Principle and Control Design of a Novel Hybrid Arc Suppression Device in Distribution Networks

A single line-to-ground (SLG) fault may lead to a more severe line-to-line fault and power supply interruption if the ground-fault current exceeds a certain value. Arc suppression device (ASD) is a good solution to eliminate the ground-fault current. A novel hybrid ASD is proposed in this article, which consists of a passive device and an active one. The passive device utilizes multiterminal breakers and an isolation transformer to couple a secondary voltage of a zig-zag grounding transformer to the neutral point to compensate the majority of the ground-fault current. The active device uses a single-phase voltage source inverter to eliminate the residual fault current due to the leakage inductance of the zig-zag grounding transformer in the passive device. A dual-loop voltage and current control method for the active device is designed for the accurate residual current compensation. Results of simulation and prototype experiment validate the effectiveness of the proposed hybrid ASD. The proposed hybrid ASD does not need to detect distributed line-to-ground parameters, and it has lower cost, less control complexity, higher reliability, and better performance, compared to other ASDs.


I. INTRODUCTION
R ELIABLE power supply is of great significance in distribution networks [1]. To enhance the reliability, a neutral noneffectively grounded distribution system is usually allowed to maintain power supply for about 1 h before a single lineto-ground fault (SLG) is isolated. However, if the fault current Manuscript received May 28, 2020; revised September 9, 2020 and November 23, 2020; accepted December 21, 2020 exceeds a certain value, the fault-induced arcs may not selfextinguish, resulting in arc overvoltage and insulation damage.
In this way, a tolerable SLG fault could lead to more severe line-to-line faults and power supply interruptions eventually [2], [3]. Therefore, it is highly desirable to extinguish fault-induced arcs when a SLG fault occurs [4], [5].
The existing arc-suppression methods can be divided into two categories: current-based compensation and voltage-based compensation. One typical current-based compensation method is to earth the system neutral through an arc suppression coil (ASC) [6]- [8]. The ASC can suppress arcs by injecting an inductive current to compensate the capacitive fault current, so that nonpermanent SLG faults can be self-extinguished without breakers operations. However, the ASC with fixed parameters is difficult to fully compensate the capacitive current (lead to series resonance otherwise), active power, and harmonic current. Another current-based compensation method is to earth the system neutral through a current-type inverter, which is able to adjust the injected inductive current based on the measured capacitive current [9]- [11]. Nevertheless, it is difficult to accurately measure the capacitive current in distribution networks, especially under an SLG fault condition [12].
The voltage-type arc-suppression methods are also proposed in literature [13]- [15]. The ground-fault transfer device can automatically compensate the ASC by quickly grounding the fault phase [14]. However, it relies on accurate faulty phase selection to prevent line-to-line fault. Moreover, its cost is very high when applying to medium-voltage distribution networks. Recently, voltage-source type single-phase inverter is proposed to control the neutral-to-ground voltage to the opposite of the faulty phase voltage. This method does not rely on the distributed parameters and the residual current. Nevertheless, the major shortcoming of the voltage-type inverter-based arc-suppression method is that it requires an electronic device with very large capacity to extinguish the fault-induced arc [16].
Moreover, an additional zig-zag grounding transformer is usually needed to create a neutral point for most mediumvoltage distribution networks, which adopt the high-resistance grounding or resonant grounding (RG) method [17]. When an SLG fault occurs, the neutral-to-ground voltage may have phase shift problem in neutral point as the leakage inductance (a zero-sequence impedance) of the zig-zag grounding transformer can build up a neutral voltage shift [18]. This makes it difficult to control the neutral-to-ground voltage to the expected value. Thus, the conventional arc-suppression methods are inadequate to increase the neutral-to-ground voltage to the line-to-ground voltage and reduce the residual current to zero at the same time [19]. There is an urgent need in a comprehensive arc suppression device (ASD) that can suppress the ground-fault current under different ground-fault resistances but require less electronic equipment to reduce the cost.
To address the abovementioned issues, a novel hybrid suppression device (ASD) is proposed in this article. The grounding system is composed of a zig-zag grounding transformer, a singlephase isolation transformer, multi terminal breakers, and a voltage source inverter (VSI). The hybrid ASD employs two control algorithms. One is the passive arc-suppression method through the zig-zag grounding transformer for high-power compensation, and the other is the active arc-suppression method through the VSI to eliminate the effect of neutral point phase shift caused by the zig-zag grounding transformer. By regulating inverter's output voltage, the active method can generate the compensated current for the neutral voltage error from phase shift in the neutral point. Together with the passive method, the neutral-to-ground voltage can be controlled to exactly match the opposite of the faulty phase voltage. Meanwhile, the residual fault current can be reduced to near zero to completely distinguish the fault-induced arcs.
The rest of the article is organized as follows. In Section II, the principle of the proposed hybrid ASD is introduced, and VSI-based dual-loop control is presented in Section III. The proposed hybrid ASD is simulated in a typical neutral noneffectively grounded 10 kV distribution network model in MAT-LAB/Simulink in Section IV. Experiment results on a 10 kV prototype are presented in Section V. Finally, Section VI concludes this article.

II. PRINCIPLE OF HYBRID ASD
A typical neutral noneffectively grounded 10 kV distribution network with the proposed hybrid ASD is illustrated in Fig. 1. The hybrid ASD is comprised of a passive ASD and an active ASD. The passive ASD consists of a zig-zag grounding transformer, a single-phase isolation transformer, and multiterminal breakers. The Zig-Zag grounding transformer (T 1 ) is used to create the neutral point, while the single-phase isolation transformer (T 2 ) is utilized to inject a voltage to the neutral point. The secondary sides of the both transformers are connected by multiterminal breakers (S an , S bn , S cn , S ap , S bp , S cp ) to select the suitable line-to-line voltage (u q ) for the injected neutral voltage. The active ASD consists of an uncontrollable rectifier and a single-phase voltage source inverter. The active ASD is connected to the secondary side of the zig-zag grounding transformer through three input inductors, while connected to the isolation transformer through a LC filter.
The passive ASD is used for high-power compensation, but the leakage inductance of the zig-zag grounding transformer may introduce phase error, resulting in the phase error of the neutral voltage [17]. Meanwhile, as a lower power device, the active ASD can inject the voltage to compensate the phase error of the neutral voltage and the SLG fault residual current. The next two sections will introduce the principles of both the passive ASD and the active ASD. The related variables are summarized in Table I.

A. Principle of Passive ASD
As shown in Fig. 1, the connection type of the zig-zag grounding transformer is Zny11. The phasor relationships among the line-to-line voltages (u ab , u bc , and u ca ) and the three-phase positive-sequence voltages (U A , U B , and U C ) are presented in Fig. 2. The line-to-line voltage is totally opposite to the associated phase positive-sequence voltage, e.g., u ca and U A , u ab , and U B , or u bc , and U C . The magnitude relationship in each group depends on the value of turn ratio N T1 . The line-to-line voltage u q can be expressed as below. . (1) According to the ideal transformer principle, When ignoring the sign, the neutral-to-ground voltage (U N ) can be equivalent to the line-to-line voltage (U X ) if the turn ratio number of N T2 equals that of N T1 , meaning that Therefore, if an SLG fault occurs on phases A, B, and C, respectively, we manage to have U N = u ca = −U A , U N = u ab = −U B and U N = u bc = −U C within short notice, and that can be expressed as follows: In this way, the passive arc suppression can be achieved for the high-power compensation.
When an SLG fault occurs in phase C, as further arcsuppression needs, a set of breakers S bn and S cp are turned ON to generate the line-to-line voltage u bc as the input of the secondary-side in the single-phase isolation transformer (T 2 ), u q . Similarly, if we assume that SLG fault (10 Ω) occurs in phases A and B, respectively, then multiterminal breakers will turn ON the breakers S an , S cp , and the breakers S bn , S ap for generating the correct line-to-line voltage to control the neutral voltage at the neutral point in order to suppress the fault arc.
However, the passive arc suppression is incapable of compensating the phase shift caused by the leakage impedance of transformers and generated during operation process in multiterminal breakers. In order to solve this problem, the active arc suppression devise is introduced. Thus, if we consider the phase error impact, the leakage impedance of T 1 shall be included in further arc suppression analysis, and u Q is the isolation transformer secondary-side voltage after considering inductive leakage reactance Z lk-T1 and Z lk-T2 .

B. Principle of Active ASD
The topology of distribution network in Fig. 1 is converted to the low-voltage side of the isolation transformer and shown in Fig. 3.  According to Kirchhoff voltage law (KVL), we can obtain

1) Symmetric Distribution Network:
Without loss of generality, it is assumed that an SLG fault occurs on phase C. The zero-sequence circuit of the distribution network with the passive ASD can be represented in Fig. 4. Fig. 4(a) is the simplified distribution network without implementing ASD and Fig. 4(b) adopts passive ASD. G ࢣ denotes the impedance of the distribution network and is shown in (6), where Y X is the line-to-ground admittance, i.e., (X = A, B, or C) and ω 0 denotes the fundamental angular frequency.
Assume SLG fault occurs on phase-C and adopt only the passive ASD with studying in the influence of the leakage impedance of the transformers Z ࢣlk = N T2 2 Z lk-T1 + 3Z lk-T2 , we consider to short-circuit U i (equivalent as active ASD) above.
Therefore, in Fig. 4(a), according to KVL, we can simply attain the neutral voltage in (7) for successfully eliminating the fault arcs if the fault current is assumed to be equal to zero (I F = 0 and U F = 0), That is, by adopt only the passive ASD, the fault current can be constrained down to near zero. But to achieve complete U N = −U C , we need to further analyze −3Z ࢣlk G ࢣ −1 U C , which can be treated as residual voltage of neutral point U N _ res , as shown in (8) and Fig. 4(b). U N _ res is caused by the phase error of neutral voltage from leakage impedances Z ࢣlk exists in transformers.
Notice that when the influence of the leakage impedances Z ࢣlk is considered, the problem will remain in the passive ASD, making the controlled neutral voltage hard to reach expectation value, Therefore, the exact equation of residual current I res can be obtained for further ground-fault compensation, by the principle of KVL and voltage divider method.
The passive ASD can suppress the fault arc down to certain promised range, leading to the arc condition of not being rekindled. However, it cannot compensate fully the neutral voltage error and residual fault current. Therefore, the low-power active ASD method is co-operated with passive ASD method in Fig. 5, named as hybrid ASD. It constrains the residual fault current down to zero for erasing the neutral voltage error. Then, according to (7), it achieves U N = −U C .
An equivalent voltage source U i is treated as the output and the control target of the dual-loop VSI. According to the Kirchhoff theorem, the flowing current I Gࢣ through the impedance of the distribution network and the flowing current I lk through the leakage impedances Z ࢣlk can be obtain as For further analysis of active ASD, the only two nodes on Fig. 4(b) is marked for the principle of Nodal-Voltage method, and the voltage between node "a" and "b" U ab can be obtained (10) According to the Kirchhoff theorem, the flowing current I Gࢣ through the impedance of the distribution network and the flowing current I lk through the leakage impedances Z ࢣlk can be obtain as An equivalent voltage source U i is treated as the output and the control target of the dual-loop VSI. According to Kirchhoff's current law, the expression of the output current of inverter i i is as follows: It should be noticed that expression of i i for compensating the residual fault current to zero can be obtained via setting I res to zero. The current I full _ com for full ground-current compensation is defined as the current injected to neutral, which ensures the ground current to be zero under the conditions of ground impedance variation.
Obviously, if the injected current i i is controlled to be equivalent to I full _ com , the faulty phase voltage and ground-fault residual current will be limited to zero, and thus the fault arc can be extinguished. The neutral-to-ground voltage can be adjusted to expectation value, then, the neutral voltage error can be erased.
Therefore, it is necessary to analysis the relationship between i i and U i , since the active ASD is treated as voltage source in our case. According to Thevenin-Norton theorem, the active ASD equivalent impedance Z i is in parallel with ground-fault resistance and the impedance of the distribution network as well as the leakage impedances of transformers, expressed as follows: Then, following the Ohm's law, the active ASD voltage U i can be obtained as follows: The active ASD voltage U i is controlled to reach the expectation value as (16) so that the output current i i can be equal to −i sum , which compensates the neutral voltage error and restrain residual fault current at the same time. The compared amplitudes of active power between passive ASD and active ASD are shown as bellow, where the passive ASD equivalent impedance is Z F // Take notice that the zero-sequence impedance of the transformers has a magnitude Z ࢣlk = N T2 2 Z lk-T1 + Z lk-T2 of less than 1 from Table II, the magnitude of i i is smaller than that of U N , and they both share the same denominator in (17) and (18). Therefore, the active power of active ASD P active is way less than that of passive ASD P passive . According to values in Table II and (17) and (18), the capacity of active part is 12.4% of total in the proposed method.
2) Asymmetric Distribution Network: Principle of passive ASD method is the same in either symmetric or asymmetric system. Assume passive ASD has been activated as SLG fault on phase C, resulting on ground-fault current I F being constraint down to the residual current I res . The algorithm of active ASD method in asymmetric distribution network is presented below, where all the parameters are shown in Table I. In asymmetric distribution network, the neutral-to-ground voltage becomes unbalanced. Hence, U 00 denotes neutral unbalanced voltage, after the operation of passive part in hybrid ASD, we assume U 00 = −U C . Parameter relationships among three phases can be simplified by using the rotating coefficient a = e j2Π/3 . Thus, analysis of the principle of Nodal-Voltage method in Fig. 3 without the active ASD part is presented in (19) and (20), the voltage between dot " 1 " and " 0 ", U 10 can be obtained, where Y ࢣ10 denotes equivalent admittance between dot " 1 " and " 0 ".
According to Kirchhoff's theorem, the new expression of the output current of inverter i i-asy in Fig. 3 can be calculated as follows: where That is, via setting I F to zero and the output current of inverter i i can be set as to be the opposite of summed current i sum-asy , then complete arc suppression can be achieved. Basically, the advantage by using hybrid ASD method is that it costs less for sustaining little active device capacity. Passive ASD method is applied for high-power compensation in groundfault current, then, the inverter-input (active ASD) voltage is controlled to generate the current for the compensation to the neutral voltage error, which allows the neutral-to-ground voltage to be controlled as expectation value.

A. Model of Distribution Network
Typical parameters for case study are presented in Table II. The distribution network is a 10 kV power system shown in Fig. 6. As the bolted ground fault rarely happens, the groundfault resistance Z F is chosen to be 10 to 10 kΩ in case study [20]. Thus, the power stage of the distribution network is in fundamental frequency, in which the distribution network can be treated as series connected voltage source E 0 and zero-sequence impedance of the distribution network Z 0 , expressed in (22) and (23), where R S , C S are set as the symmetric line-to-ground parameters of three-phase We assume SLG fault happens on phase C. The equivalent voltage source in (22) is the symmetric voltage caused by the line-to-ground parameters while the equivalent impedance Z 0 in (23) stands for the load of the proposed ASD. Take notice that the leakage impedance Z lk-T2 of single-phase isolation transformer T 2 is neglected as it is relatively small (less than 7% of the output inductance) in our case.
Since the converted voltage and impedance is in parallel with the line-to-ground voltage u Q series connected with the leakage impedance Z lk-T1 of zig-zag grounding transformer T 1 . Assume the two transformers are completely under control, and then the distribution network model can be converted to the low-voltage side of the single-phase isolation transformer T 2 . According to  Thevenin theorem, we can obtain an equivalent voltage E eq and an equivalent impedance Z eq from the distribution network with the influence of leakage impedance from T 1.

B. Voltage Control Method
In order to guarantee that the neutral-to-ground voltage (U N ) can track the inverse of the line-to-neutral voltage of faulty phase, a typical dual-loop control method is employed in the proposed hybrid ASD. No need to measure the zero-sequence admittance, capacitor current and so forth, the feedback controls the amplitude and phase of the injected current, forcing the voltage of the fault phase to zero, compensating the phase error of neutral voltage by adjusting the neutral voltage to the expectation value, which can erase the neutral voltage error. Then, the hybrid ASD method reaches to the purpose of voltage arc suppression.
The method includes a neutral-to-ground voltage outer loop and an output inductor current inner loop and the control algorithm determines the dynamic performance of the single-phase inverter. The error of U N is then regulated by a PR regulator to generate the reference of the injected current I * , which is the control objective of the inner-control loop. Thus, the filter inductor current feedback (ICF) is adopted in the inner loop for enhancing the system tolerance ability to load disturbances, as shown in Fig. 7, where U Ns * is the opposite voltage of faulty phase supply voltage, u o is output voltage from single-phase inverter.
The voltage control diagram in Fig. 7 can be further simplified as Fig. 8, where the simplified transfer function of filter G f (s) is (sL o + r) −1 , G inv is inverter gain K inv and the transfer function of PR controller G i (s) is shown in (26), where ζ denotes damping coefficient, ω 0 is fundamental frequency, K cp is proportionality coefficient, and K cr is resonant coefficient. (26) From Fig. 8, the transfer function of the voltage control system can be attained. The stability error of U Ns (considered low frequency) near the power frequency is very subtle because of the PR controller. Hence, the transfer function of both current-open loop H i and inner loop G inner are expressed as follows: The additional function G eq to the forward path comparing to the ICF method is marked in Fig. 8.
Apparently, G eq is a first-order high pass filter with a small gain, which depends on the ground-fault resistance and distributed capacitance. Fig. 9 shows the Bode diagram of G eq when R F increases, demonstrating that the fundamental gain in lowresistance grounding is lower than the one in high-resistance. Fig. 10 presents the Bode diagram of as C S varies. When C S increases, the high-band gain decreases, which indicates that the light load has lower antiinterference ability than heavy load. Both Figs. 9 and 10 can prove the great stability of the dual-loop control under different conditions. Furthermore, the outer voltage loop of the controller is designed to achieve zero steady-state error in fundamental frequency and enough damping to compensate the ground-fault phase error of neutral voltage. In order to avoid the calculation burden of the processor, simply PI plus resonant controller is applied to achieve good control performance [20].
(29) Therefore, the transfer function of U Ns can be obtained from Fig. 8, where we can find the transfer function of the entire controller expressed as follows: Where The control parameters are tuned by using the frequency domain design method. The current loop has a cutoff frequency of 1/10 of the switching frequency f sw and the voltage loop has a cutoff frequency of 1/2 of the current loop cutoff frequency. Their phase margins are both set to 45°. These values can be used to calculate the control parameters and slight change of the parameters is necessary due to the mismatch of the theoretical and practical model.

C. Implementation
For potential practical application, it should be noted that the hybrid ASD mainly focuses on the full compensation of grounding fault current, while the recognition of faulty state is fully and precisely accomplished before the hybrid ASD takes actions [21].
The implementation flowchart of the proposed hybrid ASD is given in Fig. 11. First, when the neutral point displacement voltage exceeds 15% of nominal phase voltage, an SLG fault is confirmed and the faulty phase is identified [21]. Second, the corresponding multiterminal breakers (refers to Table I) are closed for primarily regulating the zero-sequence voltage U N . Finally, to fully compensate the voltage regulation error brought by leakage inductances of T 1 and T 2 , the active ASD then inputs current i i to the neutral point by the voltage control method  in Section III, part B. Therefore, the system can achieve the accurate arc suppression after the SLG fault.

IV. SIMULATION VERIFICATION
In order to validate the performance of the proposed hybrid ASD, the distribution network and the hybrid ASD as shown in Fig. 1 are modeled and simulated in MATLAB/Simulink. The method introduced in Section III.C is used to design the controller. The parameters of the distribution network are listed controller. The parameters of the distribution network are listed in Table II, and the controller parameters are listed in Table III. In the simulations, an SLG fault is applied to phase A of the distribution network at 0.4 s. To better compare the control effect of the passive ASD and the hybrid ASD, the passive ASD is activated at 0.5 s with the active ASD deactivated, while the active ASD is activated at 0.6 s with the passive ASD activated   Similarly, when the same fault occurs on phase A with 10 kΩ fault resistance, the simulation results are given in Fig. 13. With the passive ASD activated only, the fault phase voltage (U F ) can decrease from 8500 to 452.2 V at 0.5 s. It can be further reduced to 96.4 V after active ASD activated at 0.6 s.
Moreover, the proposed hybrid ASD is tested under bidirectional power flow conditions. The results are summarized in Table IV. It is demonstrated that the residue fault current and fault phase voltage can further decrease after the active ASD is activated.

V. EXPERIMENT VERIFICATION
To further validate the performance of the proposed hybrid ASD in a more realistic condition, a hybrid ASD prototype has been developed and tested in an emulated distribution network.
As shown in Fig. 14, 10 kV power supply system is formed by a 380 V power supply, a self-coupling transformer and a boosting transformer. The main circuit of the active device consists of a three-phase control rectifier and a single-phase full-bridge inverter with an LC type filter. The rated voltage and current of the IGBT modules of the inverter are 1200 V and 400 A, respectively. The digital signal controller TMS320F28335 from Texas Instruments is used as the controller of power electronic switches. The other parts of the experimental system are the same as in Fig. 1. The parameters of the distribution network and the hybrid ASD are the same as the simulation in MAT-LAB/Simulink in Section IV. The test scenarios in Section IV are duplicated in this experiment for comparison. Fig. 15 gives the neutral-to-ground voltage (U N ), phase A positive sequence voltage (U A ), the fault phase voltage (U F ), Fig. 15. Experiment result under a SLG fault on phase A (ZF = 10 Ω). CH1: Neutral-to-ground voltage (6000 V/div, t:0.1s/div). CH2: Positive-sequence voltage (6000 V/div, t:0.1s/div)CH3: Ground-fault voltage (280V/div, t:0.1s/div)CH4: Active ASD injected current (250A/div, t:0.1s/div). Fig. 16. Experiment result under a SLG fault on phase A (ZF = 10 kΩ). CH1: Neutral-to-ground voltage (6000 V/div, t:0.1s/div). CH2: Positive-sequence voltage (6000 V/div, t:0.1s/div)CH3: Ground-fault voltage (280V/div, t:0.1s/div)CH4: Active ASD injected current (250A/div, t:0.1s/div) and the injected current from the inverter (i i ), when the fault resistance is 10 Ω. Again, the passive ASD is first activated after the SLG fault occurs. About 0.5 s later, the active ASD is activated. As shown in Fig. 15, the fault phase voltage is reduced to 221 V when only passive ASD is activated, and further reduced to 58.5 V after the active ASD is activated. Meanwhile, the faulty residual current is reduced to 5.74 A with the hybrid ASD, and the faulty current rejection ratio is 0.97%.
When the fault resistance is 10 kΩ, since the voltage measurements are out of the range of the existing oscilloscope, additional sensors are added to capture the accurate experiment results. The experiment results are given in Figs. 16 and 17.
The faulty phase voltage and current are reduced to 123 V and 0.013 A, respectively, after the hybrid ASD is applied. The faulty current rejection ratio is 2.07%. In Fig. 16, fast dynamic response can be observed from the faulty phase voltage. According to the theoretical analysis, the voltage error after the change of capacitance is treated as the open-loop bandwidth decrease. Therefore, it is demonstrated that the proposed hybrid ASD can handle SLG faults with different ground-fault resistances.
The trajectory of phase error of neutral voltage compensation is given in Fig. 18, where four different ground-fault resistances : Active ASD injected current (250A/div, t:0.02s/div)CH3: Zero-sequence voltage (6000 V/div, t:0.02s/div)CH4: Positive-sequence voltage (6000 V/div, t:0.02s/div). (10 Ω, 100 Ω, 1 kΩ, and 10 kΩ) are compared. The SLG fault occurs at 0.1 s, and the faulty phase can be all successfully recognized within 0.06 s. During 0.16 to 0.23 s, the ASD without the assistance of single-phase VSI is activated for confirming the fault arc cannot be rekindled. After 0.23 s, the phase error is fully compensated for complete-elimination in ground-fault arc, simply by injecting the current from the VSI within few microseconds. The phase error of neutral voltage in multiple ground-fault resistances is compensated by the hybrid ASD. Eventually, all the trajectories overlap to each other, which demonstrate the feasibility of the hybrid ASD.
Note that the advantage of the ASD is presented with the compared performances in two groups and a detailed trajectory of compensation in phase error of neutral voltage in different resistance grounding faults.
Most of conventional methods are equipped with isolation transformers; the advantage of the proposed method is that it only needs to add six breakers to the conventional methods with small capacity electronic devices to achieve flexible arc suppression with full compensation of fault current. For the purpose of demonstrating great performance in hybrid ASD, conventional methods under same conditions are recorded in Table V with major performance parameters, where T dev is the total time of the process and C ap is capacity of the needed electronic device in ASD. Note that hybrid ASD has the fastest response (T dev ≤ 0.03 s) as SLG fault occurs in the distribution network, while the performance of hybrid ASD surpass the other two conventional methods, with total capacity of electronic equipment is far less than voltage-based inverter ASD. Great advantages in hybrid ASD have inhibited in Table VI. Different characteristics are listed in Table VI for comparison of different arc-suppression methods. Peterson coil has high reliability and low-control complexity, but it needs to detect the capacitive current, which is difficult to be guaranteed. Voltage-based ASD does not need to detect the capacitive current and has better arc suppression effect. However, large capacity power electronics devices suffer from low reliability and high cost. Fault-transfer switch, as a passive method, also has high reliability, however, it may cause line-to-line fault if the faulty phase is falsely identified.
As the proposed hybrid ASD is a voltage arc suppression method, it does not need to detect the capacitive current. The ground-fault current compensation is primarily done by the passive ASD, and the residual current is easily eliminated by the small capacity active ASD. Thus, the proposed hybrid ASD brings better arc suppression effect, relatively high reliability, low cost, and control complexity.

VI. CONCLUSION
This article presented a hybrid ground-fault arc suppression device, which took great advantage of the reliability of passive device and the flexibility of active device. Zero-sequence voltage regulation was primarily done by reliable passive device, and precise compensation of ground-fault residual current was realized by the active device, which obtained full compensation of SLG fault current.
Peterson coils, voltage-based ASD, fault-transfer switch and the proposed hybrid ASD were comparatively analyzed in detection complexity, reliability, control complexity, and cost and arc suppression effectiveness. It illustrated that the proposed device had comprehensive advantages of high reliability, low-cost, and low-control complexity, which could achieved full compensation of ground-fault current.