity of the PV–BESS–IoT architecture.

This scenario analyzes the microgrid behavior under conditions of rapid solar resource fluctuations, generated by the passage of dense cloud formations (“solar cloud transient”).

Events of this type are frequent in urban microgrids and lead to abrupt changes in the P–V curve, simultaneously challenging the photovoltaic inverter’s capacity to maintain its Maximum Power Point (MPP) and the BESS’s ability to stabilize local voltage and frequency.

For the simulation, the following disturbing event is considered:

At t = 5 s, irradiance drops suddenly from 1000 W/m² la 400 W/m²;

PV cell temperature remains high (≈55–60 °C), which accentuates the reduction in MPP voltage via the negative dependence dV/dT;

This results in a marked diminution of the active power injected by the PV source.

This combination (G ↓ + T ↑) represents a typical critical situation in real microgrids, as the voltage generated by the panels is reduced simultaneously by two distinct physical mechanisms, affecting nodal stability and power balance.

The response of the BESS is modeled using progressive (“smooth”) dynamics, reflecting the physical limitations of the static converter and the storage system, thereby avoiding the introduction of unrealistic step-type variations in power and voltage.

To highlight the impact of PV fluctuations on microgrid operation, the following parameters are evaluated:

Voltage variation at Node 15, expressed via the VDI;

Local frequency dynamics and RoCoF, as an indicator of active imbalances;

Activation of the BESS P–f and Q–V control as an automatic stabilization mechanism;

Transition of the PV inverter between power-controlled and power-hold regimes;

Recovery time and the system’s Resilience Index (RI).

The BESS response, modeled with progressive dynamics, demonstrates the system’s ability to damp rapid variations and restore the voltage profile within a short interval, while frequency deviation remains moderate (≈49.85 Hz).

Overall, the coordinated PV–BESS behavior confirms that the proposed architecture ensures microgrid stability without triggering protection mechanisms.

This scenario is fundamental as it reproduces exactly the type of disturbances that:

Induce rapid undervoltages in low-voltage networks;

Can lead to the temporary limitation or disconnection of the photovoltaic inverter;

Activate the P–f and Q–V stabilization functions of the BESS;

Test the storage system’s capacity to support power balance in the transient regime;

Allow for the evaluation of the sensitivity of VDI, RoCoF, and RI indicators under realistic operating conditions;

Provide a critical dataset for calibrating thresholds and intelligent edge-processing logic used for the early detection of pre-instability regimes.

The analysis of stability indicators highlights that, following the rapid fluctuation of photovoltaic production, the Voltage Deviation Index (VDI) increases significantly compared to the Baseline scenario, indicating the occurrence of a temporary undervoltage at Node 15. However, VDI values remain within the admissible range, due to the rapid intervention of the BESS via Q–V control.

Regarding frequency stability, RoCoF records moderate negative values associated with the temporary active power deficit, but without exceeding critical thresholds that would lead to protection tripping [40].

The short voltage and frequency recovery time determines a high value of the Resilience Index (RI), confirming the microgrid’s capacity to rapidly absorb disturbances induced by solar resource variability.

In conclusion, Scenario 2 demonstrates that the proposed PV–BESS–IoT architecture is capable of effectively mitigating rapid photovoltaic production fluctuations, maintaining voltage and frequency stability, and ensuring a high resilience index level, without triggering protection mechanisms [41].

Figure 12 represents the dynamic response of the microgrid in Scenario 2, following a solar cloud transient event (sudden irradiance drop from 1000 W/m² la 400 W/m²). The upper panel presents the decrease in photovoltaic power P_PV, the activation of the BESS stabilization control, and the variation in the nodal voltage V_node₁₅.

The lower panel illustrates the evolution of the local frequency and the RoCoF deviation, highlighting the BESS’s capacity to damp power imbalances and bring the microgrid back to the proximity of the steady-state regime.

The results confirm the critical role of the coordinated PV–BESS response in maintaining LV microgrid stability under conditions of extreme solar resource variability.

The graphical analysis of this scenario validates the critical role of the IoT layer in managing rapid transient regimes. The synchronization of the BESS intervention with the moment of the irradiance drop (centered at t = 5 s) demonstrates the superiority of local processing (Edge Computing) over the variability of the solar resource. The control algorithm instantaneously detected the steep negative gradient of the photovoltaic power (dP_Pv/dt) and commanded the compensation without the latencies specific to communication with a central server (Cloud).

Furthermore, the shape of the BESS response curve highlights the successful application of the digital power smoothing function. The IoT architecture prevented an aggressive battery discharge, instead calculating an optimal power injection trajectory, thereby transforming an abrupt production variation into a controlled evolution of voltage and frequency. This distributed intelligence was decisive for maintaining the VDI within safety limits and for preventing the disconnection of the PV inverter via undervoltage protection.

Scenario 3—Local Overload

This scenario analyzes the behavior of the PV–BESS microgrid in the presence of a sudden increase in electricity demand, a phenomenon characteristic of urban low-voltage networks, particularly during peak consumption intervals. Unlike Scenario 2, where the disturbance is induced on the photovoltaic generation side, Scenario 3 introduces energy stress on the demand side, explicitly testing the capacity of the PV–BESS–IoT architecture to maintain nodal stability and power balance under overload conditions.

Within the simulation, a local overload is applied in Zone C, corresponding to the terminal nodes of the network (consumer area), representative of the simultaneous connection of several high-power loads (HVAC systems, electric vehicle charging, appliances).

Considered Disturbance Event:

At t = 5 s, the active load increases suddenly by +2.5 kW (approximately one-third of the local load);

The reactive load increases proportionally, maintaining a constant power factor;

After t = 8 s, the load returns gradually to the initial value.

Throughout this scenario, the photovoltaic unit operates under quasi-stationary conditions, corresponding to medium irradiance (≈700–800 W/m²) and high cell temperatures (≈50–55 °C), without rapid fluctuations, so that the observed effects on the network are attributed exclusively to load variation.

To evaluate the impact of the overload on microgrid operation, the following quantities are analyzed:

Voltage variation at Node 15, quantified by the VDI;

Local frequency deviation and RoCoF dynamics;

The response of the BESS via active and reactive power injection;

The voltage and frequency recovery time;

The resilience index associated with the overload event.

Overall, Scenario 3 demonstrates that the proposed PV–BESS–IoT architecture is capable of efficiently managing significant local overloads, maintaining microgrid stability, and avoiding the triggering of protection mechanisms, confirming the critical role of the storage system in supporting operational safety in modern LV networks.

Figure 13 represents the dynamic response of the microgrid in Scenario 3—Local Overload in Zone C. The upper panel presents the evolution of the photovoltaic power P_PV, maintained constant during the event. The second panel illustrates the activation of the storage system P_BESS, which injects active power to compensate for the local load increase. The third panel highlights the nodal voltage variation V_node₁₅, recording a temporary undervoltage, maintained within admissible limits through BESS intervention. The lower panel presents the local frequency deviation f(t), characterized by a moderate drop during the overload and a progressive return to the nominal value after the disturbance clearance. The results confirm the capacity of the PV–BESS–IoT architecture to maintain voltage and frequency stability in the presence of local overloads and to ensure a controlled dynamic regime of the low-voltage microgrid.

From the perspective of quantitative stability indicators, Scenario 3 is characterized by a temporary increase in the Voltage Deviation Index (VDI), associated with the occurrence of nodal undervoltage following the local overload. The maximum VDI value remains below the admissible threshold imposed by the EN 50160 standard, confirming that the BESS intervention is sufficient to maintain the voltage within acceptable limits.

Simultaneously, the local frequency variation results in moderate values of the RoCoF indicator, significantly lower than those associated with fault-type defects, indicating a controllable energy imbalance. The activation of the BESS P–f control rapidly limits frequency deviation and contributes to damping transient oscillations.

The voltage and frequency recovery time following the overload clearance results in an intermediate Resilience Index (RI) value, situated between the values obtained in Scenario 2 and Scenario 4. This value reflects the more severe nature of the overload compared to PV fluctuations, yet less critical compared to a transient electrical fault.

The results highlight a temporary nodal voltage drop and a moderate frequency deviation, generated by the sudden imbalance between generation and consumption. The activation of the P–f and Q–V control mechanisms of the BESS leads to the rapid injection of active power, compensating for the overload and limiting the amplitude of voltage and frequency deviations. The smooth dynamics used for modeling the BESS response reflect the physical limitations of the power converter and ensure a realistic representation of dynamic behavior, avoiding the occurrence of unrealistic step-type variations [42].

The graphical analysis explicitly validates the efficiency of the proposed IoT architecture, highlighting the superiority of local processing (Edge Computing) over centralized Cloud solutions. The reaction speed observed at t = 4 s confirms that the compensation decision did not suffer from latencies specific to remote transmissions but was taken locally, in the Edge Gateway, based on high-resolution data provided by smart sensors (PMU/Smart Meters) at Node 15.

Furthermore, the IoT architecture managed the transition of the BESS inverter from self-consumption mode to grid support mode exactly at the moment of detecting the violation of alert thresholds for VDI and RoCoF. The ‘smooth’ profile of the power injection (rounded trapezoid shape) is the direct result of the digital control algorithm implemented in the IoT layer, which imposes power ramps to protect power electronics and to prevent sudden shocks (step response) that could further destabilize the grid. Thus, the digital system not only reacts but also optimizes the physical response of the equipment.

Scenario 4—Transient Fault

This scenario analyzes the behavior of the PV–BESS microgrid in the presence of a transient electrical fault, specific to urban low-voltage networks, with a direct impact on local voltage and frequency stability. Unlike Scenarios 2 and 3, which investigate disturbances of an energetic nature associated with photovoltaic production variability and the occurrence of local overloads, Scenario 4 introduces a severe electrical disturbance characterized by a rapid and temporary voltage collapse.

The fault is modeled as a severe voltage sag applied to a branch of the low-voltage network in Zone C, considered equivalent, regarding its effects on voltage, to a single-phase short circuit typical of LV networks. The simulation does not explicitly model phase short-circuit currents, an approach frequently utilized in microgrid stability and resilience studies when the primary focus is the response of voltage, frequency, and distributed sources.

The disturbing event is defined by a sudden drop in the nodal voltage at Node 15 to 0.65 p.u. at the moment of fault occurrence. This value is maintained for a fixed duration of 200 ms, representative of temporary faults cleared by fast protection mechanisms or electric arc self-extinction.

Following fault clearance, the voltage returns progressively toward the nominal value of 1.0 p.u. without introducing artificial transient overvoltages, ensuring that the post-fault dynamics reflect a realistic network recovery process. The selection of the 0.65 p.u. level corresponds to a severe yet realistic voltage sag for low-voltage networks, being sufficiently pronounced to test the coordinated reaction of the photovoltaic source and the BESS without leading to the complete collapse of the microgrid.

During the transient fault, the photovoltaic inverter is constrained to operate in a power-hold regime, limiting active power injection to comply with the operating conditions imposed by the low nodal voltage level. This strategy reflects the real-world behavior of inverters connected to low-voltage networks, which prioritize system stability and equipment protection during voltage sag regimes.

Simultaneously, the Battery Energy Storage System (BESS) is activated to provide dynamic support to the microgrid by injecting active and reactive power in a controlled manner, according to the P–f and Q–V control laws.

The BESS response is modeled with smooth dynamics, limited by the maximum admissible power rate of change, ensuring realistic converter behavior and avoiding the occurrence of unrealistic step-type variations.

Following the fault clearance, the photovoltaic unit progressively returns to the power level corresponding to pre-fault operating conditions, while the BESS gradually reduces its energy contribution, preventing the occurrence of post-fault oscillations.

As a result of this coordinated action, the microgrid voltage and frequency are restored to the vicinity of nominal values, demonstrating the capacity of the PV–BESS architecture to efficiently manage transient electrical disturbances.

To evaluate the impact of the fault on microgrid operation, several electrical quantities and synthetic stability and resilience indicators are analyzed. The variation in the nodal voltage at Node 15 is monitored alongside the Voltage Deviation Index (VDI) to quantify the severity and duration of deviations from the nominal regime.

In parallel, the evolution of the local frequency and the dynamics of the RoCoF indicator are analyzed, highlighting the effects of the temporary power imbalance. The reaction of the storage system is evaluated by monitoring the injected active and reactive power, underscoring its role in damping fault effects and supporting local voltage. Furthermore, voltage and frequency recovery times are determined, serving as the basis for calculating the Resilience Index (RI) associated with the fault event.

By its nature, Scenario 4 represents the most severe test from an electrical perspective among all analyzed scenarios, as it combines a rapid voltage collapse, a sudden power flow imbalance, and the simultaneous stressing of the control mechanisms of both the photovoltaic source and the BESS.

The analysis of the obtained results highlights the capacity of the PV–BESS–IoT architecture to maintain voltage stability, limit frequency deviations and RoCoF values, and rapidly restore the normal operating regime, thereby preventing the propagation of the disturbance into a local blackout.

Figure 14 illustrates the dynamic response of the PV–BESS microgrid in the presence of a transient severe voltage sag fault (Scenario 4). A voltage sag down to 0.65 p.u. is applied at Node 15 for a duration of 200 ms, representative of a temporary fault in a low-voltage network. The upper panel presents the evolution of the photovoltaic power P_PV, limited during the fault by the transition of the PV inverter into the power-hold regime. The second panel highlights the active power support provided by the BESS, injected with smooth dynamics constrained by power variation limits. The third panel presents the nodal voltage variation V_node₁₅, marking the voltage sag and the progressive return toward the nominal value, without the occurrence of post-fault overvoltages. The lower panel shows the evolution of the local frequency, characterized by a moderate deviation and rapid stabilization. The results confirm the efficiency of the coordinated PV–BESS operation in mitigating disturbances and restoring stable operating conditions following a transient electrical fault in a low-voltage microgrid.

The graphical analysis of this scenario validates the capacity of the Edge-IoT architecture to ensure the critical Fault Ride Through (FRT) function. The extremely narrow time window of the fault (200 ms) imposed processing speed requirements that a centralized Cloud solution could not have sustained.

The local IoT node detected the voltage collapse at the millisecond level and prioritized the triggering of the emergency regime, inhibiting standard protections that would have otherwise led to the nuisance tripping (unintended disconnection) of the PV inverter.

Furthermore, the system’s distributed intelligence allowed for the rapid discrimination between a grid fault (Scenario 4) and a simple load or irradiance variation, activating specific strategies: power-hold for the PV unit and rapid injection for the BESS. The voltage recovery to the nominal value without dangerous overvoltages is the direct result of the coordinated recovery algorithm implemented at the Edge level, which managed the controlled relaxation of the post-fault power injection.

4.5. Synthesis of Resilience Indicators

To enable a coherent comparative evaluation of the PV–BESS microgrid’s behavior in the presence of various disturbance types, this section presents a synthesis of the resilience indicators defined and calculated within the analyzed scenarios. The Resilience Index (RI), introduced in Section 4.2, offers a unified measure of the microgrid’s capacity to return to a safe operating regime following the occurrence of a disturbance, by correlating the duration of the disturbing event with the voltage and frequency recovery time.

The comparative analysis focuses on Scenarios 2–4, where energetic and electrical disturbances are introduced, with Scenario 1 (Baseline) being utilized exclusively as a reference for the microgrid’s nominal operation. For each scenario, the following aspects are considered: the type of disturbance, its characteristic duration, the system recovery time, and the resulting global resilience level, evaluated via the normalized form of the RI.

To facilitate the comparison between the analyzed disturbance scenarios, Table 4 synthesizes the characteristic values of the resilience index obtained in Scenarios 2–4. This synthesis highlights the influence of disturbance severity on the recovery time and the global resilience level of the PV–BESS microgrid, based on indicators calculated from local voltage and frequency measurements.

The results synthesized in Table 4 indicate a progressive decrease in the resilience level correlated with the increase in disturbance severity. Rapid fluctuations in photovoltaic production (Scenario 2) are efficiently damped by the BESS, leading to reduced recovery times and a high level of resilience (RI). The local overload analyzed in Scenario 3 generates more pronounced disturbances, reflected by an intermediate value of the RI. In Scenario 4, the transient electrical fault (severe voltage sag) produces the most severe deviation from the nominal regime, resulting in longer recovery times and, implicitly, the lowest level of resilience.

This hierarchy validates the sensitivity of the RI in quantifying disturbance severity, while simultaneously confirming the efficiency of the proposed IoT architecture. The obtained values demonstrate that microgrid performance is not merely the result of storage capacity, but the direct effect of data processing at the Edge level. The IoT system enabled the automatic identification of the event type (energetic vs. electrical) and the real-time calculation of performance indices, which were decisive elements for adapting the response strategy and maximizing resilience across all analyzed scenarios [43].

5. Results and Discussions

Based on the stability and resilience indicators defined in Section 4 and calculated via the proposed IoT architecture, this section comparatively discusses the behavior of the PV–BESS microgrid across the four analyzed scenarios, highlighting the role of intelligent monitoring and coordinated intervention in preventing operational degradation and local blackouts.

5.1. Comparative Analysis of Voltage and Frequency Stability (VDI and RoCoF)

The stability analysis of the PV–BESS microgrid in the four investigated scenarios is performed based on the Voltage Deviation Index (VDI) and the Rate of Change of Frequency (RoCoF), defined in Section 4.1 and Section 4.3. These indicators are calculated based on the voltage and frequency time series presented in Figure 11, Figure 12, Figure 13 and Figure 14 and allow for the characterization of the instantaneous severity of disturbances, as well as their impact on the microgrid’s local stability.

To provide a quantitative overview of the proposed system’s performance, Table 5 centralizes the key metrics extracted from the simulation results for the four test scenarios. This synthesis highlights the disturbance hierarchy and the effectiveness of the IEP logic in maintaining microgrid stability, demonstrating the system’s ability to restore nominal parameters within an extremely short timeframe.

It is important to emphasize that without the BESS Q-V control intervention, the voltage deviations would have been significantly higher. For instance, in Scenario 4 (grid fault), estimates suggest that |VDI| would have exceeded the critical threshold of 0.45 p.u., potentially leading to a cascading disconnection of the remaining PV sources due to under-voltage protections. The rapid BESS intervention maintains the |VDI| at 0.345 p.u. and ensures a return to a stable state in only 210 ms.

In Scenario 1 (Baseline), the microgrid operates in a quasi-stationary regime, without significant external disturbances. As observed in Figure 11, the nodal voltage at Node 15 remains very close to the nominal value of 230 V, leading to very low VDI values, well below the admissible threshold of 0.1 imposed by EN 50160. Simultaneously, the local frequency oscillates insignificantly around the nominal value of 50 Hz and RoCoF presents values close to zero, indicating the absence of rapid active power imbalances. This behavior confirms the existence of a stable equilibrium between photovoltaic production, residential load, and the moderate intervention of the BESS, with the Baseline scenario constituting the reference for evaluating the disturbed scenarios.

In Scenario 2—Rapid Variations in Photovoltaic Production, the solar cloud transient event produces a sudden drop in irradiance from 1000 W/m² la 400 W/m², generating a temporary active power deficit. According to Figure 12, this disturbance determines a transient undervoltage at Node 15, reflected by a temporary increase in the VDI.

The rapid intervention of the BESS Q–V control limits the undervoltage amplitude to approximately 5% relative to the nominal value, well below levels that could lead to protection tripping, whereas a situation without voltage support could have led to estimated deviations exceeding 15%. This result is consistent with frequency support strategies analyzed by Raoufi, H. et al. [44], yet the proposed architecture demonstrates a superior reaction speed. Unlike approaches tolerating higher communication latencies described by Cebulla, F. et al. [2], local processing enabled maintaining the VDI below critical thresholds without depending on a central server. Voltage recovery occurs within a short interval, on the order of under 300 ms, highlighting the system’s capacity to react rapidly to energetic disturbances.

From the perspective of frequency stability, the local frequency drops moderately to approximately 49.85 Hz, and RoCoF records moderate negative values, corresponding to the temporary active power deficit. However, RoCoF values remain below the critical thresholds used in protection applications, confirming that the energy imbalance is efficiently damped through the activation of the BESS P–f control.

In Scenario 3—Local Overload, the sudden increase in active load by +2.5 kW at t = 4 s induces a more persistent energy imbalance than in Scenario 2. As observed in Figure 13, this event leads to a more pronounced nodal undervoltage, reflected by higher VDI values compared to the PV fluctuation scenario. However, maximum VDI values remain within the admissible range, due to the voltage support provided by the BESS via Q–V control. The efficiency of this local compensation strategy confirms that reported in similar studies regarding low-voltage networks with high PV penetration [11,15], validating the capacity of distributed storage systems to manage local congestion. The local frequency records a moderate drop during the overload, and RoCoF indicates a controllable energy imbalance, significantly lower than that associated with severe electrical faults. The activation of the BESS P–f control contributes to limiting frequency deviation and damping transient oscillations until the load returns to the initial value.

Scenario 4—Transient Fault represents the most severe situation from the electrical stability perspective. According to Figure 14, applying a severe voltage sag down to 0.65 p.u. for a duration of 200 ms determines the highest VDI values among all analyzed scenarios. This rapid voltage collapse reflects the electrical nature of the disturbance, significantly more aggressive than variations in energetic origin. From the frequency perspective, the fault induces more accentuated deviations and high RoCoF values, associated with the sudden power imbalance. However, limiting photovoltaic power via the power-hold regime and the rapid intervention of the BESS allow for frequency stabilization and the progressive return of voltage toward the nominal value, without the occurrence of post-fault overvoltages.

The comparative analysis of VDI and RoCoF indicators, based on Figure 11, Figure 12, Figure 13 and Figure 14, highlights a clear hierarchy of disturbance severity: stable regime in Scenario 1, rapid but well-damped energetic disturbances in Scenarios 2 and 3, and finally, a severe electrical disturbance in Scenario 4. The VDI and RoCoF indicators thus prove to be sensitive tools for evaluating instantaneous stability and for differentiating disturbance types in low-voltage microgrids.

The indicators analyzed in this section are calculated in real-time based on measurements provided by the proposed IoT architecture and constitute the essential inputs for evaluating the microgrid’s recovery capacity, discussed subsequently via the resilience index RI. While VDI and RoCoF indicators characterize the instantaneous severity of disturbances and their immediate impact on voltage and frequency stability, the microgrid’s capacity to return to a safe regime following these events is evaluated synthetically by the resilience index RI, analyzed comparatively in Section 5.2.

Here is the academic translation of Section 5.2, maintaining the terminology and formatting consistent with the previous sections.

5.2. Interpretation of the Resilience Index (RI)

The Resilience Index (RI) offers a synthetic evaluation of the PV–BESS microgrid’s capacity to return to a safe operating regime following the occurrence of a disturbance, integrating into a single parameter the combined effects of voltage deviations, frequency imbalances, and recovery dynamics. Unlike the VDI and RoCoF indicators, which describe the instantaneous severity of disturbances, the RI reflects the global behavior of the system in the post-disturbance phase [45].

The resilience index values obtained for Scenarios 2–4 are synthesized in Table 4 and highlight a clear correlation between the nature of the disturbance, the system recovery time, and the global resilience level of the microgrid.

In Scenario 2, characterized by rapid variations in photovoltaic production, the microgrid exhibits a high level of resilience. Although sudden irradiance fluctuations induce transient undervoltages and moderate frequency deviations, the rapid intervention of the BESS limits their severity and allows for rapid voltage and frequency recovery. The short return time leads to a high RI value, confirming that disturbances of an energetic origin, even rapid ones, can be efficiently absorbed by the PV–BESS architecture when managed through adequate control mechanisms.

In Scenario 3, the local overload introduces a more persistent energy imbalance, which is reflected in a longer recovery time compared to Scenario 2. Although voltage and frequency deviations remain within admissible limits due to the support provided by the BESS, the duration required to restore the nominal regime is longer. This aspect results in an intermediate value of the resilience index RI, situated between that obtained for PV fluctuations and that associated with severe electrical faults. The result indicates that local overloads represent more demanding disturbances for the microgrid than production variations yet remain manageable through coordinated control strategies.

Scenario 4, corresponding to a transient fault, generates the lowest resilience index value among all analyzed scenarios. The severe voltage sag down to 0.65 p.u. induces major deviations in VDI and RoCoF indicators and simultaneously stresses the control mechanisms of the photovoltaic source and the BESS. Although the microgrid manages to recover without collapse, the recovery time is significantly longer compared to disturbances of an energetic nature, which is reflected in a minimum RI value. This situation confirms the critical nature of transient electrical faults for the stability and resilience of low-voltage microgrids.

The comparative analysis of the RI for Scenarios 2–4 highlights a coherent hierarchy of disturbance severity, in full accordance with the observations derived from the VDI and RoCoF indicators. Rapid energetic disturbances lead to high resilience values, local overloads determine an intermediate level, while severe electrical faults generate the lowest recovery capacity. This quantification, validated through simulation, complements theoretical resilience assessment frameworks proposed in the literature [46] and confirms recent hypotheses stating that IoT integration transforms resilience from a passive monitoring concept into an active operational parameter [43].

By integrating recovery time into a single indicator, RI offers an efficient tool for the comparative evaluation of microgrid performance and for assessing the impact of implemented control strategies. The obtained values confirm that the analyzed PV–BESS architecture is capable of managing a wide range of disturbances yet simultaneously highlight the importance of rapid and coordinated intervention in the case of severe electrical faults, an aspect discussed in detail in the following section dedicated to the role of the IoT architecture and edge processing [47].

5.3. The Role of IoT Architecture and Edge Processing

The results presented in Section 5.1 and Section 5.2 highlight that the stability and resilience of the PV–BESS microgrid do not depend exclusively on the existence of distributed sources and the storage system, but essentially on the system’s capacity to rapidly detect disturbances and trigger corrective interventions within a timeframe compatible with the dynamics of transient phenomena. In this context, the proposed IoT architecture and edge-level processing play a central role in transforming VDI, RoCoF, and RI indicators from simple monitoring quantities into active decision-making instruments.

In the analyzed architecture, nodal voltage, local frequency, active and reactive powers, and the BESS State of Charge are acquired in real-time by IoT sensors distributed at critical microgrid nodes. These raw data are transmitted via the MQTT protocol to edge processing nodes, where they are processed with low latency for the calculation of stability indicators VDI and RoCoF. Thus, rapid voltage and frequency deviations are identified immediately upon occurrence, without depending on a centralized offline analysis process. This decentralized approach overcomes the latency and bandwidth limitations of cloud-centric architectures identified in the review conducted by Yıldırım, F et al. [20], confirming the advantages of real-time data processing demonstrated by Molokomme, D.N. et al. [21] for critical Smart Grid applications.

Specifically, the achieved response time of under 300 ms underscores a significant ‘edge advantage’ over traditional SCADA (Supervisory Control and Data Acquisition) systems, which typically operate with cycle times between 2 and 10 s. By performing local processing, the system also avoids the inherent latencies and jitter of cloud-based round-trip communications, ensuring that primary stability supports (Q-V and P-f) are activated within the millisecond windows required by the microgrid’s transient dynamics, a performance metric that aligns with recent high-impact research in edge-enabled stability [2].

The quantitative superiority of the proposed decentralized approach is visually synthesized in Figure 15. The timeline diagram illustrates the sequence of events following a disturbance at t = 0, highlighting that only the Edge-IEP architecture operates within the ‘Critical Stability Window’ (under 300 ms). In contrast, Cloud-based analytics and traditional SCADA systems exhibit response latencies that exceed the timeframe required to prevent microgrid instability, thereby validating the essential role of edge-level processing for real-time primary support.

The VDI indicator is utilized at the edge level as the primary mechanism for detecting voltage disturbances. Rapid increases in VDI above established thresholds signal the occurrence of local undervoltages or overvoltages and automatically trigger voltage support strategies via the BESS Q–V control. Similarly, RoCoF offers early detection of active power imbalances, being significantly more sensitive than absolute frequency deviation. High RoCoF values indicate an imminent energy imbalance and allow for the preventive activation of the BESS P–f control, before the frequency reaches critical limits.

While VDI and RoCoF ensure rapid disturbance detection and the triggering of immediate interventions, the Resilience Index (RI) is utilized at the edge level for evaluating the recovery dynamics of the microgrid. By monitoring the time required for voltage and frequency to return to admissible intervals, RI allows for the classification of event severity and the assessment of the efficiency of applied control strategies. Thus, RI is not merely a post-event performance indicator but becomes a feedback parameter utilized for adapting microgrid operating strategies.

The integration of these indicators into the edge processing logic facilitates the realization of a hierarchical control loop: VDI and RoCoF ensure rapid detection and immediate intervention triggering, while RI provides a global assessment of recovery capacity and post-disturbance stability. Based on RI values, triggering thresholds and BESS support strategies can be adjusted, or alerts can be generated for the upper monitoring level for long-term analysis.

The implementation of this hierarchical control via deterministic ‘if-then-else’ logic (as illustrated in Figure 10) provides a high degree of predictability and hard real-time execution. Unlike probabilistic Machine Learning models, which may introduce uncertainty in decision-making, this deterministic approach ensures a reliable and explainable safety mechanism, which is vital for the operational security of critical electrical infrastructure.

The results obtained in the four analyzed scenarios demonstrate that this IoT–edge architecture allows for corrective interventions within a timeframe comparable to the dynamics of the analyzed disturbances. In Scenarios 2 and 3, the rapid detection of VDI increases and RoCoF deviations leads to the prompt activation of energy support, limiting disturbance severity and resulting in high resilience index values. In Scenario 4, although the disturbance is significantly more severe, the IoT–edge architecture enables the rapid coordination of the photovoltaic inverter’s power-hold regime and the BESS power injection, preventing microgrid collapse and facilitating controlled post-fault recovery.

Consequently, the proposed IoT architecture is not limited to the passive monitoring of the microgrid but constitutes an active distributed control element, capable of transforming VDI, RoCoF, and RI indicators into operational decisions with a direct impact on system stability and resilience. This approach highlights the potential of IoT–edge solutions for preventing local blackouts and enhancing the operational safety of low-voltage microgrids, aspects with direct implications for the operation of modern European networks.

5.4. Practical Implications for European Low-Voltage Networks

The results obtained in this work have direct implications for the operation and modernization of European low-voltage (LV) networks, which are undergoing an accelerated transition toward decentralized systems with high penetration of renewable sources and storage systems. In this context, Distribution System Operators (DSOs) face challenges related to maintaining power quality, voltage stability, and preventing local blackout incidents, given a traditional infrastructure designed for unidirectional power flows.

The analysis of the investigated scenarios reveals that the analyzed disturbances—rapid variations in photovoltaic production, local overloads, and transient faults—are representative of real-world situations encountered in European urban and peri-urban LV networks. The VDI and RoCoF indicators highlight that these disturbances can rapidly lead to violations of admissible voltage and frequency limits if not managed through adaptive control mechanisms. Consequently, conventional monitoring, based on infrequent measurements and centralized analysis, becomes insufficient for ensuring operational stability.

From the perspective of compliance with the EN 50160 standard, the results demonstrate that utilizing a coordinated PV–BESS, supervised via an IoT architecture with edge-level processing, allows for maintaining nodal voltage within the admissible range even in the presence of severe disturbances. In Scenarios 2 and 3, voltage deviations remain below critical thresholds, while in Scenario 4, although a severe voltage sag occurs, the rapid recovery prevents disturbance propagation and reduces the duration of non-compliance. This behavior is essential for DSOs, who are responsible for adhering to power quality requirements at consumer delivery points.

An aspect of particular relevance for distribution operators is the prevention of local blackouts and cascading protection trips. The obtained results indicate that the early detection of VDI increases and RoCoF deviations, followed by the rapid intervention of the BESS, allows for limiting disturbance severity before classical protections are triggered. Particularly in the case of transient faults, the proposed architecture contributes to maintaining load supply and reducing the risk of uncontrolled microgrid separation, a critical aspect for distribution service continuity [48].

Regarding practical implementation, the proposed IoT architecture presents significant advantages in terms of scalability and progressive integration into existing networks. The use of low-cost IoT sensors, lightweight communication protocols (MQTT), and distributed processing at the edge enables system expansion without major modifications to the communication infrastructure or existing SCADA systems. This approach is compatible with the European strategy for distribution grid digitalization and the development of the “active distribution networks” concept [49].

Furthermore, the clear separation between local monitoring, edge processing, and centralized analysis offers DSOs the flexibility to adapt the automation level according to node criticality and available resources. Microgrids can operate autonomously during transient regimes, benefiting from rapid local decisions, while aggregated data can be utilized for long-term analysis, planning, and infrastructure investment optimization.

Consequently, the results of this work highlight the potential of IoT–edge–BESS architectures to support the transition of European LV networks toward more resilient, flexible, and secure systems. The integration of VDI, RoCoF, and RI indicators into a smart monitoring and distributed control framework represents a pragmatic solution for distribution operators aiming to enhance network reliability and reduce the risk of local blackouts in the context of the continuous growth of distributed renewable sources.

Here is the academic translation of Section 5.5, maintaining the formal tone required for the “Limitations and Future Work” section of a scientific paper.

5.5. Limitations and Future Research Directions

Although the results presented in this work demonstrate the efficiency of the proposed PV–BESS–IoT architecture in improving the stability and resilience of low-voltage microgrids, the study presents a series of inherent limitations that must be acknowledged and which open clear directions for future research.

A primary limitation of the current study is its focus on a single, radial low-voltage (LV) network topology based on the CIGRE benchmark. While this is representative of many distribution feeders, the generalizability of the results to meshed configurations or medium-voltage (MV) networks warrants further discussion. In such systems, the R/X ratios and impedance characteristics differ significantly, which may affect the propagation of disturbances and the sensitivity of the VDI and RoCoF indicators. Future research will aim to validate the IoT–edge architecture across diverse network topologies to ensure its robustness in varied operational environments.

Another limitation is related to the simulative nature of the study. The analysis is conducted based on a detailed model of the CIGRE European LV network, implemented in the MATLAB/Simulink environment, which allows for the rigorous investigation of the microgrid’s dynamic behavior under controlled conditions. However, the results are not yet validated through real experimental measurements or Hardware-in-the-Loop (HIL) tests. Differences between theoretical power converter models and actual equipment behavior may influence the observed transient dynamics, particularly during severe disturbance regimes.

A second limitation is associated with electrical fault modeling. In this work, fault-type defects are represented by an equivalent voltage sag model, without the explicit modeling of phase short-circuit currents or detailed electromagnetic phenomena. This approach is adequate for analyzing voltage stability, frequency stability, and distributed source response, but it does not allow for the evaluation of classical protection behavior or detailed interactions between relays and converters under severe short-circuit regimes.

While the current study primarily focuses on stability indicators, the coordination with specific protection settings, such as relay grading and selectivity, was not explicitly modeled. This represents a clear direction for future work, where the interaction between fast edge-based control and legacy protection systems will be analyzed to ensure seamless integration in larger distribution networks. However, the implemented current saturation limits ensure that the BESS support does not reach levels that would typically cause upstream protection misoperation. Additionally, the current BESS control assumes balanced power injection; future iterations of the IEP logic will explore per-phase optimization to specifically address high degrees of load asymmetry in low-voltage feeders.

Furthermore, the edge processing architecture utilized in this study relies on a deterministic decision logic, grounded in thresholds for VDI, RoCoF, and RI indicators. Although this approach is robust, explainable, and suitable for implementation on embedded devices with constrained resources, it does not fully exploit the potential of advanced machine learning methods. The predictive models used can be extended in the future toward machine learning or deep learning algorithms, capable of learning complex patterns from historical data and anticipating disturbances before their full manifestation.

In this context, recent advancements in anomaly detection frameworks, such as the use of invariant normal region prototypes combined with Segment Anything Models (SAM) for electrical infrastructure components, demonstrate the potential of integrating high-level computer vision and deep learning into edge-based diagnostic systems [50].

Another important limitation is related to the absence of real IoT communication variability. Within the current simulation model, an ideal, latency-free communication between the IoT edge nodes, gateways, and the cloud platform is assumed. While the results demonstrate the theoretical effectiveness of the IEP logic, real-world network constraints such as jitter, packet loss, and variable latency were not explicitly modeled. Analyzing the impact of these communication network factors on microgrid stability and resilience represents an important direction for future research, which will focus on integrating network emulation tools (e.g., NS-3 or Mininet) to assess the impact of communication delays on the overall stabilization time of the microgrid.

Regarding future development directions, an essential step constitutes the experimental validation of the proposed architecture. Implementing the system on real hardware platforms, using industrial controllers or dedicated microcontrollers (ESP32, STM32, ARM Cortex-M), in combination with commercial inverters and BESSs, would allow for performance evaluation under real operating conditions. Integrating HIL tests would represent a valuable intermediate stage between simulation and field implementation.

A complementary direction is the extension of the architecture toward islanded operation and grid reconnection scenarios, as well as the analysis of the interaction between multiple interconnected microgrids. These situations are increasingly relevant in the context of active network development and the concept of community microgrids.

Additionally, integrating advanced machine learning algorithms at the edge and cloud levels could enable a transition from reaction to disturbances toward a predictive control strategy, oriented toward the active prevention of instabilities. Beyond HIL validation, future research will explore ‘hybrid intelligence’ models, where the deterministic edge layer handles safety-critical, millisecond-range control, while cloud-based Machine Learning performs higher-level tasks, such as long-term energy optimization and asset health forecasting, using aggregated data streams from the edge nodes. Utilizing historical data collected via the IoT infrastructure could lead to the adaptive optimization of VDI and RoCoF thresholds and a more precise real-time estimation of the resilience index.

In conclusion, although this work demonstrates the feasibility and efficiency of an IoT–edge–BESS architecture for enhancing low-voltage PV microgrid resilience, it simultaneously constitutes a starting point for future research oriented toward experimental validation, functional expansion, and the integration of advanced artificial intelligence into distributed energy systems.

6. Conclusions

This paper analyzed the stability and resilience of a low-voltage microgrid with photovoltaic integration and a battery energy storage system (PV–BESS), utilizing a distributed IoT architecture and a simulation framework based on the CIGRE European LV reference network. The study aimed to evaluate the dynamic behavior of the microgrid in the presence of disturbances representative of modern LV networks, with an emphasis on preventing operational degradation and local blackouts.

Based on the defined quantitative indicators—Voltage Deviation Index (VDI), Rate of Change of Frequency (RoCoF), and Resilience Index (RI)—the comparative analysis of the disturbance scenarios led to the following specific conclusions:

Scenario 2 (Rapid Variations in PV Production): Irradiance fluctuations induce moderate disturbances, which are efficiently damped by the rapid intervention of the BESS. The short return time to nominal parameters is reflected in high values of the Resilience Index (RI).

Scenario 3 (Local Overload): Energy imbalances caused by sudden load increases generate more pronounced voltage and frequency deviations. Although the system remains stable, the time required for recovery is longer, resulting in an intermediate level of resilience.

Scenario 4 (Transient Fault): This proved to be the most severe disturbance, characterized by a deep voltage sag (down to 0.65 p.u.) and a sudden power imbalance. The scenario generated the highest VDI and RoCoF values and, implicitly, the lowest value of the Resilience Index.

A core trade-off validated by this study is the achievement of high-speed, robust control for critical grid functions through explainable, deterministic edge logic, rather than more complex and data-intensive AI/ML approaches. While machine learning offers advanced predictive capabilities, the deterministic ‘if-then-else’ logic implemented at the edge ensures the hard real-time execution and predictability vital for the operational security of critical electrical infrastructure, especially during the millisecond windows of transient phenomena.

An essential result of the study is the demonstration of the critical role of the IoT architecture and Edge-level processing in managing these regimes. Simulations validated that VDI and RoCoF indicators can be calculated in real-time based on local measurements, allowing for the detection of pre-instability regimes and the triggering of BESS interventions with minimal latencies. Thus, the proposed indicators transcend the role of simple monitoring, becoming active operational decision tools that allow for the discrimination between fault types and the adaptation of the response strategy.

From a practical perspective, the results confirm that integrating storage systems coordinated via intelligent monitoring contributes significantly to increasing the resilience of low-voltage networks, maintaining power quality parameters within the limits of the EN 50160 standard. The proposed architecture is scalable and offers a viable solution for Distribution System Operators (DSOs) aiming to reduce the risk of local blackouts in areas with high penetration of renewable sources. Specifically, it provides an incremental modernization path that complements existing infrastructure, as it adds a resilient ‘distributed intelligence’ layer without requiring the replacement of legacy SCADA systems.

Overall, the paper demonstrates that the integrated assessment of stability and resilience, supported by distributed data processing, represents a necessary direction for the transition toward smart grids. The obtained results constitute a solid basis for future research, including Hardware-in-the-Loop (HIL) experimental validation and the integration of machine learning algorithms for advanced disturbance prediction.

Author Contributions

Conceptualization, D.C.L., T.L. and D.P.; methodology, D.C.L., D.C.P. and G.B.; software, F.G.P. and A.M.T.; validation, D.C.L., D.P. and T.L.; formal analysis, D.I. and A.M.T.; investigation, D.C.L. and F.G.P.; resources, T.L. and G.B.; data curation, F.G.P. and G.B.; writing—original draft preparation, D.I. and D.P.; writing—review and editing, T.L., D.C.L. and D.P.; visualization, T.L., D.I. and D.C.L.; supervision, D.P., T.L. and F.G.P. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.

Acknowledgments

The authors have reviewed and edited the output and take full responsibility for the content of this publication.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:

LV	Low-voltage
PV	Photovoltaic
IEP	Intelligent Edge Processing
BESS	Battery Energy Storage System
VDI	Voltage Deviation Index
RoCoF	Rate of Change of Frequency
RI	Resilience Index
DSOs	Distribution System Operators
PV-BESS	Photovoltaics and Battery Energy Storage System
EV	Electric vehicle
PCC	Point of Common Coupling
ENS	Energy Not Supplied
EMS	Energy Management System
FDD	Fault Detection and Diagnostics
ONAN	Oil Natural Air Natural
PLL	Phase-Locked Loop
MPPT	Maximum Power Point Tracking
STC	Standard Test Condition
MPP	Maximum Power Point
NOCT	Nominal Operating Cell Temperature
IncCond	Incremental Conductance
P&O	Perturb and Observe
IoT	Internet of Things
DC	Direct Current
AC	Alternating Current
PI	Proportional-Integral
THD	Total Harmonic Distortion
BMS	Battery Management System
SOC	State of Charge
FRT	Fault Ride Through
HIL	Hardware-in-the-Loop
SCADA	Supervisory Control and Data Acquisition
CIGRE	International Council on Large Electric Systems
XAI	Explainable AI
RES	Renewable energy sources
HVAC	Heating, Ventilation and Air Conditioning
TSN	Time-Sensitive Networking
URLLC	Ultra-Reliable Low-Latency Communications
MQTT	Message Queuing Telemetry Transport
OPC UA	Open Platform Communications Unified Architecture
IEC	International Electrotechnical Commission
MV	Medium Voltage
SOA	Safe Operating Area
IGBT	Insulated-Gate Bipolar Transistor
ZIP	Constant Impedance–Constant Current–Constant Power
RTU	Remote Terminal Unit
TCP	Transmission Control Protocol
EI	Edge Intelligence
SAMs	Segment Anything Models
AI	Artificial Intelligence
ML	Machine Learning

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1 Timeline comparison of system response latencies for stability recovery following a grid disturbance at t = 0 s. The proposed Edge-IEP architecture achieves a response in the sub-300 ms range, outperforming cloud-based analytics (limited by internet jitter) and traditional centralized SCADA systems (constrained by polling cycles).

Figure 2 Radial topology of the CIGRE LV network, featuring single-phase and three-phase loads, a Photovoltaic (PV) source, and a Battery Energy Storage System (BESS).

Figure 3 I–V and P–V curves for a 400 W PV module.

Figure 4 Daily solar irradiance profile, G(t).

Figure 5 Daily profile of the PV cell temperature, T_cell(t).

Figure 6 Performance comparison of Incremental Conductance (IncCond) and Perturb and Observe (P&O) MPPT algorithms under rapid power variations.

Figure 7 Dynamic response of the photovoltaic power and voltage at Node 15 during a sudden drop in solar irradiance.

Figure 8 Proposed hardware and communication architecture, illustrating the interconnection between physical, edge, and cloud layers. The different background colors distinguish the Physical (green), Edge (yellow), and Cloud (blue) layers. Solid blue lines represent physical/wired connections, while dashed red lines indicate wireless communication protocols (MQTT, HTTPs).

Figure 9 Intelligent edge processing architecture for PV–BESS microgrid stability assessment.

Figure 10 Flowchart of the deterministic IEP decision logic for stability assessment and BESS/PV control coordination.

Figure 11 Dynamic response of the microgrid in the Baseline scenario: P_PV, P_BESS, V_node₁₅(t) and f(t).

Figure 12 Microgrid response in Scenario 2: (a) PV power evolution (blue line); (b) BESS activation (dashed red line); (c) nodal voltage variation (black line); and (d) frequency dynamics (green line) following the solar cloud transient event.

Figure 13 Dynamic response of the microgrid to local overload: (a) PV power evolution (blue line); (b) BESS activation (dashed red line); (c) nodal voltage variation (black line); and (d) frequency dynamics (green line).

Figure 14 Voltage, frequency, and power dynamics during a severe voltage sag in the PV–BESS microgrid: (a) PV power evolution (blue line); (b) BESS activation (orange line); (c) nodal voltage variation (green line); and (d) frequency dynamics (purple line).

Figure 15 Timeline diagram of the response latencies for a disturbance at t = 0: Comparative analysis between the proposed Edge-IEP architecture, Cloud-based analytics, and traditional SCADA systems.

Table 1

CIGRÉ European LV network parameters used in the simulation.

Parameter	Symbol/Unit	Value
Nominal Voltage	V_nom [V]	400/230
Nominal Frequency	f_nom [Hz]	50
Number of Nodes	—	18
Total Grid Length	— [m]	900
Main Conductor Type	—	NAYY (Al) 150 mm²
Phase Resistance	R^′ [Ω/km]	0.206
Phase Reactance	X^′ [Ω/km]	0.080
Neutral Impedance	—	included (CIGRÉ model)
Grid Type	—	Radial, 3-ph + N
MV/LV Transformer	20 kV/400 V	400 kVA
Transformer Resistance	R_tr [%]	1.275
Transformer Reactance	X_tr [%]	4.0
Single-phase Loads	—	Nodes 1–7
Three-phase Loads	—	Nodes 8–13
PV/BESS Integration Zone	—	Nodes 15–16
Load Power Factor	—	0.95 inductive

Table 2

Photovoltaic module parameters used in the simulation (single-diode model).

Parameter	Symbol/Unit	Value
Nominal Module Power	P_MPP [W]	400
Reference Irradiance	G_ref [W/m²]	1000
Reference Temperature	T_ref [°C]	25
Current at Maximum Power	I_MPP [A]	9.86
Voltage at Maximum Power	V_MPP [V]	40.6
Short-Circuit Current	I_sc,ref [A]	10.80
Open-Circuit Voltage	V_oc,ref [V]	49.50
Temperature Coefficient of I_sc	α [%/°C]	+0.05
Temperature Coefficient of V_oc	β [%/°C]	–0.29
Number of Cells in Series	Ns	72 (sau 144 half-cut)
Total PV Generator Power	P_{PV, nom} [kW]	3.2

Table 3

BESS parameters used in the simulation.

Parameter	Symbol/Unit	Value
Rated Power	P_{BESS, max} [kW]	±2.5
Reference Voltage	V_ref [p.u]	1.0
Total Capacity	C_nom [kWh]	30
Initial SOC	SOC₀ [%]	70
Lower SOC Limit	SOC_min [%]	20
Upper SOC Limit	SOC_max [%]	90
P–f Control Coefficient	K_f [W/Hz]	20
Q–V Control Coefficient	K_v [var/V]	5
Global Efficiency	Ƞ [%]	95

Table 4

Characteristic values of the resilience index for the PV–BESS microgrid in the analyzed disturbance scenarios.

Scenario	Disturbance Type	Disturbance Duration (T_dist)	Recovery Time (T_rec)	Resilience Index (RI_norm)	Level
Scen. 2	Rapid PV fluctuations	continuous variation	short (≈0.24 s)	high	Good
Scen. 3	Local overload	4 s	medium (≈0.26 s)	intermediate	Medium
Scen. 4	LV Fault—severe voltage sag (0.65 p.u.)	200 ms	long (≈0.21 s)	low	Low

Table 5

Summary of key stability metrics across the four test scenarios.

Scenario	Microgrid State	\|VDI\|_max [p.u.]	\|RoCoF\|_max [Hz/s]	T_rec [ms]
Scenario 1	Normal	<0.005	<0.01	N/A
Scenario 2	Alert	0.059	0.15	≈240
Scenario 3	Alert	0.050	0.12	≈260
Scenario 4	Pre-instability	0.345	0.45	≈210

References

1. Blaabjerg, F.; Chen, Z.; Kerekes, D. Power Electronics as Efficient Interface in Dispersed Power Generation Systems. IEEE Trans. Power Electron.; 2004; 19, pp. 1184-1194. [DOI: https://dx.doi.org/10.1109/TPEL.2004.833453]

2. Cebulla, F.; Naegler, T.; Pohl, M. Electrical energy storage in highly renewable European energy systems: Capacity requirements, spatial distribution, and storage dispatch. J. Energy Storage; 2017; 14, pp. 211-223. [DOI: https://dx.doi.org/10.1016/j.est.2017.10.004]

3.EN 50160 Voltage Characteristics of Electricity Supplied by Public Electricity Networks; CENELEC: Brussels, Belgium, 2022.

4. Yu, W.; Liang, F.; He, X.; Hatcher, W.G.; Lu, C.; Lin, J.; Yang, X. A Survey on the Edge Computing for the Internet of Things. IEEE Access; 2018; 6, pp. 6900-6919. [DOI: https://dx.doi.org/10.1109/ACCESS.2017.2778504]

5. Sadabadi, M.S.; Meng, X.; Liu, Z. Resilient and Robust Voltage Regulation in Shipboard DC Microgrids With ZIP Loads Under Actuator and Parameter Uncertainties. IEEE Trans. Transp. Electrif.; 2026; 12, pp. 578-588. [DOI: https://dx.doi.org/10.1109/TTE.2025.3619129]

6. Wang, H.; Song, Y.; Yang, H.; Liu, Z. Generalized Koopman Neural Operator for Data-Driven Modeling of Electric Railway Pantograph–Catenary Systems. IEEE Trans. Transp. Electrif.; 2025; 11, pp. 14100-14112. [DOI: https://dx.doi.org/10.1109/TTE.2025.3609347]

7.IEC 61727 Photovoltaic (PV) Systems—Characteristics of the Utility Interface; International Electrotechnical Commission: Geneva, Switzerland, 2004.

8. Shafiullah, G.; Arif, M.T.; Oo, A.M. Mitigation strategies to minimize potential technical challenges of renewable energy integration. Sustain. Energy Technol. Assess.; 2018; 25, pp. 24-42. [DOI: https://dx.doi.org/10.1016/j.seta.2017.10.008]

9. Zhu, Y.; Liu, C.; Dai, R.; Liu, G.; Xu, Y. Optimal Battery Energy Storage Placement for Transient Voltage Stability Enhancement. Proceedings of the 2019 IEEE Power & Energy Society General Meeting (PESGM); Atlanta, GA, USA, 4–8 August 2019; pp. 1-5. [DOI: https://dx.doi.org/10.1109/PESGM40551.2019.8973610]

10. Massucco, S.; Pongiglione, P.; Silvestro, F.; Paolone, M.; Sossan, F. Siting and Sizing of Energy Storage Systems: Towards a Unified Approach for Transmission and Distribution System Operators for Reserve Provision and Grid Support. Electr. Power Syst. Res.; 2021; 190, 106660. [DOI: https://dx.doi.org/10.1016/j.epsr.2020.106660]

11. Gomila, D.; Carreras, B.A.; Reynolds-Barredo, J.M.; Martinez-Barbeito, M.; Colet, P.; Gomis-Bellmunt, O. Reducing Blackout Risk by Segmenting European Power Grid with HVDC Lines. J. Mod. Power Syst. Clean Energy; 2025; 13, pp. 1556-1567. [DOI: https://dx.doi.org/10.35833/MPCE.2024.000768]

12. Ferkous, K.; Guermoui, M.; Menakh, S.; Bellaour, A.; Boulmaiz, T. A Novel Learning Approach for Short-Term Photovoltaic Power Forecasting—A Review and Case Studies. Eng. Appl. Artif. Intell.; 2024; 133, 108502. [DOI: https://dx.doi.org/10.1016/j.engappai.2024.108502]

13. Marcu, M.D.; Popescu, F.G.; Pana, L. Modeling and simulation of power active filter for reducing harmonic pollution using the instantaneous reactive power theory. Environ. Eng. Manag. J.; 2014; 13, pp. 1377-1382. Available online: https://eemj.eu/index.php/EEMJ/article/view/1951 (accessed on 31 January 2026). [DOI: https://dx.doi.org/10.30638/eemj.2014.148]

14. Rosso, R.; Wang, X.; Liserre, M.; Lu, X.; Engelken, S. Grid-forming converters: Control approaches, grid-synchronization, and future trends—A review. IEEE Open J. Ind. Appl.; 2021; 2, pp. 93-109. [DOI: https://dx.doi.org/10.1109/OJIA.2021.3074028]

15. Ma, Q.; Wei, W.; Mei, S. Health-aware coordinate long-term and short-term operation for BESS in energy and frequency regulation markets. Appl. Energy; 2024; 356, 122363. [DOI: https://dx.doi.org/10.1016/j.apenergy.2023.122363]

16. Joshi, P.; Hasanuzzaman, M.; Thapa, C.; Afli, H.; Scully, T. Enabling all in-edge deep learning: A literature review. IEEE Access; 2023; 11, pp. 3431-3460. [DOI: https://dx.doi.org/10.1109/ACCESS.2023.3234761]

17. Zhang, X.; Nie, J.; Huang, Y.; Xie, G.; Xiong, Z.; Liu, J.; Niyato, D.; Shen, X.S. Beyond the cloud: Edge inference for generative large language models in wireless networks. IEEE Trans. Wirel. Commun.; 2024; 24, pp. 643-658. [DOI: https://dx.doi.org/10.1109/TWC.2024.3497923]

18. Shi, W.; Cao, J.; Zhang, Q.; Li, Y.; Xu, L. Edge computing: Vision and challenges. IEEE Internet Things J.; 2016; 3, pp. 637-646. [DOI: https://dx.doi.org/10.1109/JIOT.2016.2579198]

19. Huang, W.; Deng, X. Real-time tracking railway intruders using multiple-agent cooperated large language models with edge stream processing engine. J. Netw. Comput. Appl.; 2025; 242, 104231. [DOI: https://dx.doi.org/10.1016/j.jnca.2025.104231]

20. Yıldırım, F.; Yalman, Y.; Bayındır, K.Ç.; Terciyanlı, E. Comprehensive Review of Edge Computing for Power Systems: State of the Art, Architecture, and Applications. Appl. Sci.; 2025; 15, 4592. [DOI: https://dx.doi.org/10.3390/app15084592]

21. Molokomme, D.N.; Onumanyi, A.J.; Abu-Mahfouz, A.M. Edge Intelligence in Smart Grids: A Survey on Architectures, Offloading Models, Cyber Security Measures, and Challenges. J. Sens. Actuator Netw.; 2022; 11, 47. [DOI: https://dx.doi.org/10.3390/jsan11030047]

22. Minh, Q.N.; Nguyen, V.-H.; Quy, V.K.; Ngoc, L.A.; Chehri, A.; Jeon, G. Edge Computing for IoT-Enabled Smart Grid: The Future of Energy. Energies; 2022; 15, 6140. [DOI: https://dx.doi.org/10.3390/en15176140]

23. Poyyamozhi, M.; Murugesan, B.; Rajamanickam, N.; Shorfuzzaman, M.; Aboelmagd, Y. IoT—A Promising Solution to Energy Management in Smart Buildings: A Systematic Review, Applications, Barriers, and Future Scope. Buildings; 2024; 14, 3446. [DOI: https://dx.doi.org/10.3390/buildings14113446]

24. Cavalieri, S.; Cantali, G.; Susinna, A. Integration of IoT Technologies into the Smart Grid. Sensors; 2022; 22, 2475. [DOI: https://dx.doi.org/10.3390/s22072475] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35408089]

25. Ghasempour, A. Internet of Things in Smart Grid: Architecture, Applications, Services, Key Technologies, and Challenges. Inventions; 2019; 4, 22. [DOI: https://dx.doi.org/10.3390/inventions4010022]

26. Ameedeen, M.A.; Kamarudin, I.E.; Ab Razak, M.F.; Zabidi, A. Integrating Edge Computing and Software Defined Networking in Internet of Things: A Systematic Review. Iraqi J. Comput. Sci. Math.; 2023; 4, 11. [DOI: https://dx.doi.org/10.52866/ijcsm.2023.04.04.011]

27. Kirmani, S.; Mazid, A.; Khan, I.A.; Abid, M. A Survey on IoT-Enabled Smart Grids: Technologies, Architectures, Applications, and Challenges. Sustainability; 2023; 15, 717. [DOI: https://dx.doi.org/10.3390/su15010717]

28. Gao, Y.; Li, S.; Yan, X. Assessing Voltage Stability in Distribution Networks: A Methodology Considering Correlation among Stochastic Variables. Appl. Sci.; 2024; 14, 6455. [DOI: https://dx.doi.org/10.3390/app14156455]

29. Ghaeb, J.; Salah, S.; Obeidat, F. Intelligent Control for Voltage Regulation in the Distribution Network Equipped with PV Farm. Energies; 2023; 16, 360. [DOI: https://dx.doi.org/10.3390/en16010360]

30. Andras, A.; Popescu, F.D.; Radu, S.M.; Pasculescu, D.; Brinas, I.; Radu, M.A.; Peagu, D. Numerical Simulation and Modeling of Mechano–Electro–Thermal Behavior of Electrical Contact Using COMSOL Multiphysics. Appl. Sci.; 2024; 14, 4026. [DOI: https://dx.doi.org/10.3390/app14104026]

31. Valuva, C.; Chinnamuthu, S.; Khurshaid, T.; Kim, K.-C. A Comprehensive Review on the Modelling and Significance of Stability Indices in Power System Instability Problems. Energies; 2023; 16, 6718. [DOI: https://dx.doi.org/10.3390/en16186718]

32. Meegahapola, L.; Sguarezi, A.; Bryant, J.S.; Gu, M.; Conde D., E.R.; Cunha, R.B.A. Power System Stability with Power-Electronic Converter Interfaced Renewable Power Generation: Present Issues and Future Trends. Energies; 2020; 13, 3441. [DOI: https://dx.doi.org/10.3390/en13133441]

33. Cárdenas, P.A.; Martínez, M.; Molina, M.G.; Mercado, P.E. Development of Control Techniques for AC Microgrids: A Critical Assessment. Sustainability; 2023; 15, 15195. [DOI: https://dx.doi.org/10.3390/su152115195]

34. Pasculescu, D.; Riurean, S.; Ilieva-Obretenova, M.; Lazar, T.; Tatar, A.M.; Fita, N.D. Intelligent Modeling of PV–BESS Microgrids for Enhanced Stability, Cyber–Physical Resilience and Blackout Prevention. Energies; 2026; 19, 148. [DOI: https://dx.doi.org/10.3390/en19010148]

35. He, C.; Geng, H.; Rajashekara, K.; Chandra, A. Analysis and control of frequency stability in low-inertia power systems: A review. IEEE/CAA J. Autom. Sin.; 2024; 11, pp. 2363-2383. [DOI: https://dx.doi.org/10.1109/JAS.2024.125013]

36. Radaelli, L.; Martinez, S. Frequency Stability Analysis of a Low Inertia Power System with Interactions among Power Electronics Interfaced Generators with Frequency Response Capabilities. Appl. Sci.; 2022; 12, 11126. [DOI: https://dx.doi.org/10.3390/app122111126]

37. Singh, B.; Slowik, A.; Bishnoi, S.K.; Sharma, M. Frequency Regulation Strategy of Two-Area Microgrid System with Electric Vehicle Support Using Novel Fuzzy-Based Dual-Stage Controller and Modified Dragonfly Algorithm. Energies; 2023; 16, 3407. [DOI: https://dx.doi.org/10.3390/en16083407]

38. Wang, Y.; Liu, C.; Liu, Z.; Wang, T.; Ke, F.; Yang, D.; Zhang, D.; Miao, S. A Hierarchical Cooperative Frequency Regulation Control Strategy of Wind-Storage-Load in a Microgrid Based on Model Prediction. Energies; 2023; 16, 1886. [DOI: https://dx.doi.org/10.3390/en16041886]

39. Fan, Y.; Yang, Y.; Wu, F.; Qiu, H.; Ye, P.; Xu, W.; Zhong, Y.; Zhang, L.; Chen, Y. Coordinated reactive power–voltage control in distribution networks with high-penetration photovoltaic systems using adaptive feature mode decomposition. Energies; 2025; 18, 2866. [DOI: https://dx.doi.org/10.3390/en18112866]

40. Pana, L.; Grabara, J.; Moraru, R.I.; Pasculescu, D.; Pasculescu, V.M. Optimal Quality Management Algorithm for Assessing the Usage Capacity Level of Mining Transformers. Pol. J. Manag. Stud.; 2018; 18, pp. 233-244. [DOI: https://dx.doi.org/10.17512/pjms.2018.18.2.19]

41. CIGRÉ Task Force C6.04.02. Benchmark Systems for Network Integration of Renewable and Distributed Energy Resources; Technical Brochure, Update 2020/2021 CIGRE: Paris, France, 2014.

42. Al-Ali, A.; Aburukba, R. Role of Internet of Things in the Smart Grid Technology. J. Comput. Commun.; 2015; 3, pp. 229-233. [DOI: https://dx.doi.org/10.4236/jcc.2015.35029]

43. Amin, M.R.; Negnevitsky, M.; Franklin, E.; Alam, K.S.; Naderi, S.B. Application of Battery Energy Storage Systems for Primary Frequency Control in Power Systems with High Renewable Energy Penetration. Energies; 2021; 14, 1379. [DOI: https://dx.doi.org/10.3390/en14051379]

44. Raoufi, H.; Vahidinasab, V.; Mehran, K. Power Systems Resilience Metrics: A Comprehensive Review of Challenges and Outlook. Sustainability; 2020; 12, 9698. [DOI: https://dx.doi.org/10.3390/su12229698]

45. He, J.-H.; Lin, J.-H. Review of Microgrids to Enhance Power System Resilience. Eng. Proc.; 2025; 92, 82. [DOI: https://dx.doi.org/10.3390/engproc2025092082]

46. Aghajan-Eshkevari, S.; Ameli, M.T.; Azad, S.; Pashiri, N.T. Chapter 15—Resilience enhancement of distribution networks by optimal scheduling of hydrogen systems. Future Modern Distribution Networks Resilience; Elsevier: Amsterdam, The Netherlands, 2024; pp. 353-374. [DOI: https://dx.doi.org/10.1016/B978-0-443-16086-8.00003-8]

47. Riurean, S.; Fîță, N.-D.; Păsculescu, D.; Slușariuc, R. Securing Photovoltaic Systems as Critical Infrastructure: A Multi-Layered Assessment of Risk, Safety, and Cybersecurity. Sustainability; 2025; 17, 4397. [DOI: https://dx.doi.org/10.3390/su17104397]

48. Zhou, D.; Pan, X.; Sun, X.; Hu, F. Resilience Assessment Framework for High-Penetration Renewable Energy Power System. Sustainability; 2025; 17, 2058. [DOI: https://dx.doi.org/10.3390/su17052058]

49. Padmavathy, R.; Singh, S.K.; Sindhu, M.; Jasim, L.H. Archana Saxena and Sukhvinder Singh Dari, Enhancing Power Grid Resilience against Cyber Threats in the Smart Grid Era. E3S Web Conf.; 2024; 540, 10021. [DOI: https://dx.doi.org/10.1051/e3sconf/202454010021]

50. Yang, H.; Liu, Z.; Cui, H.; Ma, N.; Wang, H.; Zhang, C.; Song, Y. An Electrified Railway Catenary Component Anomaly Detection Frame Based on Invariant Normal Region Prototype With Segment Anything Model. IEEE Trans. Transp. Electrif.; 2026; 12, pp. 1391-1402. [DOI: https://dx.doi.org/10.1109/TTE.2025.3628607]

Word count: 24022

Show less

© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license. Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.

Abstract

Translate

The transition toward active distribution networks requires advanced control solutions capable of handling the rapid dynamics of distributed energy resources. This paper proposes a low-cost, intelligent IoT architecture designed for the real-time optimization and analysis of energy systems within low-voltage networks. Unlike centralized monitoring approaches constrained by communication latency, the proposed solution leverages Intelligent Edge Processing (IEP) implemented on ESP32 embedded nodes to optimize data flow and decision-making. This architecture executes stability assessments directly at the network edge, calculating critical analysis indicators such as the Voltage Deviation Index (VDI) and Rate of Change of Frequency (RoCoF). The system was validated on the CIGRE European LV benchmark under severe stress scenarios, including rapid solar transients and voltage sags. The results demonstrate that the proposed architecture effectively coordinates storage interventions, ensuring voltage recovery within 300 ms and maintaining power quality within EN 50160 limits even during severe voltage sags. The study validates the feasibility of using industrial IoT edge computing as a resilient, non-wire alternative for modernizing complex energy systems.

Details

Title

Real-Time Energy System Optimization and Resilience Analysis in Low-Voltage Networks Using Intelligent Edge Computing

Author

Lazar, Dan Cristian¹; Petrilean, Dan Codrut²

; Lazar Teodora¹; Popescu, Florin Gabriel¹; Ionescu Daria³; Tatar, Adina Milena⁴

; Buica Georgeta⁵; Pasculescu Dragos¹

¹ Automation, Computer, Electrical and Power Department, University of Petrosani, 332006 Petrosani, Romania; [email protected] (D.C.L.); [email protected] (T.L.); [email protected] (F.G.P.)
² Department of Mechanical, Industrial and Transportation Engineering, University of Petrosani, 332006 Petrosani, Romania; [email protected]
³ Industrial Engineering and Management Department, University of Petrosani, 332006 Petrosani, Romania; [email protected]
⁴ Department of Industrial and Automation Engineering, University “Constantin Brancusi” of Targu-Jiu, 210135 Targu-Jiu, Romania; [email protected]
⁵ Electrical and Mechanical Risks Laboratory, INCDPM “Alexandru Darabont”, 35A Blvd Ghencea, 6th County, 062692 Bucharest, Romania; [email protected]

First page

660

Publication year

2026

Publication date

2026

Publisher

MDPI AG

e-ISSN

22279717

Source type

Scholarly Journal

Language of publication

English

DOI

https://doi.org/10.3390/pr14040660

ProQuest document ID

3307533044

Real-Time Energy System Optimization and Resilience Analysis in Low-Voltage Networks Using Intelligent Edge Computing

Jump to:

Full text

1. Introduction

1.1. The Context of the Energy Transition and the Role of PV–BESS Microgrids

1.2. Specific Challenges in Monitoring Low-Voltage PV–BESS Microgrids

1.3. The Role of IoT Systems and Intelligent Edge Processing

1.4. Scope and Contributions of the Paper

1.5. Organization of the Paper

2. System Modeling

2.1. The CIGRÉ LV Network—Structure, Parameters and Topology

2.2. Photovoltaic (PV) Unit Model

2.2.1. Photovoltaic Generator

2.2.2. MPPT Algorithm and the Selection of the Incremental Conductance Method

2.2.3. Performance Parameters of the MPPT Algorithm

2.2.4. Relevance of the Photovoltaic System for IoT Monitoring

2.2.5. Photovoltaic (PV) Inverter Model

2.2.6. PV Power and Microgrid Operation Mode

2.3. Battery Energy Storage System (BESS) Model

2.3.1. General Architecture of the BESS

2.3.2. Dynamic Battery Model

2.3.3. BESS Control Strategy Within the Microgrid

2.3.4. The Role of BESS in Enhancing PV–BESS Microgrid Resilience

2.3.5. Relevance of BESS for IoT-Based Monitoring

2.4. Real Load and Generation Profiles (Load and PV Production Profiles)

3. Proposed IoT Architecture

3.1. Integrated IoT Sensors (Voltage, Frequency, P, Q, RoCoF)

3.2. Node-RED/MQTT/Modbus/ESP32—Communication and Edge Processing Infrastructure

3.3. Intelligent Edge Processing for Microgrid Stability Estimation

3.4. Complete Data Flow: “Measurement → Edge Processing → Cloud Logging”

4. Methodology and Stability Indicators

4.1. Voltage Deviation Index (VDI)

4.2. Resilience Index (RI)

4.3. RoCoF (Rate of Change of Frequency)

4.4. Test Scenarios (Faults, PV Variation, Overloads)

4.5. Synthesis of Resilience Indicators

5. Results and Discussions

5.1. Comparative Analysis of Voltage and Frequency Stability (VDI and RoCoF)

5.2. Interpretation of the Resilience Index (RI)

5.3. The Role of IoT Architecture and Edge Processing

5.4. Practical Implications for European Low-Voltage Networks

5.5. Limitations and Future Research Directions

6. Conclusions

Abstract

Details