Optimized fault detection and control for enhanced reliability and efficiency in DC microgrids

Abstract

This paper introduces a comprehensive framework for fault detection and control in DC microgrids (DCMGs) integrating diverse energy sources. A resistance-based fault detection scheme is proposed to address intermittent DC link faults, enabling efficient operation without complete system shutdown. Perturb and Observe (P&O) techniques are employed for PV and wind power tracking, while proportional-integral (PI) controllers manage fuel cell (FC) and battery energy storage systems (BESS). Fuzzy logic controllers (FLCs) demonstrate superior performance over traditional PI controllers in mitigating voltage and current (V-I) fluctuations. To optimize DC-link V-I levels, a genetic algorithm-tuned PI controller (GA-PIC) and evolution-inspired PI controller are utilized. The proposed method is validated using Opal-RT simulations under various scenarios, demonstrating improved performance over un-optimized configurations. The key achievement of this research is a validated, optimized control and protection scheme that significantly enhances the stability and reliability of DCMGs under fault conditions. Specifically, the work develops a distributed fault detection and control method to improve protection and address stability and power quality in DCMGs. It also presents a GA-based PI-optimized controller for DCMGs with FC and battery storage, and an optimized controller integrating FLCs and GA-tuned PI-Cs to reduce V-I fluctuations. Furthermore, an integrated DC protection scheme is implemented, demonstrating enhanced fault detection speed and accuracy compared to individual schemes. The effectiveness of the proposed GA-PI-C is validated through Opal RT real-time simulations, confirming the efficacy of FLCs in dynamic system responses and contributing to more robust and reliable DCMG operation.

Introduction

DC microgrids (DCMGs) are increasingly utilized due to their reliability, stability, and efficiency benefits within modern energy world¹. However, ensuring consistent electrification from sources to loads in DCMGs presents challenges due to intermittency². The DCMG represents a future power system integrating distributed generation (DG) sources (i.e., PV, wind energy (WE), fuel cell (FC), utility grid (UG)), loads (i.e., , , , and ), and two different battery energy storage systems (BESS)³. The integration of energy sources with storage systems is crucial for ensuring a reliable energy supply. Renewable energy sources are unpredictable and intermittent. At the same time, the demand for clean energy is rising to meet growing power needs. Installing PV and WE in DCMGs is relatively straightforward, and they operate without fuel consumption⁴. Similarly, fuel cells can provide stable power, addressing energy storage/supply challenges, and dynamically adjust their output voltage as needed^3,5. So battery and fuel cells are used as backup units. The progress in the power converter’s controller has made a good choice for energy storage solutions and the growth of power resources in DCMG¹. Although there are many advantages, it also provides challenges such as protection, stability, control, and power quality.

The DCMG is considered healthy through various protection schemes when the voltage and current (V-I) are at the source and loads terminal. DCMG distribution lines/cables may experience persistent issues with interruptions and power balancing, despite careful design and operational methods⁶. Following faults in the system, the voltage and current fluctuations on the DC link increase sharply (or spike), causing the entire system to de-energize^7,8. Therefore, in practical terms, a strong protection scheme is essential to maintain V-I within safe limits at the DC link. This ensures the continuous and reliable operation of DCMGs^6,8. Similarly, the system’s safety measures, accurate stability design and effective energy control methods are required⁹. Therefore, relays () and circuit breakers can be installed to quickly detach faulty sections. These defects are marked by relays and circuit breakers (CB). The fault is resolved quickly based on the transmitted fault information within a short period of operation^2,7. The statement regarding the distance (80 m to 100 m) between relay protection lines was incorrectly marked and has been revised for clarity^10,11. Identifying faults in microgrids presents a significant challenge. This is due to the presence of many sources and the deliberate blocking of the fault currents () by DC-DC converters^12–17. The DCMG system has short-circuit (SC) defects including line-to-line (LL), line-to-ground (LG) defects and arc. The current-based methods are used in DC systems, but voltage sag-based methods can cause CB tripping with incorrect localization¹⁰. This scheme detects and mitigates low and high impedance faults (LIF/HIF) in DC grids without requiring a full shutdown. The SC faults in DCMG systems highlight the need for an efficient protection relay algorithm for reliability. Although current approaches address these concerns, they still have limitations. So, the main objective is to design a reliable algorithm for fault detection¹¹. The biggest challenge in DCMG control is managing complex interactions between non-linearities and variable power output from distributed sources. The fault conditions further complicate system stability and reliability¹⁸. From a practical standpoint, improving fault detection and control in DCMGs is crucial. It ensure the safety and longevity of the system components. It prevents costly downtime, and maintain a stable power supply to critical loads.

The fault detection scheme uses conditional algorithms induced by abnormal events. It must be carefully selected for each DC line to ensure accurate fault detection. In², a high-speed fault detection and localization scheme is proposed, offering superior speed and accuracy over traditional methods. It effectively addresses the critical need for rapid and precise fault clearing. In⁷, the multi-terminal DC improves localization in microgrids using local measurements. It provides fault information, which is essential for accurately identifying and isolating the target fault. In⁸, a real-time resistance-based technique was developed, enabling rapid and reliable fault detection in zonal low-voltage DC Microgrids. This approach significantly enhances protection for these specific configurations. These studies emphasize the importance of fault identification schemes designed for different DC Microgrid topologies. In^12,16, medium-voltage DC Microgrids investigated deficiencies by coordinating supply power converters and bus contactors for improved protection. This highlights the necessity of comprehensive control and protection strategies for enhanced reliability. In¹³, analyzing fault characteristics across various converter topologies offers valuable insights for robust protection design. It highlights the significant impact of converter technology on abnormal system behavior. In⁶, the impedance-based relaying scheme for hybrid AC/DC microgrids introduces complexities in protecting these advanced systems. In¹⁵, proposes high-speed differential protection, while¹¹ studies inter-connection protection. The fault detection scheme’s performance is assessed through incorrect identity evaluation and protection rate analysis. A comparative study with existing technologies is necessary to ensure its effectiveness and reliability. Rather than identifying a fault, effective control strategies are required.

In¹⁷, the proportional load has developed distribution control for sharing and voltage control, improving energy management and stability¹. In the focus on controlling PV-WE-based microgrids with hybrid power storage. In⁴, the converter-lace solar PV control strategy. Optimization techniques are crucial to improve both incorrect identification and control performance. In¹⁸, the proposal of vague-oriented power-sharing methods for microgrids shows better performance compared to conventional drop control. In¹⁹, the PI-Cascade Controller has been studied for parameter optimization in linear servo mechanisms. It demonstrates the impact of different optimization algorithms on system performance. In²⁰, optimization-based PI controller tuning is implemented for non-deal differential boost inverters, improving the dynamic response and stability of the system. In²¹, Self-adaptive virtual inertial control using fuzzy logic was implemented to improve frequency stability in microgrids with high renewable penetration. It effectively addresses challenges arising from fluctuating renewable energy generation. In²², natural genetics adopted control for autonomous WE-battery-based microgrids, demonstrating its effectiveness in optimizing system performance and stability. In²³, a support vector machine integrated with a genetic algorithm was utilized for fault diagnosis in high-voltage circuit breakers. This approach highlights the effectiveness of machine learning techniques in predictive maintenance. These control strategies, including optimized and intelligent control methods, contribute to efficiency and stability.

The new protective relay algorithm, implemented in MATLAB, is used (or was implemented) to improve the real-time monitoring of power distribution systems. The algorithm consolidates protective measures such as over-voltage, under-voltage and over-current protection, which improves system reliability²⁴. The response to DC defects and the magnitude of the affected fault current are influenced by resistance considerations. Proper assessment of resistance is crucial for accurate fault analysis and mitigation. This is due to the maintenance of distribution buses interconnected by power converters. The reference current for maximum power extraction and air in PV is produced according to the P&O maximum power point tracking method. The current prevailing control procedures for BESS and FC have PI control¹⁹. This reference value is maintained by controlling the boost converter input current. The study focuses on the V-I interruption that responds by optimizing PI control associated with FC and BESS to reduce DC link V-I fluctuations²⁰. The dimming suspicion system is proposed for the use of the system compatible and effective. The proposed research aims to fill this space by consolidating Adaptive FL-C based on automatic shift and sequence compensation methods¹⁸. The use of different control topologies and converters reduces the efficiency of the parallel operation. Improvements in coordinating Distributed Energy Sources (DES), BESS, and loads are needed for proper power supply control at the DCMG link. These adjustments affect the stability of the PI controller in DCMG^25–32. It can be resolved through a single control loop that effectively solves the energy part. The membership function of MPPT resolves V-I variations using the FLC trial-and-error tracking method. It effectively enhances system performance by addressing resistance-based shortcomings. Thus FLC effectively maintains V-I levels due to its non-linear and dynamic behavior.

The consolidation of AI-based control methods, including ANFIS, AO, EO, SA, and Fuzzy, enhances system efficiency. Among these, FLC demonstrates superior performance in improving energy quality and stability in systems. The PI-C and FLC-based controller implementation work together to reduce voltage and current distortions. The GA-based tuned PI controller method is implemented due to the consistent expected performance level. This remains unchanged even when using different PI and FL controller types. This comprehensive approach aims to improve reliability and resiliency by reducing deviations. The ultimate goal is to develop an integrated fault detection and control strategy for DCMG. The proposed DCMG protection system uses GA to optimize PI-C parameters by providing an efficient control design^19,20. This approach allows for precise adjustment of the PI controller’s settings, leading to the best possible overall performance of the system²². This approach builds on successful solutions from previous attempts, leading to continual improvement. This approach also minimizes the chance of getting stuck on a solution that isn’t quite the absolute best ( “global optimum”). This leads to more reliable and effective results³³. It maintains V-I control for both stable and changing conditions. This method refines the controller’s settings to achieve the best possible performance within set limitations. Efficient collective search also provides parallelization, resilience, and computational efficiency²³. It provides a valuable approach to obtaining optimal parameters for PI-C in DCMG^34,35. This research introduces a design methodology for parameter optimization of PI-C using GA. Finally, the proposed protection system demonstrates improved effectiveness in enhancing microgrid performance during fault conditions. The methodology is assessed using MATLAB simulations and validated with the OPAL-RT real-time simulator. It analyzes transient and steady-state responses to short-circuit faults effectively. The proposed method’s practicality and accuracy are demonstrated by comparing un-optimized and optimized configurations. Simulink and OPAL-RT simulations validate performance improvements with FLCs in dynamic system responses.

The following are the main contributions of the paper:

• i.This approach introduces a distributed fault detection and control method. It aims to improve protection in DCMGs while addressing stability and power quality issues.
• ii.Proposing a PI-optimized controller based on a genetic algorithm for DCMG with FC and battery storage.
• iii.Implements an optimized controller that integrates FLC and GA-based tuned PI-Cs, leading to significant reductions in voltage and current fluctuations in the DCMG.
• iv.This article describes an integrated DC protection scheme that performs better than individual schemes and enables faster fault detection in DCMGs. The proposed GA-PI-C provides excellent performance under various loads and supply voltages.
• v.Validation through Opal RT real-time simulator to evaluate fault response and system performance.

The remainder of the paper is structured as follows: Section II provides an in-depth review of existing fault detection and control techniques in DCMGs, highlighting their limitations. Section III describes the proposed fault detection methodology, including the resistance-based fault detection scheme and optimization techniques. Section IV presents the control strategy, detailing the implementation of FLCs, GA-PI-Cs, and energy management techniques. Section V discusses the simulation setup, experimental validation using OPAL-RT, and comparative performance analysis. Section VI concludes the study by summarizing the key findings and discussing future research directions in DCMG fault detection and control.

Description of the DC micro-grid system

A DCMG consists of distributed energy sources (i.e. PV, WE, FC, and UG), four resistive loads, and two different types of EV batteries^2,14. The structure of the DCMG model with different terminal types is shown in Fig. 1 and specification parameters are given in Table 1. In Fig. 1, solar PV and WE are connected to DCMG through DC-DC converters operating in P&O-based maximum power point tracking (MPPT) control mode⁶. Since the output generated by the WE system is AC in nature, a diode bridge rectifier (DBR) operating in constant power control mode is used to interface with the DCMG. The utility grid is connected to the DCMG through a DBR and the fuel cell is connected to the DC-DC boost converter.

Fig. 1.

Schematic diagram of DCMG consists of different sources, loads, converters, and CBs (-).

Table 1.

DCMG components and parameters.

Components	Parameters
Switching Frequency ()	5000 Hz
Desired terminal output voltage ()	120 V
ypical input voltage of the converter ()	200–400 V
Converter efficiency	85%
Load resistance (, , and )	25
The inductor value of the converter (L)	7.14 mH
The input capacitor ()	473.4693
Output capacitor ()	473.4693
Resistance Fault range (, ,)	0.1–10
,, , and	1.3 kW
BESS voltage	12 V
BESS capacity	35 Ah
PV, WE, FC, Grid DC-DC Converter	20 kW
	0.62 kW
voltage	24 V,12 V
capacity	35 Ah
Resistances of the cable () Cable length	12.1 90 m

The and are connected to the DCMG as a bi-directional converter voltage regulator via DC-DC converters. Generally, converters connected to the DC bus require capacitors to connect to the output. In this configuration, all types of energy sources, loads, and energy storage systems are connected to converters through DC wires across a distance of 82 to 100 m¹⁶. A DCMG system typically incorporates several DC - and relays (-). These components are strategically positioned along the DC line to detect and isolate fault segments promptly by issuing trip signals at various branches. This architecture allows the DCMG to operate in different modes, typically used in residential applications with EV stations. FL controllers are recommended to manage V-I fluctuations in DCMG links. It performs better than PI controllers in DES and storage systems converters, ensuring stability and improved operational efficiency.

External support for the DC link in an MG

In DCMG, DC-Link receives stable support from sources and BESS via a boost converter (BC). Additionally, the bi-directional buck-boost converter (Bi-Di) ensures voltage stability during source and load changes, as shown in Fig. 1. The DC link power balance of DCMG is defined by Eq. (1)⁴.

1

Where, , , , , represents the power of PV, WE, FC, UG, battery energy storage systems, and resistive loads output in a DC sub-grid. signifies the DC-Link power demand.

DC sub-grid with solar PV energy

Solar energy is preferred because it is free, clean, and environmentally friendly. The PV cells are connected in parallel and series to convert electrical power. The mathematical model for expressing the performance of PV is given by⁴,

2

Where, is the current generated by the PV cells (A), signifies the reverse saturation current (A), and is the voltage at an output of PV. And also, is an output current of the PV system, is the thermal voltage (V), implies the series resistance (Ω) and refers to the parallel resistance (Ω).

In this article, solar energy connects to the DC link for several reasons:

• Grid enhances power quality.

• Reduces pressure on the converter’s rated voltage and demand on the supply side.

• Protect the load from network interruptions.

The solar energy module can be further expanded with a boost converter in series to produce DC power. The SP allows real-time energy exchange during load changes and DC-Link capacitor charging. The MPPT controller is crucial for optimizing solar power extraction. Figure 2 shows the Solar controller with MPPT, and BC for voltage boosting. The SP output power, , is calculated using Eq. (2). Furthermore, the performance improvement achieved by using the fuzzy system is analyzed. In this way, the power with a fuzzy system is calculated using Eq. (3)³⁶.

3

Fig. 2.

Solar PV with Boost Converter (a) MPPT Algorithm (b) Fuzzy logic controller.

The difference between the present solar PV voltage and the previous solar PV voltage is given to a fuzzy logic controller in Eq. (4)³⁷.

4

DC sub-grid with wind energy

A wind turbine uses a Permanent Magnet Synchronous Generator (PMSG) to convert the WE into electricity. The PMSG’s electrical output is converted to direct current (DC) and amplified by a boost converter. The wind turbine of a power () determined by¹

5

Where Air density (ρ), rotor blade radius (R), wind speed (), and power coefficient (). This coefficient evaluates turbine and blade efficiency based on the tip speed ratio ) and pitch angle ().

The duty cycle of the boost converter is continuously adjusted to extract the most power from a WE Conversion System (WECS). As mentioned above in Solar PV systems the performance of the MPPT controller is crucial to optimize wind power extraction. Figure 3 shows the wind controller with MPPT, and BC for voltage boosting. The wind output power, , is calculated using Eq. (5). The improvement of the energy source is also compared with the fuzzy system³⁶. This way, the power with a fuzzy system is calculated using Eq. (7).

6

Fig. 3.

Wind energy with Boost Converter (a) MPPT Algorithm (b) Fuzzy logic controller.

For fuzzy Eq. 

7

The difference between the present WE current error and the previous WE current error is given to a fuzzy logic controller in Eq. (8)³⁷.

8

The controller manages the generator’s output by adjusting the current flowing through the boost converter’s inductor.

DC sub-grid with fuel cell energy

With the rise of renewable energy sources like hydrogen, there’s growing interest in its potential as a clean energy carrier for the future. FCs emerge as a particularly attractive option due to their efficiency. High-temperature FCs offer a distinct advantage: the ability to directly handle fuels like natural gas internally. It simplifies the power generation process and contributes to the reliability of FCs for stationary applications. However, a drawback of high-temperature FCs is the difficulty in quickly stopping them, limiting their use to stationary power generation. An additional benefit of FC systems is the ability to recapture excess water vapour and utilize it in steam turbines for further electricity generation, boosting overall system efficiency. Within the FC itself, oxygen is supplied to the cathode while hydrogen enters the anode. This separation leads to reduction occurring at the cathode and oxidation at the anode. Overall, fuel cell technology presents a promising path forward for direct power generation and applications in microgrids. As a highly efficient converter of chemical energy into electricity, FCs offer a reliable and clean solution for the future of power generation.

Figure 4 shows a constant DC bus voltage via PI-C with fuel cell and BC. The fuel cell output power, , is calculated using Eq. (9)³⁶.

9

Fig. 4.

Fuel cell with Boost Converter (a) PI controller (b) Fuzzy logic controller.

The difference between the fuel cell voltage () and the reference voltage () is defined by Eq. (10), which is supplied to a PI controller^3,5.

10

The fuel cell reference current is calculated by using Eq. (11).

11

Similarly, when controlling BC with fuzzy logic it is calculated by Eqs. (12)-(13). The relationship between a fuel cell’s ideal voltage () and its actual operating voltage () is described by a specific Eq. (12) supplied to a PI controller.

12

The difference between the present error voltage of fuel cell energy and the previous error voltage of fuel cell energy is given to a fuzzy logic controller in Eq. (13)³⁷.

13

DC sub-grid with battery energy storage system (BESS)

Figure 5 shows a constant DC bus voltage via PI-C with ES battery and Bi-BBC. The state of charge (SOC) performance of the battery can be significantly enhanced by Eq. 14^4,38.

14

Fig. 5.

Battery storage with Bi-Directional Buck Boost Converter (a) PI controller (b) Fuzzy logic.

Equation (15) defines the SOC constraints in solar charging or discharging.

15

The calculation of involves subtracting , while is determined through a PI-C as expressed in Eq. (16). In this context, represents the variation between and as specified in Eq. (16)²².

16

Where

For fuzzy Eqs^36,37.

17

controller.

A protection algorithm for enhanced fault detection

DC microgrids often utilize voltage source converters (VSCs) to connect loads and distributed generators to the DC bus via a common point of connection (PCC) as in Fig. 6. Traditional fault detection methods designed for grid-connected systems may not be suitable for evaluating the performance of short-circuit (SC) fault protection in this configuration^8,12,14,39. As illustrated in Fig. 1, a fault near a specific load (e.g., load ) can trigger existing protection schemes to isolate the entire section (i.e., all loads), disconnecting all loads connected through that PCC. This over-isolation is inefficient. Implementing these methods individually for each load branch increases both cost and system complexity. This work aims to develop a method for fault detection, location, and isolation within individual load branches, thereby minimizing protection system cost and complexity. Figure 1 depicts a DC microgrid powered by a fuel cell, wind turbine, and battery, each connected via DC-DC converters. Loads are grouped into zones based on proximity, effectively making each zone a local load for its respective converter. These converters dynamically adjust power delivery to meet fluctuating load demands. Converter-I (PV-side boost converter) typically operates in maximum power point tracking (MPPT) mode, while Converter-II (battery-side bidirectional converter) regulates the DC bus voltage. A fault drastically alters converter operating points. Failure to detect these changes can lead to bus voltage collapse and system-wide shutdown, potentially causing damage or even fire depending on the fault’s nature.

Fig. 6.

Equivalent circuit of the DCMG an L-L fault in the load zone.

This study assumes negligible cable inductance contribution to equivalent impedance, as the inductive component is dominant (i.e. ). The cable resistances between loads and points of common coupling (PCC) A and B are = - , while the resistance between converters is . The Distributed Converter Microgrid (DCMG) fault analysis uses an equivalent circuit with four loads ( - ) connected to the DC grid via cable resistances ( - ). Loads , , , and correspond to capacitors C₄ and C₅, respectively. represents the resistance line between source and load converters. A line-to-line (L2L) fault with resistance is introduced near load (zone 1). During the fault, bus voltages and drop rapidly, de-energizing the system. Simultaneously, capacitors ( to ) discharge quickly, primarily feeding the fault. Distributed Energy Resources (DER) and Bidirectional (BI-DI) converters are active sources connected to the converter output.

The capacitor discharge behaviour is key to fault detection. KCL is applied at nodes A and B to determine each source’s contribution, as defined in equations (18). The relationship between the resistance seen by the capacitor and its average current is generalized in Eq. (19). A 120 V initial voltage across a 473.4692µF capacitor, storing 0.7 J of energy, is examined under varying equivalent resistances () to cover both high and low values. The capacitor’s discharge rate is influenced by the fault resistance (), with lower leading to faster discharge. As the fault is assumed to be in the load zone, the load’s contribution to is disregarded. The equivalent circuit representing this simplified system is depicted in Fig. 7. This model facilitates analysis of the capacitor’s behaviour during fault conditions, are shown in Fig. 11(a-c).

18

19

Fig. 7.

Simplified equivalent circuit of the DCMG.

, , , , and Are the output currents of the different converters. consists of various resistances like , and .

Fault detection approach

In a DC microgrid, a DC-link fault triggers a rapid response from filter capacitors (FCs). Utilizing this dynamic behaviour, fault detection is achieved by observing the sudden spike in capacitor current directed towards the fault. The energy stored in a capacitor is given by . For a constant voltage of 115 V across the capacitor, power calculations are similar. Despite the RC circuit, power is dissipated through a resistive path associated with the capacitor. The peak discharge current is dependent on . Expressed as , the discharge current represents the equivalent resistance of parallel loads, cables, and fault resistances. Therefore, the peak discharge current serves as an indicator of the fault type. Comparing current patterns under dynamic conditions is crucial for accurate fault detection. (k) and (k) are considered in this analysis.

Where is the cable (line) resistance, is the resistance of the different load zones (i.e., , , , and ), and is the fault resistance.

Figure 8 shows the proposed fault detection algorithm. Following setup, relays (Rs) collect capacitor currents (i_j=1 to 6). The algorithm samples, averages, and calculates error indices, as described below.

Fig. 8.

Flowchart of the proposed protection scheme.

Sampling capacitor currents ()

Analog-to-digital converters (ADCs) sample currents at instants (k) and (k-1). Equations (20) and (21) define present and previous average current patterns for the i^th capacitor. The sampling time difference is ​ and . Error equations for load zone 1 use capacitor current and average sample currents and ​(k − 1) for capacitor , as shown in Fig. 6.

20

21

It can be noticed here that, At the time of the fault.

Fault index ()

The fault index “” is calculated using (22), which is nothing but the change between the present and past samples

22

23

At the time of the fault, the initial voltage of the capacitor is , As the capacitor voltage is maintained at the same voltage, i.e., 115 V. Thus, (23) can be written as (24)

24

Fault detection and characterization

In this DCMG setup, six FCs necessitate six relays to monitor their currents. The fault detection process begins by assigning states to relays ( = 1 for “on,” = 0 for “off”), as shown in Fig. 5. Relay states are determined through sampling and communicated to circuit breakers (CBs). Sampled capacitor currents () undergo averaging, yielding present and past current models: (k) and (k − 1). An error index (δ) is then calculated using Eq. (22). The RC circuit’s time constant is inversely proportional to fault resistance and maximum discharge current. Consequently, complete capacitor discharge time is shorter for low-impedance faults (LIFs), exhibiting a higher peak magnitude compared to high-impedance faults (HIFs). Post-fault, i_j decays exponentially. Upon reaching steady state, fault isolation becomes challenging. A detection threshold (δ ) is implemented for the capacitor, enabling HIF isolation. Lower () and upper () characterization thresholds are defined, dependent on , which itself relies on . > 1Ω signifies HIF, while 0< <1Ω indicates LIF^8,39.

In a 100 W/50V DCMG, the maximum discharge current of capacitors and was analyzed using (19) for fault resistance = 0.1 to 10 Ω. The resulting currents ranged from approximately 80 A to 5 A for C₂ and 5 A to 1.5 A for C₃. Immediately post-fault, the calculated δ exceeded the 0.4 A threshold (δ). A fault is identified when δij values surpass δ. Analysis suggests a minimum fault characterization threshold (δ) of 5 A for effective fault/disturbance differentiation. However, using 3 A and 40 A for the lower () and upper () thresholds, respectively, led to mischaracterization. A “no-fault” condition is registered below . Exceeding indicates a low-impedance fault (LF), while high-impedance faults (HF) are identified when δ falls between these thresholds. Equation (15) provides the misclassification index.

25

Working proposed scheme

This flowchart outlines a fault detection method designed to enhance the performance of a DC microgrid by rapidly identifying and classifying faults. The process begins with initializing a line status flag (S_j) to 1, indicating a healthy state. It then proceeds to measure voltage and current at sources and loads, alongside acquiring and averaging capacitor current samples . By comparing the measured current () against predefined thresholds (, ), the system preliminarily identifies potential fault conditions. A crucial step involves calculating the difference in the fault index (), derived from consecutive averaged current values, to ascertain the rate of current change. This change is then compared against other lines () to isolate the fault location⁴⁰. Based on the current magnitude and the rate of change, the system distinguishes between low and high-impedance faults, enabling tailored protection responses^41,42. Throughout the process, if no fault is detected or the condition resolves, the line status flag is reset to 1. This approach facilitates quick fault detection, differentiation between fault types, and precise localization, ultimately contributing to a more resilient and stable DC microgrid operation.

Imagine a scenario where an LG fault occurs in a DC microgrid. The current on the affected line will quickly rise. The sensors detect this change and send the data to the processing unit. The system calculates the fault index difference () and compares it with the predefined thresholds and values from other lines. If the rate of change is high and the current exceeds the high threshold, the system identifies a low-impedance fault. It then sends a trip signal to the circuit breakers protecting that line, isolating the faulty section from the rest of the microgrid. This rapid response prevents further damage and ensures the stability of the microgrid, are shown in Fig. 8.

Controller’s and PI-tuning for DCMG

The PV and WE systems employ the MPPT technique, while the FC and BESS systems utilize PI control. Similarly, the DCMG is linked with the utility grid to supply additional power. The control of adaptive PV and WE converters by the MPPT algorithm is shown in Fig. 2(a) and Fig. 3(a). The PI controller control of adaptive FC and BESS converters is shown in Fig. 4(a) and Fig. 5(a). When the BESS reaches maximum capacity, the local DC link voltage exceeds the reference value allowing the WE and PV to operate off-MPP. The impact of the DCMG fault causes changes in other sources and loads. Controlling the PI considering the sources and loads involved with converters is essential to ensure stable DC link voltage levels and safety. The PI controller is controlled to demonstrate how well FLC-based techniques can regulate the DC link voltage and current in the DCMG. Therefore, the focus is on control strategies for improving the voltage, and current of the converters through the optimization method. Thus, applying different controller techniques did not change the results as much as expected, So the GA-PI tuning technique was adopted. The hierarchical control system is recommended to ensure voltage stability during faults in DCMG.

Boost converters using MPPT algorithm

The boost converters effectively control the DC-link voltage by managing the current flow of PV and WE, enabling the MPPT²⁸. The MPPT algorithm is a software-based flowchart that uses electrical devices to optimize power generation, as shown in Fig. 9, where p(k) > p(k-1) represents the MPPT implementation. It is used in controlled DCMG connected to PV and WE through converters³⁰.

Fig. 9.

Flowchart of the proposed MPPT algorithm.

The symbol ∆D represents small incremental changes in the DC-link voltage. The control algorithm is illustrated in Fig. 8 where the effect of L, and We examine the converters for voltage conversion from source to load. A duty ratio-based MPPT technique with perturbation and observation (P&O) is used for PV and WE. MPPT is used to achieve zero derivatives of voltage for input control in current and voltage. The MPPT controller generates the reference voltage for DC-link operation.

Fuzzy logic controller

Fuzzy control ensures energy sources (PV, WE, FC) adapt to real-time loads by dynamically tracking MPP in converters. Because it is robust, it improves the accuracy of complex rules and is easy to design^18,36. Similarly, resize model knowledge of PV, FC, WE, and Bess is not required. Fuzzy logic control is made up of three essential components: fuzzification module, fuzzy inference engine, and defuzzification module, as shown in Fig. 10.

Fig. 10.

Structure of fuzzy logic controller.

Fuzzification

In the process of fuzzification, numerical values from the input are transformed into linguistic terms using a membership function, are shown in Fig. 11(a-c). For this MPPT system, the voltage and current of the photovoltaic (PV) module are measured continuously. Power is then calculated as the product of voltage and current (P = V * I). The control strategy is determined based on two key input control variables: error (E) and change in error (CE). Error (E) signifies the slope of the P-V characteristic curve at a specific sampling instance (k). Change in error (CE) represents the variation in error between consecutive sampling intervals³⁷.

26

Fig. 11.

(a) Membership function for error signal. (b) Membership function for change error signal. (c) Membership function for the duty Cycle (D).

In this context, P(k) represents the power output, and V(k) signifies the voltage. The controller regulates the output voltage in a DC-DC converter by adjusting the duty cycle (D). This adjustment is based on the error signal, E(k), representing the difference between the desired and actual voltage. The controller manipulates the duty cycle to drive the error signal towards zero, effectively guiding the converter towards the optimal operating point. In the fuzzy controller, the input values (E and CE) are transformed into variables like positive high (PH), positive medium (PM), positive small (PS), zero error (ZE), negative small (NS), negative medium (NM), and negative high (NH). These terms are defined using basic fuzzy sets. The membership functions for these fuzzy sets, for both the input and output variables, are shown in Fig. 11(a-c).

27

Inference engine

The fuzzy controller relies on fuzzification to convert crisp input values (E and CE) into fuzzy terms the system can understand. This is necessary before applying the rules in the inference engine, as shown in Fig. 12. Table 2 showcases the controller’s rule table. This table defines the fuzzy output (D) by combining fuzzy inputs (E and CE)³⁷. The fuzzy controller utilizes a rule table (Table 2) containing 49 rules. These rules govern the DC-DC converter’s operation to achieve MPP tracking for the PV, WE, FC, and BESS modules. The core principle behind these rules is to adjust the DC-DC converter’s duty ratio based on the deviation of the operating point from its MPP. The controller aims to drive towards its MPP by increasing or decreasing the duty ratio.

Fig. 12.

Surface view of fuzzy input versus output functions.

Table 2.

The fuzzy rules that are used in the proposed method.

E/CE	PH	PM	PS	ZE	NS	NM	NH
PH	ZE	ZE	ZE	NH	NH	NH	NH
PM	ZE	ZE	ZE	NH	NM	NM	NM
PS	ZE	ZE	ZE	NS	NS	NM	NM
ZE	NS	NS	ZE	ZE	ZE	PS	PS
NS	PM	PM	PS	NS	ZE	PS	ZE
NM	PM	PM	PM	PH	ZE	ZE	NS
NH	PH	PM	PM	PH	ZE	ZE	ZE

Defuzzification

For the DC-DC converter to function properly, it needs a precise control signal at its input. To achieve this, the fuzzy controller’s output, which is fuzzy information, must be converted into a specific numerical value. This process of transforming fuzzy data into a crisp (exact) value is called defuzzification. During defuzzification, the fuzzy logic controller’s output, represented by linguistic terms (“high” or “low”), is converted into a numerical value. There are two main methods for defuzzification: the Center of Area (COA) method and the Max Criterion Method. The COA method, which calculates the center of gravity of the final combined fuzzy set, is the most commonly used technique. The final fuzzy set is obtained by combining all outputs from the fuzzy rule set using the maximum membership function. The center of gravity () can then be calculated using the following formula³⁷:

28

and the actual duty ratio D is calculated by:

29

Proposed PI for converter’s

A PI-C is essential to adjust the variables of the proposed DCMG and tune the parameters with the corresponding transfer function. Unlike proportional components, these components exhibit variations depending on their respective transfer functions (TFs). A consistent gain term plays an important role in minimizing long-term errors^19,24,40. The transfer function of the PI-C is expressed by the following equation:

30

Here, indicates the output of the PI controller, and indicates the input signal. represents the proportional gain constant and represents the integral constant. PI-Cs improve the frequency results of distributed systems, thereby increasing regulation efficiency. Figure 12 shows the PI-C architecture. serves as an input to the PI differential controller, which produces an output based on the transfer function. As a result, the PI-C efficiently tunes the proposed controller^43,44. This block diagram illustrates the use of a weighted combination of ITAE as an objective function to derive PI parameters by a GA.

31

GA-based PI-tuning optimization

Figure 13 shows a simple block diagram of a PI-C, incorporating both proportional and integral elements. Optimize the PI gain to minimize the difference between the desired output (r) and the actual output (y) to achieve better performance. The central role is played by the control rule (u), e = r-y, and the important parameters and . Using integrated time absolute error (ITAE) as the error criterion optimizes PI setting values, as shown in Fig. 13. The control strategy aims to minimize the error between the set value and the actual value, thereby improving the system’s overall performance¹⁹. GA is a stochastic search technique that handles nonlinear systems and solves complex problems. It uses probabilistic transformation rules to manage potential solutions called individuals or chromosomes. Each generation evaluates individual performance using an objective function, assigning appropriate values ​​to each iteration. Using a survival of the fittest strategy based on chromosome fitness, the algorithm evolves iteratively, guided by the error value to evaluate fitness.

Fig. 13.

Structure of PI controller tuned based on optimization.

PI tuning using genetic algorithms

The GA starts with a small randomly generated population for fast optimization. The PI parameters, and , are encoded as binary strings (chromosomes) and converted to real-valued PI parameters to calculate the goodness of fit. These values ​​are evaluated in the PI controller, which determines the system response through cost functions such as the integration time absolute error (ITAE) as in Tables 3 and 4. Figure 14 shows the best and average fitness characteristics of GA. Steps 2 and 3 are repeated several times until optimal coverage is achieved. GA targets overall PI values ​​( and ) with minimal coverage across the entire operating range. Table 3 presents the main GA simulation parameters. The evolution and comparison of design parameter values of the GA as in Table 5. Figure 15 shows the comparative analysis of different PI parameters, highlighting GA’s superior performance over other techniques. The final selection of GA demonstrates its optimal parameter tuning, ensuring enhanced system efficiency and stability.

Table 3.

Parameters used in GA.

Parameter	Type/value
Maximum generations	10
Population size	50
Encoding	Binary
Selection	Uniform
Crossover	Single point crossover
Mutation	Uniform

Table 4.

PI parameter range.

PI parameters
Lower bound	0.001	0.001
Upper bound	200	200

Fig. 14.

Characteristic of Genetic algorithm.

Table 5.

Evaluation and comparison of design parameters.

Method	K p1	K i1	K p2	K i2
PI-C	1	1	1	1
Fuzzy	0.85	10	0.01	10
ANFIS	1.42	2.43	1	1
AO	2.473	4.718	1.486	3.037
EO	2.252	4.449	1.513	3.061
SA	66	251	70	56
GA	162.4255	109.4393	33.2547	189.0824

Fig. 15.

Comparison of different PI parameters.

Genetic Algorithm Steps:

Step 1: Initialization: GA parameters (crossover/mutation rates, population size, generations) are defined. A population of potential PI values is randomly generated as binary strings, each representing a solution.

Step 2: Fitness Evaluation: The Integral Time Absolute Error (ITAE) serves as the fitness function, aiming for minimization. Each individual’s fitness is calculated: Lower ITAE indicates better fitness.

ITAE = (31).

Step 3: Selection: Parent solutions are selected using a rank-based approach. Individuals with superior fitness (lower ITAE) are more likely to be chosen for reproduction.

Step 4: Crossover: Single-point crossover combines genetic material from selected parent pairs. A random crossover point is chosen, and genetic information is exchanged up to that point, creating offspring.

Step 5: Mutation: Random bit flips are introduced into offspring binary strings, preventing premature convergence.

Step 6: Iteration: The fitness of the updated population is evaluated. If the minimum ITAE criterion is met, the algorithm terminates, returning optimized PI values. Otherwise, steps 3–6 are repeated.

Result and discussion

A DCMG is simulated using MATLAB/Simulink with DES, a BESS, and load as shown in Fig. 1. The DESs are connected to boost converters, and the BESS are connected to buck-boost converters. All the power tracking controller techniques are also connected to converters. This study uses BESS as a backup system for the UG, WE, PV, and FC connections in DCMG. System implementation, efficiency, quality of required power, and energy management for the proposed system are managed by MATLAB Simulink of DCMG as shown in Fig. 1. To ensure the proposed algorithm’s stability, various situations, including fault and non-faulty transients, are addressed. It is expected that the proposed method will correctly respond to these interruptions by isolating the fault via tripping of the relevant breakers. DC cable faults (LG, LL, internal/external, arcing/static) are simulated to assess fault detection methods. For targeted performance evaluation, relays with the algorithm are labelled by their DC link connection, as shown in Figs. 8, 16.

Fig. 16.

Experimental setup of the proposed system in opal-RT.

The proposed algorithm’s findings for relays , , and installed on cables are explored further below. In Fig. 1, the faults , , and are simulated at = 1.0s, 1.4s, and 1.8s in the line connecting the UG, , , and DC connections. Figure 1 shows the voltage responses with relays , , and linked at both ends. The current will be high if the L2G fault has a low impedance. For a high-impedance fault, the current will be low. The impact of , , and circuit breaker operation on , , and relay performance evaluates the algorithm’s ability to deal with external disturbances. Figures 17, 18 and 19 shows the response of the relay , , and to variations in the rated value of voltage, current, and fault. Figure 17 presents a comparative analysis of voltage and current behavior in various distributed generation (DG) sources within a DCMG under fault conditions. Four distinct DG units solar PV, WE, FC, and a UG system-were examined. Each system was analyzed using PI, fuzzy, and GA based control methods. Three distinct remote faults (,, and ), each with a maximum severity of 0.05 ms, were introduced to assess system resilience. Figure 17(a) illustrates the voltage and current variations in the solar PV system. The introduced faults disrupt the terminal voltage and current, leading to fluctuations in the power balance at the PV terminal. Similarly, Fig. 17(b) depicts the impact of the same faults on the WE system. Here, the faults induce fluctuations not only in power balance but also in current intensity at the wind turbine terminal. Figure 17(c) shows the voltage and current variations in the Fuel Cell system. The faults again disrupt the terminal voltage and current, resulting in fluctuations in both power balance and the current level. Finally, Fig. 17(d) presents the results for the UG system. The faults cause fluctuations in power balance and a significant current level increase at the UG terminal. Notably, these faults, especially those occurring at the UG terminal, propagate through the DCMG, impacting the current levels and subsequently disrupting the operation of other connected source, load, and storage systems. This highlights the importance of robust control strategies for mitigating the impact of faults on the overall DCMG stability. Figure 18 illustrates the impact of faults on various load terminals within the DCMG. Four distinct loads (,, and ) were analyzed under fault conditions, with voltage and current measurements taken at each terminal. PI, fuzzy, and GA-based control methods were employed for analysis. Three remote faults (,, and ), each with a severity of up to 0.05 ms, were introduced. Figure 18(a) shows the voltage and current behavior at . All three faults disrupt the terminal voltage and current, leading to power balance fluctuations. Specifically, causes a significant increase in current and a corresponding decrease in voltage, potentially affecting other DCMG components. Figure 18(b) shows the response of . Here, the faults generally reduce both current and voltage levels, also impacting power balance. Figure 18(c) presents the results for . Again, the faults cause power balance fluctuations due to voltage and current disruptions. , in particular, induces correlated fluctuations in both voltage and current levels, with a voltage decrease associated with the current change. Finally, Fig. 18(d) shows the impact on . The faults disrupt the terminal voltage and current, leading to a decrease in power balance at this load. These results demonstrate the varying impact of faults on different load terminals within the DCMG, highlighting the need for effective fault management strategies. Figure 19 examines the impact of faults on two battery energy storage systems ( and ) within the DCMG. Voltage and current measurements were taken at each BESS terminal, and analysis was performed using PI, fuzzy, and GA-based control methods. Faults ,, and each with a severity of up to 0.05 ms, were introduced. Figure 19(a) shows the impact on . The faults disrupt both voltage and current, leading to a decrease in power balance. Furthermore, faults at other grid terminals also reduce the terminal voltage, causing significant current variations.

Fig. 17.

V-I comparative methods at the terminal faults of sources. (a) PV. (b) Wind. (c) Fuel cell. (d) Utility grid.

Fig. 18.

V-I comparative methods at the terminal faults of Resistive loads. (a) . (b) . (c) . (d) .

Fig. 19.

V-I comparative methods at the terminal faults of energy storage systems. (a) . (b) .

Figure 19(b) presents the results for . Similar to , the introduced faults disrupt voltage and current, reducing power balance. Again, faults at other grid locations decrease the terminal voltage, resulting in extreme current variations. A PI control strategy was employed to mitigate fault effects. PI controller parameters were optimized using manual tuning, fuzzy logic, and a GA. The FLCs demonstrated superior V-I regulation during fault events, achieving a 20% reduction in voltage overshoot and a 15% decrease in settling time compared to PI controllers. Opal-RT simulations validated FLC’s enhanced dynamic response, resulting in a 10% improvement in fault detection accuracy and a 5% reduction in steady-state error. Specifically, FLC-based systems minimized operational disruptions by reducing fault clearance time by 8%. The non-linear behavior of FLCs proved advantageous in handling rapid dynamic changes during faults, offering better robustness and stability compared to linear PI controllers. This highlights FLC’s effectiveness in enhancing DCMG resilience under fault conditions. The Genetic Algorithm (GA) optimizes PI controllers, surpassing both unoptimized PI and FLC methods in DCMG control. GA-tuned PI achieved a 12% reduction in V-I fluctuations and a 10% stability improvement over standard PI. Compared to FLC, GA-PI provided a 7% enhancement in dynamic response due to precise parameter optimization. GA’s global search avoids local optima, ensuring optimal controller settings. Opal-RT validation confirmed GA-PI’s robustness under varying loads, showing superior performance consistency. GA’s efficient parameter tuning minimizes steady-state error and improves fault clearance time, demonstrating its advantage in complex DCMG control. GA-based optimization significantly outperformed other methods, demonstrating its suitability for DCMG performance enhancement.

The suggested control is robust against short-circuit faults in DCMG. Short circuit fault responses are summarized as shown in Fig. 7. The DC bus is considered to have undervoltage, overvoltage, and overcurrent protection. In this situation, the variance in V-I after fault occurrence is minimized. By connecting the short-circuit fault line is triggered at = 1.0 s, 1.4s, 1.8s, and the fault is cleared at = 1.05s, =1.45s, and =1.85s. The power level of the DC link fluctuates due to a fault in the DCMGs, and the backup supply prevents it from falling. Figure 2 shows the impact of MPPT on PV current and power. Power from PV and FC are connected to ensure system stability. A PMSG wind system uses a set of equations (Eqs. 5–8) to create an optimal control approach that maximizes power output, as seen in Fig. 3. A consistent wind speed and a power coefficient (Cp) of 0.48 suggest functioning at the MPP. This maintains a constant 120 V DC connection voltage at the source. The battery is versatile, allowing for multi-directional charging and discharging while maintaining a constant 98 V DC connection voltage. To address potential load fluctuations, the voltage and current loops within the battery converter are managed by a typical Proportional-Integral (PI) controller. However, research indicates that an FLC strategy provides higher performance. FLC has faster settling times and less overrun during operation, making it a promising control option for this system. Even with continuous nonlinear constant changes, the FLC-based control approach has good reference tracking ability. Another major goal of controls is to sustain different load fluctuations. The GA-based tuned PI is used due to the low charging and discharging current at the BESS’s DC link with the FL controller. GA-based tuned PI controllers track steady state faster and perform better in PI controller response. Figure 13 shows the performance of a traditional PI controller using objective functions such as IAE, ITSE, and ISE. The impact of over/under load and significant voltage drop in a DCMG on the performance of the proposed GA-PIC algorithm is investigated. A DCMG with different DES and loads is modelled in the OPAL-RT 5600 simulator. The OPAL-RT 5600 generates DCMG output voltage and current, sent into the dSPACE DL350 as shown in Fig. 16. The proposed GA-PI control method is implemented on the dSPACE DL350. The obtained voltage and current outputs (Figs. 20, 21 and 22) are compared against results from the OPAL-RT simulation. An Opal-RT system validated the simulation results, ensuring accuracy and reliability. Figure 20 presents a comparative analysis of voltage and current in various distributed generation (DG) sources, focusing on results obtained using the GA method. Three remote faults (,, and ), each with a severity of up to 0.05 ms, were applied to each DG system. Figure 20(a) illustrates the impact on the solar PV and WE systems. Waveforms 1 and 2 show the voltage and current response of the solar PV system. The introduced faults disrupt the terminal voltage and current, affecting system stability. Waveforms 3 and 4 depict the corresponding behaviour in the WE system, where the faults similarly cause disturbances in voltage and current. Figure 20(b) presents the results for the fuel cell and utility grid (UG) systems. Waveforms 1 and 2 show the voltage and current variations in the fuel cell system due to the applied faults. Waveforms 3 and 4 illustrate the response of the UG system. Notably, fault in the UG system causes current variations that consequently alter the overall DCMG power balance. These results highlight the effectiveness of the GA method in maintaining stability under fault conditions across different DG sources.

Fig. 20.

V-I based Real-time simulation results at the terminal faults of sources (a) solar PV and wind energy. (b) fuel cell and utility grid.

Fig. 21.

V-I based Real-time simulation results at the terminal faults of resistive loads (a) and . (b) and .

Fig. 22.

V-I based Real-time simulation results at the terminal faults of energy storage systems ( and ).

Figure 21 presents the impact of faults on various load terminals (,, and ), while Fig. 22 shows the effect on the battery energy storage system (). All results shown in these figures were obtained using the GA method. Three remote faults (,, and ), each with a severity of up to 0.05 ms, were applied. Figure 21(a) illustrates the voltage and current behaviour at and . Waveforms 1 and 2 show the impact on , all faults disrupt the terminal voltage and current, with notably increasing the current and affecting the DCMG power balance. Waveforms 3 and 4 show the response of , where the faults also disrupt voltage and current. Figure 21(b) presents the results for and . Waveforms 1 and 2 show the impact on ; again, all faults disrupt voltage and current, and alters the current, affecting the DCMG power balance. Waveforms 3 and 4 depict the behavior of , where the faults disrupt the terminal voltage and current.

Finally, Fig. 22 (waveforms 1 and 2) shows the impact on . The faults disrupt both the voltage and current at the terminal. These results demonstrate the varying effects of the applied faults on different loads and the energy storage system within the DCMG. Figure 22 (waveforms 3 and 4) shows the voltage and current at , obtained using the GA method. Faults , , and (0.05 ms severity) disrupt terminal voltage and current. The waveform changes in are similar to those in due to their identical parameters. To achieve this verification, the dSPACE DL350 simulator V-I settings were adjusted to encompass 16 levels. Similarly, it can be observed that the V-Is are verified with the results obtained from the Opal-RT simulation.

Comparative analysis

Table 6 presents a summary of how existing non-AI-based methods perform, specifically regarding their parametric capacity⁴². According to this comparison, the proposed model offers advantages, notably a strong ability to correctly identify faults even with high fault resistance, thus minimizing false alarms and misoperations. Furthermore, it requires a low sampling rate and has reduced computational demands. Importantly, the model does not necessitate synchronized data acquisition.

Table 6.

Comparative analysis of DC microgrid protection methods.

Protection method	42	Proposed Methodology
Detection method	Improved LSTM	Resistance-based fault detection
Fault resistance accuracy	High (480.0 )	Low (0.1 )
Noise Issue	No	No
Sampling rate (kHz)	50.0	5.0
Communication Link	No	No
Calculation rate	Low	High
Other remarks	Small-time window. No communication is required for an intelligent detection model. Detail modelling is unnecessary, and the noise impact is robust, with good fault resistance accuracy. Applicable for long-distance HVDC transmission. Because of the plug-and-play capability, this solution is ideal for multivendor applications.	Focuses on enhancing DC microgrid reliability via optimized fault detection and control. Employs resistance-based fault detection for intermittent faults. Combines P&O, PI, and FLC methods. Utilizes GA-tuned PI controllers. Validates results with Opal-RT, showing improved V-I stability.

Table 7 presents a comparative fault analysis at different bus terminals of ​. The Fuzzy Logic controller performs better than the PI controller by reducing voltage fluctuations and minimizing current overshoot, enhancing system stability. However, the GA controller optimizes both parameters more efficiently, as seen in Table 7, where it maintains better voltage regulation and lower current variation. When fault ​ occurs at the UG bus terminal, the performance of other terminals is affected. PV, WE, and FC exhibit moderate voltage drops, while and remain relatively stable due to their energy storage capabilities. Resistive loads (​–​) maintain operational stability with minimal deviation. Figure 23 illustrates the fault impact on different terminals. UG, WE, and FC experience a 40–60% performance reduction, whereas Bat1, Bat2, and ​–​ show less than a 20% impact.

Table 7.

Comparative fault results at different bus terminals of .

Faults at different terminal	PI				Fuzzy				GA
	Voltage (V)		Current (A)		Voltage (V)		Current (A)		Voltage (V)		Current (A)
	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max
	115.3	117.5	16.94	17.38	118.3	120.6	16.85	17.36	115.15	117.6	16.91	17.4
	115.7	116.7	79	80	118.8	122.2	76.6	77.9	115.7	116.7	78.8	79.8
	113.3	116.3	47.6	48.5	118.8	120.2	50.4	51	112.5	115.5	47.7	48.5
	115.6	116.7	26.5	28	118.8	120.1	27.1	28.6	115.7	116.5	26.5	27.9
	90.6	91.8	4.4	4.6	100.3	101.3	0.06	-0.09	90.8	92	4.5	4.6
	90.6	91.8	4.4	4.6	100.3	101.3	0.06	-0.09	90.8	92	4.5	4.6
	90.6	92.5	3.48	3.54	100.3	101.3	3.84	3.88	90.6	92	3.48	3.52
	90.7	92	3.48	3.54	100.3	101.3	3.84	3.88	90.7	92	3.48	3.525
	90.5	92	3.46	3.54	100.3	101.3	3.84	3.88	90.7	92	3.48	3.525
	90.7	92	3.48	3.54	100.3	101.3	3.84	3.88	90.7	92	3.48	3.525

Fig. 23.

Comparison of the performance of fault results at different Bus terminals of .

Table 8 presents the transient response of different bus terminals after fault ​. The Fuzzy Logic controller outperforms the PI controller by effectively reducing transient spikes, leading to a more stable system response. However, the GA controller provides the most optimized results by further minimizing voltage and current deviations. As observed in Table 7, the Fuzzy controller limits the transient voltage at UG to 4.2 V, whereas PI holds it at 2 V. The GA controller further stabilizes UG with only 1.8 V deviation, demonstrating its efficiency in transient fault mitigation. During the occurrence of ​ at UG, Table 7 indicates varying impacts on other terminals. PV, WE, and FC exhibit transient disturbances, with FC peaking at 2.4 V under Fuzzy but optimally controlled to 1.8 V under GA. Battery terminals (Bat1, Bat2) exhibit near-zero transient impact under both Fuzzy and GA, ensuring stable energy storage. Resistive loads (​–​) experience minimal fluctuations, where GA effectively maintains current deviation below 0.4 A. Figure 24 shows that UG and FC suffer a 50–70% transient deviation, while PV, WE, and resistive loads remain relatively stable.

Table 8.

Comparative transient results of after fault at different bus terminals of .

Faults at different terminal	PI		Fuzzy		GA
	Voltage (V)	Current (A)	Voltage (V)	Current (A)	Voltage (V)	Current (A)
	2	0.34	2	0.30	1.8	0.1
	1	0.12	2	1.4	0.8	1.0
	2	0.5	2.4	0.6	1.8	0.5
	2	0.04	4.2	0.4	1.8	0.03
	1.0	0.6	0	0	0.8	0.2
	1	0.6	0	0	0.8	0.2
	1.4	1.2	1	0.05	0.04	0.14
	1.6	0.6	1	0.025	1	0.4
	1	0.06	1	0.04	0.8	0.04
	1.6	0.6	1	0.025	1	0.4

Fig. 24.

Comparison of the performance of transient results of after fault at different Bus terminals of .

Table 9 presents the comparative fault results for different bus terminals under ​. The PI controller exhibits significant voltage and current deviations across all terminals. In contrast, the Fuzzy Logic controller enhances stability by reducing fluctuations, and the GA controller further optimizes fault handling. During at ​, the voltage at ​ is 79 V under PI but improves to 88 V under Fuzzy, while GA stabilizes it at 79 V with minimal current variation. Other bus terminals experience varying impacts. PV remains relatively stable, with a minor voltage dip from 113.3 V (PI) to 112.7 V (GA). WE shows improved regulation under GA at 115.5 V compared to 113.5 V in PI. FC is well-regulated in GA, reducing fluctuations in voltage and current. UG maintains stability under Fuzzy at 116.4 V, while GA further optimizes it to 112.6 V. Battery terminals (Bat1 and Bat2) show instability under Fuzzy with negative current spikes but remain stable under GA. Figure 25 indicates that ​ impacts around 50–70% of bus terminals, particularly WE and UG.

Table 9.

Comparative fault results at different bus terminals of .

Faults at different terminal	PI				Fuzzy				GA
	Voltage (V)		Current (A)		Voltage (V)		Current (A)		Voltage (V)		Current (A)
	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max
	113.3	116.25	16.7	17.5	116.5	118.8	16.75	17.49	112.7	115.8	16.35	17.7
	113.5	116.5	79	81	116.5	118.8	76.8	78.8	112.5	115.5	79.5	81.5
	113.2	116.4	46.3	47.9	116.5	118.5	48.8	50.3	112.8	115.8	46.2	47.8
	113.4	113.5	25.4	25	116.4	118.6	25.3	25.55	112.6	115.5	24.95	25.2
	80	85	3.5	6.3	88.5	100	-0.6	0.6	78	85	3.5	5.8
	80	85	3.5	6.3	88.5	100	-0.6	0.6	78	85	3.5	5.8
	79	91	3.5	9	88	101	3.9	10.2	78	90	3.5	9
	80	91	3.05	3.5	88.5	93	3.4	3.56	79	91	3.01	3.5
	80	91	3.05	3.5	91	100.5	3.4	3.85	79	91	3	3.5
	80	91	3.05	3.5	88.5	93	3.4	3.56	79	91	3.01	3.5

Fig. 25.

Comparison of the performance of fault results at different Bus terminals of .

Table 10 presents transient results after fault ​ at ​. The PI controller exhibits higher voltage and current fluctuations across terminals. Fuzzy logic reduces these variations, enhancing system stability, while GA further optimizes transient response. In Fig. 26, PI shows significant transient deviations, particularly at FC (2 V), WE (1.6 V), and UG (1.2 V), whereas fuzzy logic improves regulation, reducing WE and UG fluctuations to 1 V. GA demonstrates superior fault tolerance, minimizing voltage deviations across all bus terminals. During ​, ​ voltage drops significantly, while , , ​ experience minor disturbances. PV voltage improves under fuzzy (1.2 V) but remains low in GA (0.04 V). WE stabilizes at 1 V in fuzzy and drops to 0.2 V in GA. FC exhibits enhanced stability under GA, reducing transients effectively. UG voltage variations are minimized under GA (0.4 V). Battery terminals maintain improved performance under GA, reducing transient spikes. Figure 26 indicates that ​ impacts approximately 50–70% of bus terminals, with GA mitigating effects efficiently.

Table 10.

Comparative transient results of after fault at different bus terminals of .

Faults at different terminal	PI		Fuzzy		GA
	Voltage (V)	Current (A)	Voltage (V)	Current (A)	Voltage (V)	Current (A)
	0.4	0.18	1.2	0.56	0.04	0.02
	1.6	0.7	1	1	0.2	0.06
	0.5	0.2	2	0.2	0.4	0.02
	1.2	0.16	1	0.1	0.4	0.02
	1	0.4	0	0	0.2	0.1
	1	0.4	0	0	0.2	0.1
	0.8	0.06	0.4	0.05	0.2	0.045
	0.8	0.02	1.4	0.015	0.4	0.005
	0.8	0.02	0.4	0.16	0.4	0.006
	0.8	0.02	1.4	0.015	0.4	0.005

Fig. 26.

Comparison of the performance of transient results of after fault at different Bus terminals of .

Table 11 presents the comparative fault results for ​ occurring at ​. The PI controller exhibits larger voltage deviations, particularly for FC (112.5–114 V) and UG (112.5–114.5 V). Fuzzy logic improves voltage regulation, maintaining FC at 115.7–117.5 V and UG at 115.5–117.5 V. GA further optimizes stability, minimizing fluctuations across all terminals. In Fig. 27, PI shows higher variations across loads and energy sources, whereas fuzzy logic reduces transients, and GA achieves optimal regulation. During ​, ​ undergoes a voltage drop, while , , ​ maintain stability with minimal variations. PV voltage remains highest under fuzzy logic (115.7 V), while GA slightly reduces it (111.7 V). WE voltage stabilizes under GA (111.3–114 V), while UG exhibits enhanced performance (111.8–114 V). Batteries experience transient effects under PI but improve with fuzzy logic and GA. The figure shows that ​ affects 40–60% of bus terminals, with GA minimizing impact effectively. These results highlight the superiority of GA over PI and fuzzy logic, ensuring optimized fault detection and control for DC microgrids.

Table 11.

Comparative fault results at different bus terminals of .

Faults at different terminal	PI				Fuzzy				GA
	Voltage (V)		Current (A)		Voltage (V)		Current (A)		Voltage (V)		Current (A)
	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max	Min	Max
	112.5	114.35	16.86	17.54	115.7	117.7	16.8	17.38	111.7	114.15	16.6	17.56
	112.5	115.5	79.5	81	115.5	117.5	78.5	79.5	111.3	114	80	81.2
	112.5	114	45.3	45.8	115.7	117.5	46.2	46.7	111.8	113.5	45.2	45.7
	112.5	114.5	23.1	25	115.5	117.5	22.8	25	111.8	114	23.05	25
	75.5	78.5	4.5	6.4	84	92	-1	0.4	74	79	4.5	6.6
	75.5	78.5	4.5	6.4	84	92	-1	0.4	74	79	4.5	6.6
	75	84	6.3	8.7	83.8	93	7.5	10.4	74	83.5	6.2	8.6
	75	83.8	2.88	3.22	84	93	3.2	3.55	74	83.5	2.83	3.2
	75	84	3.2	9.1	84	93	3.8	10.8	74.5	83.5	3.2	9.1
	75	83.8	2.88	3.22	84	93	3.2	3.55	74	83.5	2.83	3.2

Fig. 27.

Comparison of the performance of fault results at different Bus terminals of .

The transient analysis in Table 12; Fig. 28 highlights the superiority of the GA over fuzzy logic and PI controllers in enhancing fault response. GA effectively optimizes fault mitigation in DC microgrids, ensuring stability. When fault occurs at , PI control results in higher transient voltage and current variations across the system, with values reaching 1 V at , 1.6 V at UG, and 2.4 A at WE. The fuzzy controller improves system stability by reducing voltage deviations, especially at FC (2 V), while maintaining UG stability (1.6 V, 0.2 A). However, GA outperforms both, minimizing transient effects with a maximum voltage of 0.4 V across most terminals and significantly reducing current to below 0.1 A. When occurs at , the transient voltages at other bus terminals (, , ) remain stable under GA compared to PI and fuzzy. Under PI, , , and experience voltage fluctuations around 1 V, whereas under GA, they remain 0.2 V, demonstrating improved damping. Figure 28 highlights the percentage impact of on all terminals. PI control exhibits a sharp response, causing fluctuations exceeding 70% in UG and WE, whereas fuzzy logic dampens this effect by nearly 30%. GA further improves stability, restricting fluctuations to below 10% across all buses.

Table 12.

Comparative transient results of after fault at different bus terminals of .

Faults at different terminal	PI		Fuzzy		GA
	Voltage (V)	Current (A)	Voltage (V)	Current (A)	Voltage (V)	Current (A)
	0.6	0.18	1.1	0.4	0.4	0.02
	0.6	0.1	2.4	1.4	0.4	0.06
	0.6	0.06	2	0.4	0.4	0.04
	0.6	0.03	1.6	0.2	0.4	0.02
	1	1	0	0	0.4	0.1
	1	1	0	0	0.4	0.1
	1	0.08	0.8	0.042	0.2	0.02
	1	0.4	0.6	0.02	0.2	0.01
	1	0.4	0.8	0.08	0.08	0.02
	1	0.4	0.6	0.02	0.2	0.01

Fig. 28.

Comparison of the performance of transient results of after fault at different Bus terminals of .

GA’s superior fault tolerance enhances microgrid reliability by optimizing fault detection and control, ensuring rapid mitigation, improved efficiency, and resilience. This analysis supports future comparative studies for advanced fault management strategies in microgrid applications, strengthening system stability and performance.

Conclusion

This study presents a novel approach for detecting line-to-ground (LG) faults in DC microgrids (DCMGs), incorporating voltage and current disturbances at terminal points. Various fault scenarios were analyzed using MATLAB simulations, integrating multiple control strategies, including PI controllers, fuzzy logic controllers (FLCs), and optimization algorithms. The proposed protection mechanism effectively detected and isolated short-circuit faults, mitigating fault current impact. A comparative evaluation between optimized and non-optimized configurations, validated through Opal-RT real-time simulations, demonstrated enhanced fault detection accuracy, improved system reliability, and minimized operational disruptions. Despite these advancements, limitations exist, such as the reliance on predefined parameters for FLC tuning and the computational complexity of genetic algorithm-based optimization. Future research should explore adaptive machine learning models for real-time fault detection, hybrid optimization techniques for enhanced controller performance, and experimental validation under large-scale microgrid conditions. Extending this methodology to AC grids and hybrid AC/DC systems could further enhance fault detection and system stability in renewable-integrated networks.

Abbreviations

VSC

Voltage source converter

LL

Line to line

LG

Line to ground

SC

Short circuit

CB

Circuit breaker

R_s

Relay

DBR

Diode bridge rectifier

List of symbols

Resistive loads

Fault currents

Output current of the PV

Current generated by PV cells

Reverse saturation current

Output voltage of the PV

Terminal voltage

Series resistance

Shunt resistance

Present voltage of the solar PV

Previous voltage of the solar PV

Change in power of the solar PV

Change in voltage of the solar PV

Change in current of the solar PV

ρ

Air density

R

Rotor blade radius

Wind speed

Power coefficient

Pitch angle

Tip speed ratio

Equivalent resistance

Resistance of the different load zones

Discharge current of time

Lower threshold current

Upper threshold current

δ

Fault characterization threshold

Author contributions

Banothu Somanna, Sushma Gupta, Jatoth Rajender,: Conceptualization, Methodology, Software, Visualization, Investigation, Writing- Original draft preparation. Muhannad Alshareef, Abdulrahman Babqi: Data curation, Validation, Supervision, Resources, Writing - Review & Editing. Borchala Namomsa, Sherif S. M. Ghoneim: Project administration, Supervision, Resources, Writing - Review & Editing.

Funding

This work is funded and supported by the Deanship of Graduate Studies and Scientific Research, Taif University.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.