Due to the global coal crisis and the rising price of crude oil in the global market, many emerging nations must now lessen their reliance on fossil fuels to meet their energy needs (Burke et al., 2018). The inception of renewable energy-based electricity generation and the introduction of an electric transportation system can help a country to become energy independent (Gielen et al., 2019). The introduction of a DC microgrid (DCMG) can also prove to be a crucial breakthrough in the process of electricity generation and utilization. The idea of a DCMG is feasible in the near future due to the significant increase in electronic loads and smart appliances that primarily use DC power (Al-Ismail, 2021).

DC microgrids have several advantages over AC microgrids like low effective resistance, absence of inductive and capacitive reactance, lesser number of power converting stages as compared to AC microgrids, no frequency related issues in DG integration, and many more (Justo et al., 2013). However, the AC microgrid persisted in being preferred over the DC microgrid because of lack of a well-defined standards, technological advancement, and trustworthy protective measures (Joos et al., 2017). The protection of DCMG is one of the challenging issues due to various reasons such as the absence of current-zero crossing points, lack of frequency-dependent parameters, absence of sequence components, quickly rising fault current, etc., (Augustine et al., 2018). As the DC microgrid possesses less physical inertia as compared to the AC system (Sun et al., 2022), hence the fault current in these systems rises rapidly to an intolerant level in a very short duration of time (Beheshtaein et al., 2019). This necessitates the requirement of very fast protection schemes for DC microgrids.

In the past various conventional protection schemes such as over-current protection (Shabani and Mazlumi, 2020), rate of change of current-based protection (Chauhan et al., 2022), differential protection (Dhar et al., 2018), resistance-based protection (Kesava Rao and Jena, 2022), (Yadav and Tummuru, 2020), etc., Have been suggested for the relaying of DC microgrid. Some of the merits of conventional relaying schemes are easy implementation and low computational burden. However, the issues associated with these conventional protection schemes are slow response; high sensitivity for temporary faults; unable to protect against high resistance faults (HRF); And poor selectivity towards temporary faults (Kant and Gupta, 2022). To attain the objective of a faster response, a travelling wave (TW)-based relaying scheme has been proposed in various works of literature. The application of TW-based schemes for the protection of high voltage DC (HVDC) transmission systems has been reported in various kinds of literature. The fact that TW-based schemes do not rely on the magnitude of the fault current is one of their key advantages; as a result, these schemes function satisfactorily against HRFs (Zhang et al., 2012). Although there are various TW-based relaying schemes available for HVDC systems, the application of TW-based protection in low and medium-voltage DC systems (MVDC) is still an unexplored area. In (Saleh et al., 2019), a TW-based relaying strategy has been suggested for an MVDC microgrid. The proposed scheme in this literature is independent of the arrival times of the travelling wave and utilizes waveshape properties of the very first TW initiated by the fault. Due to the requirement of single-ended measurement only, no communication link is required to implement such schemes. Nowadays, many advanced techniques such as wavelet transform (WT), fuzzy logic, artificial neural network (ANN), etc. Are also being applied to improve the performance of the conventional DCMG protection schemes. As discussed in (Sharma et al., 2021), machine learning (ML) techniques are also used in combination with differential protection scheme for quick fault detection. This method uses a ML based support vector machine (SVM) classifier to increase the scheme’s resilience to transient or momentary faults. The problem with this scheme is that it can detect faults up to 10Ω resistance only. In (Mohanty et al., 2021), a cosine similarity-based relaying strategy has been presented for DCMG protection. The operating time of the proposed scheme is 3 m. But again, it can detect faults up to 10Ω resistance only. In (Jayamaha et al., 2019), a WT and ANN-based relaying strategy for a DCMG has been proposed. The proposed scheme can detect the ground fault with a maximum resistance up to 300Ω, but the operating time varies between 3–5.5 m, depending on the type of fault.

Overall, there are a few things, in the available schemes, that may be improved and therefore, in this paper, an advanced short-circuit protection scheme for a DCMG is proposed. The key matters identified with previously discussed methods are a) the operating time of the relay; b) protection against HRFs; c) robustness during external fault or disturbances; and d) the operation during islanded operation. The proposed scheme is developed keeping all these issues into consideration and therefore, the main contributions of the developed scheme are as follows:

• It is an ultra-fast relaying scheme that performs a multi-resolution analysis of TW using the DWT.

The remainder of the paper is structured as follows: Section 2 describes the DC microgrid test system in detail; Section 3 offers a brief explanation on the TW and DWT; Section 4 introduces the suggested algorithm; and Section 5 includes the simulation results and analysis. Section 6 presents the real-time validation of the results, Section 7 presents a comparison of the proposed scheme with the existing schemes, and Section 8 presents the conclusion.

In this study, a 500V ring-main bipolar DCMG architecture, as shown in Figure 1, has been considered for the implementation of the suggested scheme. This test system is a modification of the system considered in (Chauhan et al., 2022). The distributed generation (DG) sources, i.e., PV and battery are connected at bus two through appropriate DC-DC converters. The system is integrated to the grid at bus 1 with the help of a bi-directional voltage source converter (VSC). The bi-directional VSC, connected at the grid interface, enables the power transfer from/to DCMG. When connected to grid, the system voltage is maintained by the grid; and in islanded mode, it is maintained by the local sources. The battery is coupled to the bi-directional DC-DC converter, which is controlled to allow the battery to either absorb or supply power from/to the DC system. The PV-side DC-DC converter is managed to provide the DC system with the maximum real power available. To maintain the positive and negative pole voltage, a voltage balancer circuit is used at each bus. The grounding on the DC side is provided at each bus at the mid-point of voltage balancing capacitors (C1 & C2). A total load of 300 kW (200 kW on bus 3 & 100 kW on bus 1) is connected to the DC system. A detailed description of the components connected to this test system is given in Table 1. The proposed scheme has been implemented for the protection of positive and negative pole lines between bus 1 and bus 2, i.e., lines P12 and N12, respectively. Hence the relays are connected at each terminal of both lines, i.e., R_P1, R_P2, R_N1, and R_N2, as shown in Figure 1. Any fault outside this protected zone is considered an external fault.

The proposed algorithm is developed by combining the concept of TW and DWT. This algorithm does not utilize the timing of the incident wave at the line terminals. Instead, this algorithm is based on the waveshape properties of the travelling waves. The travelling waves were calculated by Eqs 2, 3 (Tiwari et al., 2022). V t w f k = 1 2 Δ V f k + Z s * ∆ i f k (2) V t w b k = 1 2 Δ V b k − Z s * ∆ i b k (3) Where subscript f and b denote the forward and backward wave quantities, respectively. V _tw (k) is the voltage travelling wave at sample k and Z _s is the surge impedance of the dc cable. The change in voltage and current at the i ^th bus can be calculated as in Eqs 4, 5. Δ V i k = V i k − V i k − 1 (4) Δ I i k = I i k − I i k − 1 (5)

The schemes based on the incident time of travelling waves require high frequency transient recorders to capture the travelling waves at the line terminals, and use of such costly equipment is not very economical for distribution lines. Hence in this study, the properties of the travelling are extracted by employing the signal processing tools, i.e., wavelet transform. The first incident forward voltage travelling wave is decomposed into the various high-frequency and low-frequency components by using the DWT’s dyadic filter bank theory, as shown in Figure 2. The selection of the mother wavelet is also an important criterion for the implementation of WT. Some of the literature available (Sabug et al., 20192020) suggest that Daubechies as mother wavelet gives better result in the case of power system studies. The details of the DWT used in this algorithm have been provided in Table 2.

The energy content of the extracted DWT coefficients of the voltage travelling wave is calculated by Eq 6. E t w f = ∑ i = 1 N d t w i 2 (6) Where d _tw is the high-frequency detail coefficient of the voltage travelling wave (V _tw ), and N is the number of samples. In this case, N is taken as 16 samples. The number of samples (N) were decided based on the rigorous simulation studies. The energy content of the DWT coefficient (d ₃ ) following a PG fault and PP fault is presented in Table 3.

If this energy content is more than the pre-set value, then the fault will be detected. The proposed algorithm is also able to classify the faults. With PP fault, the energy content of the DWT coefficient at both positive and negative poles (i.e., E _twP and E _twN ) will almost be the same. Hence, the difference between both energies will be very small. If this difference energy is less than the pre-specified value, the fault will be a PP fault. While in the case of PG-fault, this energy difference will be more than the pre-set value. The pre-specified threshold value has been determined by simulating the test model for different fault and non-fault conditions and then observing the energy content of the DWT coefficients in each case. The flow chart of the proposed algorithm is shown in Figure 3.

The DCMG test system, as illustrated in Figure 1, is simulated in the MATLAB/SIMULINK environment for the validation of the proposed algorithm. The performance of the suggested algorithm is evaluated in both fault- and non-fault-related scenarios.

The operational validation of the proposed algorithm is conducted with the OPAL-RT time simulator. The laboratory set-up for the real-time validation is shown in Figure 17. The test system shown in Figure 1 is used for real-time simulation. Several cases were simulated for real-time validation of the proposed scheme, and the result of one such case has been presented here.

In Table 5, the performance of the suggested algorithm is contrasted with that of other currently available techniques. Table 6 shows that the suggested algorithm can identify a ground fault up to a fault resistance of 900 Ω in the grid-integrated mode, which is greater than the other approaches stated in the literature. The suggested algorithm operates in 0.64 m, which is faster than the majority of the algorithms described in the literature. The suggested system does not malfunction in the presence of an outside disturbance like an AC side fault, a grid outage, or a DG outage. Since it is unaffected by external or non-fault situations, this protection strategy offers superior stability than others. Investigating whether a protection strategy is adaptable enough to function well in both grid-integrated and islanded modes is another crucial situation. This situation has not been taken into account by the numerous protection strategies studied in the literature. However, the efficiency of the suggested algorithm has been carefully examined for both grid-integrated and islanded operations. The study’s findings confirm that the suggested strategy can properly function in both operating modes. Additionally, in islanded mode, it is capable of detecting PG faults up to 1,300 Ω fault resistance. As a result, the suggested scheme proves to be superior to the other existing schemes in many aspects.

For the preservation of DCMG, a novel protection system has been described in this study. In order to verify the effectiveness of the suggested approach, a test model for a 500 V ring-main bipolar DCMG was created in the MATLAB/Simulink environment. Faults such as PPG, NPG, and PP were then simulated. The suggested algorithm has been shown to be a very efficient solution for the protection of DC microgrids and can function well in both grid-integrated and islanded modes. Additionally, it is capable of detecting HRFs of up to 900Ω in grid-integrated mode and 1,300 Ω in islanded mode. The operating time of the proposed algorithm is independent of the type of faults and fault resistances, and it can detect all types of faults in 0.64 m, which is quite fast. The suggested algorithm’s resilience has been tested during faults and external disturbance transients, and it has been discovered that this scheme is unaffected by these external disturbances. The suggested approach is therefore very robust, accurate, and highly selective and offers a quick, stable, and trustworthy solution for the protection of DC microgrids.