Investigation of Impulse Aging of Energy-Absorption Elements for Hybrid DC Circuit Breakers

Abstract

The state of the energy-absorption branch MOV in the hybrid DC circuit breaker (DCCB) has a very important impact on the short fault breaking operation of the circuit breaker. Therefore, it is necessary to evaluate the state of the MOV, which is also called the “sleep component”. Due to DCCB being placed indoors, the aging is mainly caused by short-circuit impulse current. Therefore, this paper mainly focuses on the study of the short-circuit impulse aging characteristics of the energy-absorption branch MOV. The dynamic simulation system of a hybrid DCCB was built to investigate the impulse of short-circuit current on the MOV of the energy-absorption branch during the circuit breaking process. Then, an accelerated impulse aging test platform was built and the accelerated impulse aging tests of the MOV were conducted. The aging characteristics of the MOV were analyzed in detail and detailed analysis was conducted of the macroscopic parameters and microstructure changes. The results indicate that the nonlinear coefficient α could be emphasized as a basis for judging the “sleeping” component states and aging degree of the hybrid DC circuit breaker, and can be expected to be applicable to MOV condition monitoring of BCCB in the future.

1. Introduction

In recent years, with the rapid development of renewable energy technology and the requirements for long-distance transmission, DC transmission technology, which is based on high-power electronics with low losses, has gained broad prospects in transport and distribution of electric power. Hybrid DC circuit breakers (DCCBs) with fast fault current cut-off functions play an important role in the DC systems as the foundation for the operation, control, and protection of DC networks [1,2,3,4].

The hybrid circuit topology, which is commonly used for hybrid DCCBs, consists of a main flow branch, a transfer branch, and an energy-absorption branch connected in parallel [5]. Among the branches of a DCCB, the function of the energy-absorption branch is to dissipate the energy stored in the system inductance through the metal oxide varistor (MOV) during the process of breaking the fault current. Under the normal operational state, the main branch remains in a long-term flow state, while the energy-absorption branch is in a “sleep” state.

However, when the DCCB has broken a short-circuit fault, due to the impulse of short-circuit current on the MOV of the energy-absorption branch, it is difficult to determine the effectiveness of the protection function of the MOV for the next short-circuit fault. Therefore, when the energy-absorption branch enters a “sleep” state again after breaking one fault, there is the hidden danger that the short-circuit fault cannot be quickly removed. So, in the “sleep” state, predicting the effectiveness of the protection function of the MOV is the key to ensuring the short-circuit breaking performance of hybrid DC circuit breakers.

The main causes that affect the protection function of the MOV include moisture and aging [6,7,8,9,10]. Moisture caused by sealing defects and environment pollution can be avoided, and a damp MOV valve sheet can continue to be used after treatment [11]. However, aging valves must be withdrawn from operation.

The factors leading to the aging of MOV valve sheets mainly include electrical stress [12] and thermal stress [13]. Impulse current waves are the primary cause of electrical stress, which results in the aging of MOV valve sheets.

Currently, studies on MOV impulse aging principally concentrate on MOVs used for long-term grid-hung operation for lightning protection, considering the damage caused to the MOV by lightning impulse waves. Huang Lin et al. conducted continuous lightning impulse tests and multiple lightning impulse tests on metal oxide lightning arresters and zinc oxide varistors to investigate the effect of lightning impulse current on their thermal impact and aging characteristics [14]. Yoshiyasu Koga et al. studied the effect of 4 μs/10 μs lightning impulse currents of 30 kA and 100 kA on the aging of surge arresters on 6.6 kV distribution lines and assessed the deterioration of MOV performance based on the reference voltage change rate of the MOV [15]. Zhang Jie et al. conducted an impulse aging test on ZnO varistors of AC filter surge arresters in UHV DC and HVDC transmission systems under 15/35 μs impulse currents with different current amplitudes [16].

Recently, increasing attention has been paid to the issues of MOV aging in hybrid DCCBs. For example, Maike Bröker et al. conducted temperature and impulse energy injection endurance tests on the energy-absorbing element MOV in DCCBs to investigate its ability to withstand the energy pressure generated during the breaker switching operation, and the effect on MOV aging was determined by measuring the leakage current and clamp voltage [17]. However, the research on the aging characteristics of MOVs used in hybrid DCCBs under the specific impulse current wave generated during the short-circuit fault is missing in the current literature.

Impulse aging is normally a gradual process, and thermal stress is the primary cause of deterioration of the MOV valve sheet under impulse current [18]. This paper investigates the aging characteristics of the MOV under short-circuit impulse current by conducting accelerated impulse aging tests on MOV valve sheets placed indoors. Firstly, the dynamic simulation system of a hybrid DCCB was built according to the principle of similarity and per-unit value equality to investigate the short-circuit breaking process of the hybrid DCCB. Secondly, the impulse current waveform for testing was determined based on simulation results. An impulse aging test platform was designed and constructed on the basis of the parameters of the impulse current waveform to carry out accelerated impulse aging tests. Finally, the aging characteristics of the MOV under short-circuit impulse current were analyzed based on the macroscopic parameters, and the impulse aging effects on the MOV are analyzed based on examining the microstructure using a microscope.

2. Simulation Analysis of the Working Characteristics of MOVs during Short Circuit

In order to study the impulse of short-circuit current on the MOV of the energy-absorption branch during the short-circuit fault breaking process, a dynamic simulation model was designed to investigate the operation process of hybrid DCCBs. MATLAB/Simulink was used to build this model [19]. The schematic diagram of the dynamic simulation model for the hybrid DCCB is shown in Figure 1, the construction of which is based on the topology of a diode full-bridge circuit [1,13]. The simulation model mainly consists of three branches connected in parallel, namely, a main branch with a fast mechanical switch (FMS) and an auxiliary sub-module (ASM), a transfer branch mainly composed of a sub-module (SM), and an energy-absorption branch mainly consisting of a connected metal oxide varistor (MOV). Both the SM and ASM consist of an insulated-gate bipolar transistor (IGBT), an RC buffer circuit, and diode full-bridge circuit. Additionally, a DC voltage source, E, is used in the model to simulate the DC system voltage. An inductor, L, and a resistance, R_L, are used in the model to represent all inductive components and all loads of the DC system, respectively. When closing S at t₀, the short-circuit fault in the simulation model starts. The circuit parameters of the simulation model are shown in

Table 1. The circuit parameters of the simulation model.

Parameter	Value	Parameter	Value
Load Resistance (RL)	250 Ω	Buffer capacitor of SM/ASM (Cs)	15 μF
DC system inductor (L)	100 mH	Buffer resistor of SM/ASM (Rs)	200 Ω
Start time of short curcuit fault (t0)	0.02 s	Forward voltage of IGBT (Vf)	2.7 V

.

The MOV in the simulation model is a non-linear resistor, and its nonlinear volt–ampere characteristics can be expressed as follows:

$$\frac{U}{U_{ref}}=k_i{\left(\frac{I}{I_{ref}}\right)}^{1/{\alpha }_i}$$

(1)

where U_ref is the reference voltage of the MOV, I_ref is the reference voltage of the MOV, k_i and α_i are the coefficient of the ith section of volt–ampere characteristics curve. i = 1, 2, 3 represent a small current region, nonlinear region, and saturation region, respectively. The parameters of the MOV set in the model are shown in

Table 2. The parameters of the MOV.

Uref (V)	Iref (A)	(k1 and α1)	(k2 and α2)	(k3 and α3)
8000	500	(0.995, 50)	(1.0, 25)	(0.9915, 16.5)

.

Based on the design principle of per-unit value equality [19,20], the dynamic simulation model with rated value 5 kV/20 A is used to simulate a practical 500 kV/2 kA DC transmission system. The ratio of voltage and current is set to 1/100. The system parameters of the simulation model and the corresponding real system are shown in

Table 3. The system parameters of the simulation model and the corresponding real system.

	DC Voltage (kV)	DC Current (kA)	Peak of Fault Current (kA)	Rising Rate of Fault Current (A/ms)
Real system	500	2	15	5000
Simulation model	5	0.02	0.15	50
Ratio	0.01	0.01	0.01	0.01

. The short-circuit fault happened to the simulation system at t = 0.02 s, and the threshold of the action current is set to 30 A.

The simulation waveform of the whole process of breaking the short-circuit fault current of the hybrid DCCB is shown in Figure 2.

At t₁ = 20 ms, a ground short-circuit fault occurs on the load side of the DC system and the fault current rises rapidly. When the fault current reaches the pre-set action threshold of I_th = 30 A, the hybrid DCCB starts to operate, triggering the SM and shutting down the ASM, and the fault current is instantaneously switched from the main branch to the transfer branch. Subsequently, when the FMS is broken to the rated distance after Δt₁ = 2 ms, the SM is turned off, and the fault current is transferred to the energy-absorbing branch. The process can be divided into two stages. The current is transferred to the SM buffer capacitor first, and the capacitor voltage rises to the MOV motion threshold quickly. After that, the capacitor current is transferred into the MOV, which starts to absorb energy. At t₂ = 28 ms, the fault current drops to the rated value, I_rate = 20 A, at the moment the fault is considered to be removed, following a rapid drop in the fault current to 0 A, and the whole breaking process is finished. The maximum breaking current I_max in the whole opening process is about 143.5 A. Based on the simulation result shown in Figure 2, the waveform of the impulse current flow of the MOV is obtained, which will be used to calculated the circuit parameters of the impulse aging test platform in the Section 3.

3. Research on Impulse Aging Characteristics of MOVs

3.1. Construction of MOV Impulse Aging Test Platform

The schematic diagram of the test platform is illustrated in Figure 3. The impulse test platform mainly relies on the lightning impulse wave A-wave module that was independently developed by the Institute of Power Electronics Specialized Equipment of Xi’an Jiaotong University.

The specific process for impulse aging test is as follows. At the beginning, the capacitor bank C is charged by the DC charging device. When the charging voltage of the capacitor bank C reaches the set value, the trigger impulse control signal is output, and the discharge gap G breaks down instantly to connect the discharge circuit. Then, the capacitor bank C discharges to the test sample MOV through the wave-regulating resistor R and wave-regulating inductance L, forming the impulse current waveform that we need.

The impulse current waveform is measured by a Rogowski Coil, while the residual pressure of the MOV valve sheet is measured by the voltage divider. The impulse current waveform and the residual voltage waveform are recorded by the Tektronix DPO 3012 oscilloscope. The static parameters and leakage current I_L of the MOV valve sheet are measured using the MOV-III zinc oxide resistor sheet parameter tester developed independently by the Institute of Power Electronics Specialized Equipment of Xi’an Jiaotong University.

The parameters of the impulse current waveform and the elements in the current generator circuit are designed according to the simulation results. The short impulse current waveform flowing through the MOV, which is magnified from Figure 2, is illustrated in Figure 4.

It can be seen from Figure 4 that the maximum current I_MOV-max flowing on the MOV is 136 A. The wave front time of impulse current T₁ = 135 µs, and the half-peak time of the impulse wave tail T₂ = 2970 µs, and T₂/T₁ is 22. So, the impulse current generator is set to generate an impulse current with the amplitude of 150 A (i_m) and waveform of 135 µs/2970 µs for the impulse aging test.

According to the “Pulse Current Technology” published by Xi’an Jiaotong University, the normalized solution of the impulse current waveform is calculated as follows: the damping coefficient of the impulse current circuit ζ = 4.2, normalized wavefront time T₁* = 0.288, and normalized current i_m* = 0.12. The circuit parameters (capacitance C, inductance L, total circuit resistance R, and charging voltage U₀ of capacitance C) can be calculated through Formulas (2)–(4).

$$i_m^{\ast }=i_m/\frac{U_0}{\sqrt{L/C}}$$

(2)

$$T_1^{\ast }=T_1/\sqrt{L/C}$$

(3)

$$\xi =\frac{R}{2\sqrt{L/C}}$$

(4)

The calculated parameters of the impulse current generation circuit are shown in

Table 4. Partial parameters in impulse current test platform.

Parameter	Value
C (µF)	52
L (µH)	4225.5
R (Ω)	73.9
U0 (V)	11,268

.

We conducted the impulse test based on the designed circuit and obtained the impulse current waveform, residual voltage waveform, and absorbed energy waveform, which are displayed in Figure 5.

It can be seen that the peak of the impulse current applied to the test samples is 150 A, with a wave front time T₁ = 148 μs and a half-peak time T₂ = 2800 μs. The residual voltage is 7.84 kV, and the peak of absorbed energy is 4296 J (the test result is related to the selected test sample and the parameters of the impulse current generation circuit).

3.2. Accelerated Impulse Aging Test of MOVs

We explored the impulse aging characteristic of MOVs by increasing the impulse current amplitude and changing the number of impulses per round. The changes in the parameters of the MOV were continuously recorded throughout the whole test. Due to the fact that the DCCB only acts when a fault occurs, it is mostly in a sleep state; thus, the impact of temperature on the sample during the impulse aging test was ignored in this study. The process for the aging test is displayed in Figure 6.

The MOV valve sheets produced by the same manufacturer with similar initial static parameters were selected for the test. The test samples are labeled SD18/1411B, and the size is Φ32 mm × 31 mm, as shown in Figure 7.

After cleaning and drying the MOV sample with anhydrous ethanol alcohol, the following tests were performed to explore the short-circuit impulse aging characteristics of the MOV valve sheets. Four current levels were determined for the test, which were 150 A, 175 A, 200 A, and 300 A.

• Conduct multiple rounds of impulse tests with the impulse current amplitude of 300 A and one impulse per round;
• Conduct multiple rounds of impulse tests with varying impulse current amplitudes (150 A, 175 A, 200 A) and ten impulses per round;
• Conduct multiple rounds of impulse tests with an impulse current amplitude of 200 A and varying numbers of impulses per round (10, 15, 20).

After each round of impulse, the MOV sample was cooled to room temperature before the next round of impulse.

The aging degree of the MOV is generally determined according to the “Operating Surge Protector Testing Standards” published by Chinese standards press, and the MOV valve sheet is considered to be aged if the fluctuations of the U1 mA is more than 10% or leakage current IL is more than 20 μA. After each group of tests, we need to determine if the test sample has been damaged.

If there was no damage to the test sample, the static parameters of the MOV in the low-current region were measured, which include voltage values U_1mA, U₀._1mA, and U_0.01mA, to explore the aging degree of the MOV during the impulse test. We also explored the aging degree of the MOV using some other factors, including the residual voltage ratio K, the leakage current I_L (0.75 U_1mA), and the nonlinear coefficient α. If the test sample did not experience aging, we continued to perform the impulse aging test on the sample.

If the test samples were damaged, their states were observed and recorded.

3.3. Result and Discussion

By analyzing the relationship curves between the change rate of static parameters, leakage current value I_L, residual voltage ratio K, nonlinear coefficient α, and the number of impulses n, the impact of impulse current on the aging degree of the MOV valve sheets is explored.

The value of α is usually calculated from the measured voltages U₁ and U₂ at two determined currents I₁ and I₂ using Equation (5).

$$\alpha =\frac{\lg I_1-\lg I_2}{\lg U_1{}_{}-\lg U_2}=\frac{\lg (I_1/I_2)}{\lg (U_1/U_2)}$$

(5)

When I₁ and I₂ are set to 1.0 mA and 0.1 mA at room temperature, Equation (5) can be reduced to Equation (6):

$$\alpha =1/[\lg (U_{1mA}/U_{0.1mA})]$$

(6)

The residual voltage ratio K is calculated by Equation (7):

$$K=U_{res}/U_{1mA}$$

(7)

where U_res is the residual voltage.

Three tests were carried out as follows:

Test 1: The amplitude of the impulse current is 300 A, one impulse per round (300 A-1 group, impulse current amplitude, the number of impulses per round).

Three MOV samples with similar initial parameters were selected. Multiple rounds of pulsed current with an amplitude of 300 A were applied to the MOV samples, one impulse per round. The DC static parameters and leakage current I_L of the MOV in the low-current region were measured and recorded after each round of impulse. The change rate of DC static parameters, residual voltage ratio K, and nonlinearity coefficient α were calculated.

It can be seen that the statistics of the three MOV samples showed similar trends in each parameter; thus, the statistics of one of them were selected to plot the relationship curves between the change rate of DC static parameters, leakage current I_L, residual voltage ratio K, nonlinear coefficient α, and the number of impulses n, as shown in Figure 8a–d.

It can be seen that test samples were damaged after 44 impulses; therefore, the statistics during the first 44 impulses were recorded for analysis. From Figure 8a, as the number of impulses increases, U_1mA shows a trend of small increase, U₀._1mA has a trend of small increase followed by a decrease, and U_0.01mA shows a trend of decrease. The initial values of each parameter are as follows: U_1mA = 6.103 kV, U₀._1mA = 5.740 kV, U_0.01mA = 5.230 kV, I_L = 1.67 μA, K = 1.34, and α = 37.5. After the 44th impulse, it can be found that U_1mA increased by 2.95%, and U₀._1mA fluctuated by 0.52%, and U_0.01mA decreased by 4.94% from the initial value.

Figure 8b shows that before the damage occurs, the leakage current I_L exhibits an increasing trend, reaching a value of 6.90 μA. The change rate of voltage U_1mA remains within ±10% and the leakage current value I_L does not exceed 20 μA prior to the damage of the MOV valve sheet; this indicates that U_1mA and I_L have a delay in judging the aging degree of the MOV, so that the aging cannot be determined in the initial period. After the MOV is damaged, the voltage value in the small current area rapidly decreases to nearly 0 A, and the MOV is no longer insulated. The leakage current I_L rapidly increases to 249.8 µA.

Figure 8c indicates that the residual voltage ratio K shows a trend of decreasing before damage occurs. Since the residual pressure ratio K is jointly determined by U_res and U_1mA, the two parameters need to be studied simultaneously. Firstly, the change in residual voltage is analyzed. The initial value of U_res of the test sample is 8.16 kV. The U_res changed 40 V after the 44th impulse with the change rate of 0.49%, which can be regarded as basically unchanged. The V-I characteristics in the large current region are determined by the grain characteristics, and U_res is basically unchanged during the aging process, so that the grain resistance value is also basically unchanged. This indicates that the change in the residual voltage ratio K is mainly caused by the change in the grain boundary barrier. According to the double Schottky barrier theory [21], it is generally believed that the initial aging of the MOV is caused by the double Schottky barrier distortion in the grain boundary layer. At the beginning of the impulse, the trap in the grain boundary captures the injected electrons (electron trapping process), so that they cannot go to the anti-bias side, which leads to an increase in the anti-bias-side Schottky barrier holes and reduces the height of the Fermi energy level. This causes an increase in the height of the anti-bias-side grain boundary barrier, which leads to the increase in the energy required by the electron leap barrier. From the macro-perspective, there is a small increase in U_1mA, while U_res remained basically unchanged, resulting in a decrease in the residual voltage ratio K.

As shown in Figure 8d, the nonlinear coefficient α shows a trend of linearly decreasing with the increase in the number of impulses, and decreased from 37.5 to 27.0. Generally, the nonlinear coefficient α of the MOV ranges from 30 to 100 [22]. At about the 23rd current impulse, the value of α started to be less than 30. This suggested that the MOV samples entered the early stage of aging. This indicates that the nonlinear coefficient α can more accurately characterize the deterioration and aging degree of the MOV compared with other parameters.

Test 2: The amplitude of the impulse current is 150 A, 175 A, and 200 A, ten impulses per round.

Nine MOV samples with similar initial parameters were selected and divided into three groups (group 150 A-10, group 175 A-10, and group 200 A-10), with three samples in each group.

Multiple rounds of impulse current with an amplitude of 150 A, 175 A, and 200 A generated by an impulse current generator were applied to each MOV sample in the three groups, with ten impulses per round.

The DC static parameters of the MOV were measured after each round of impulses, and the main reference index values were recorded. The voltage change rate in the small current region, the residual voltage ratio K, and the nonlinear coefficient α were calculated. Since the recorded data trends for each parameter of the three groups of samples were similar, one sample was selected from each group to plot the relationship curves between the change rate of voltage in the small current region, leakage current I_L, residual voltage ratio K, nonlinear coefficient α, and the number of impulses n under different impulse current amplitudes, as shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9a–d.

Regarding the 150 A-10 group of samples, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9a show that test samples were damaged after 430 impulses, and the trends in U_1mA, U₀._1mA, and U_0.01mA were the same as those in the 300 A-1 group. After the 430th impulse, U_1mA increased by 3.83% compared to its initial value, U₀._1mA decreased by 0.86% compared to its initial value, and U_0.01mA decreased by 7.56% compared to its initial value.

As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9b, the leakage current I_L showed a rising trend and its value increased to 10.60 μA before the test samples were damaged.

Regarding the 175 A-10 group of samples, it can be seen from Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9a that test samples were damaged after 590 impulses. During the 590 impulses before the samples were damaged, U_1mA first slightly increased and then showed a decreasing trend, with its value increasing by a maximum of 5.77% compared to its initial value. U₀._1mA firstly increased and then had a decreasing trend, with its value decreasing by 2.07% compared to its initial value. U_0.01mA voltage changed with a decreased trend, with its value decreased by 13.61% compared to its initial value. As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9b, the leakage current I_L showed a rising trend and its value increased to 17.90 µA before the test samples were damaged.

Regarding the 200 A-10 group of samples, it can be seen from Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9a that test samples were damaged after 620 impulses. During the 620 impulses before the samples were damaged, U_1mA first slightly increased and then showed a decreasing trend, with its value increasing by a maximum of 2.94% compared to its initial value. U₀._1mA had a decreasing trend, with its value decreasing by 5.86% compared to its initial value. U_0.01mA voltage changed with a decreasing trend, with its value decreasing by 13.50% compared to its initial value. As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9b, the leakage current I_L showed a rising trend and its value increased to 18.67 µA before the test samples were damaged. The change rate of the U_1mA of the three test samples in groups 150 A-10, 175 A-10, and 200 A-10 did not exceed ±10%, and I_L also did not exceed 20 μA. After the samples were damaged, the voltage value in the low-current region rapidly decreased to nearly 0 A, and the MOV valve sheet no longer had insulation performance. The leakage current I_L value rapidly increased to 249.8 μA.

From Figure 9c, it can be seen that the residual voltage ratio K of the three groups of samples showed a trend of first decreasing and then increasing with the number of impulses n. Before applying the 30th, 220th, and 20th impulses, the residual voltage ratio K of the three groups of samples showed a decreasing trend, and, after this, the residual voltage ratio of each group showed a slight increase. During the process of applying the 30th, 220th, and 20th impulses to the three groups of samples (the initial stage of impulse aging), due to the electron trap process, U_1mA increased slightly macroscopically, while U_res remained basically unchanged, resulting in a decrease in the residual voltage ratio K, which can still be explained by the dual Schottky barrier distortion.

From Figure 9d, it can be seen that the nonlinear coefficient α of the three groups of samples showed a trend of rapid decrease followed by a stable decrease with the increase in the number of impulses. During the impulse process, when the value of the nonlinear coefficient α is less than 30, it can be regarded as starting to age under the impulse current. In the whole impulse process, for the 150 A-10 group, the nonlinear coefficient α decreased rapidly from 40.5 to 28.8 in the first 60 impulses, and then decreased slowly from 28.8 to 22.3 from the 60th to 430th impulses. For the 175 A-10 group, the nonlinear coefficient α decreased rapidly from 43.4 to 29.4 in the first 30 impulses, and then decreased slowly from 29.4 to 19.2 from the 30th to 590th impulses. For the 200 A-10 group, the nonlinear coefficient α decreased rapidly from 42.9 to 29.8 in the first 30 impulses, and then decreased slowly from 29.8 to 18.0 from the 30th to 620th impulses. Therefore, the 150 A-10, 175 A-10, and 200 A-10 groups can be regarded as starting to age under the 60th, 30th, and 30th impulses, respectively. By comparing the changes in the nonlinear coefficient α of the samples under the three impulse conditions, it can be found that the larger the amplitude of the impulse current, the fewer impulses are needed to make α less than 30, and the faster the aging speed. At the same time, when α is less than 30, the change rate of U_1mA, the change rate of U₀._1mA, the change rate of U_0.01mA, and I_L of the three groups of samples were 3.23%, 0.87%, −2.82%, and 5.4 µA; 3.21%, 0.62%, −3.2%, and 4.8 µA; and 1.96%, −0.40%, −1.75%, and 5.37 µA, respectively. None of these parameters reached the commonly used aging judgment criteria, especially after the samples were damaged, and none of the parameters reached the standard values. Therefore, conventional methods for judging the aging and impulse aging degree of MOVs have obvious delays.

Test 3: Impulse current amplitude of 200 A and the number of impulses per round is 10, 15, and 20.

Nine MOV samples with similar initial parameters were selected and divided into three groups (200 A-10, 200 A-15, 200 A-20). Each group contained three samples. Multiple rounds of impulse aging were conducted on each test sample of the 200 A-10 group, 200 A-15 group, and 200 A-20 group using an impulse current generator that produces the impulse current with an amplitude of 200 A. The DC static parameters of the test samples were tested after each round of impulse by changing the number of impulses, and the U_1mA, U₀._1mA, and U_0.01mA, and I_L of the test samples were recorded. The change rate of static parameters in the low-current region, residual voltage ratio K, and non-linearity coefficient α were calculated.

Since the recorded data trends for each parameter of the three groups of samples were similar, the data of one sample from each group was selected to plot the relationship curves between the change rate of DC static parameters in the small current region, leakage current I_L, residual voltage ratio K, nonlinear coefficient α, and the number of impulses n under different impulse times per round, as shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10a–d.

The initial values of each parameter were as follows: U_0.01mA = 5.133 kV, I_L = 2.47 μA, K = 1.31, α = 42.9. It can be seen from Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10a that the test samples of the group of 200 A-10 were damaged after 620 impulses. During the 620 impulses before the samples were damaged, U_1mA first slightly increased and then showed a decreasing trend, with its value increasing by a maximum of 2.94% compared to its initial value. U₀._1mA had a decreasing trend, with its value decreasing by 5.86% compared to its initial value. U_0.01mA voltage changed with a decreasing trend, with its value decreasing by 13.50% compared to its initial value. As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10b, the leakage current I_L showed a rising trend and its value increased to 18.67 µA before the test samples were damaged.

It can be seen from Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10a that test samples of the group of 200–15 were damaged after 405 impulses. During the 405 impulses before the samples were damaged, U_1mA first slightly increased and then showed a decreasing trend, with its value increasing by a maximum of 4.61% compared to its initial value. U₀._1mA had a decreasing trend, with its value decreasing by 2.34% compared to its initial value. U_0.01mA changed with a decreasing trend, with its value decreasing by 9.92% compared to its initial value. As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10b, the leakage current I_L showed a rising trend and its value increased to 15.93 µA before the test samples were damaged.

It can be seen from Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10a that test samples of the group of 200 A-20 were damaged after 460 impulses. During the 405 impulses before the samples were damaged, U_1mA first slightly increased and then showed a decreasing trend, with its value increasing by a maximum of 5.60% compared to its initial value. U₀._1mA had a decreasing trend, with its value decreasing by 2.99% compared to its initial value. U_0.01mA changed with a decreasing trend, with its value decreasing by 14.37% compared to its initial value. As shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10b, the leakage current I_L showed a rising trend and its value increased to 20.80 µA before the test samples were damaged.

For the 200 A-10 and 200 A-15 groups, the change rate of U_1mA did not exceed ±10%, and the I_L did not exceed 20 μA. After the damage of the test sample, the voltage in the low-current region rapidly decreased to nearly 0 V, indicating that the MOV valve no longer had insulation performance. The leakage current I_L increased rapidly to 249.8 μA. The I_L of the 200 A-20 group exceeded 20 μA in the first 460 impulses, and can be judged as aging.

From Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10c, it can be seen that the residual voltage ratio K of the three groups of samples showed a trend of firstly decreasing and then increasing as the number of impulses increased.

From Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10d, it can be seen that the nonlinear coefficient α of the three groups of samples exhibited a rapid decrease followed by a steady decline with the increase in the number of impulses. During the whole impulse process, for the 200 A-10 group samples, the nonlinear coefficient α decreased rapidly from 42.9 to 29.8 in the first 30 impulses, and then decreased gradually to 18.0 from the 30th to 620th impulses. For the 200 A-15 group samples, the nonlinear coefficient α decreased rapidly from 37.7 to 27.3 in the first 30 impulses, and then decreased gradually to 18.6 from the 30th to 405th impulses. For the 200 A-20 group samples, the nonlinear coefficient α decreased rapidly from 41.6 to 28.3 in the first 40 impulses, and then decreased gradually to 17.7. Therefore, the 200 A-10 group, 200 A-15 group, and 200 A-20 group can all be considered to begin aging after the 30th impulse. By comparing the changes in the nonlinear coefficient α under the three different impulse conditions, it can be observed that as the number of impulse increases, the required number of impulses for α to be less than 30 did not change significantly. This is because impulse aging is mainly thermal aging, and if the amplitude of the impulse current did not change (the same amount of energy is injected into each impulse), as the number of impulse increases, the temperature of the MOV valve plate increased to a certain extent and no longer increased, which meant that the thermal aging degree of the MOV valve plate for each impulse cycle was not significantly different, and was more affected by the quality of the MOV valve plate itself. At the same time, when α was less than 30, the change rate of U_1mA, the change rate of U₀._1mA, the change rate of U_0.01mA, and I_L of the three groups of samples were: 1.96%, −0.40%, −1.75%, 5.37 µA; 2.64%, 0.29%, −2.74%, 6.03 µA; and 3.59%, 0.92%, −3.36%, 5.73 µA, respectively. The parameters of each group of samples did not reach the commonly used aging judgment standard value, except for the 200 A-20 group sample, which did not experience damage until the leakage current I_L exceeded 20 µA.

The states of the samples after the impulse test are summarized in

Table 5. The states of the sample under different test conditions.

Test Groups	Sample 1	Sample 2	Sample 3
300 A-1	Insulation breakdown	Body breakdown	Insulation breakdown
150 A-10	Body breakdown	Insulation breakdown	Insulation breakdown
175 A-10	no damage	Insulation breakdown	Body breakdown
200 A-10	Body breakdown	Insulation breakdown	Insulation breakdown
200 A-15	Body breakdown	Insulation breakdown	Body breakdown
200 A-20	Insulation breakdown	no damage	Insulation breakdown

. The failure modes of MOV valve sheets under impulse aging can be classified into three kinds, namely, “no damage”, “insulation breakdown”, and “body breakdown”.

Among these, “no damage” indicates that there was no visible damage to the MOV when the test was completed; “Insulation breakdown” indicates that there were obvious burn-through channels or breakdown holes in the side insulation coating. “Body breakdown” indicates that the MOV body was broken down and cracked, and there were obvious traces of current flow and breakdown on the cracked surface.

Under the same test conditions, the failure modes of the MOV valve sheet were different. This was due to the fact that the quality and structure of each MOV valve sheet could not be exactly the same, so each MOV valve plate had different structural defects and different load-bearing capacities.

According to Table 5, it can be seen that body breakdown and insulation breakdown were the two most common fault modes in MOV impulse aging tests, accounting for 33.3% and 55.6% of the total experiment, respectively.

Insulation breakdown usually occurs at the interface between the edge of the MOV sintered body and the insulation coating. Initially, it only formed non-penetrated impulse holes, and then gradually developed into penetrated holes, which formed erosion marks as the insulation layer was destroyed due to energy release. The body breakdown could be observed on the surface of the MOV sintered body, with some puncture holes due to thermal expansion caused by thermal stress.

The pictures of failure modes of insulation breakdown and body breakdown are shown in Figure 11.

4. Microstructure Testing and Analysis of MOV

The main focus of the previous content on the aging characteristics of MOV is the changes in the macroscopic electrical properties of MOV before and after impulse aging.

Since the nonlinear characteristics of MOV originate from its microstructure, the impulse aging characteristics from the internal microstructure of MOV valve sheets are studied in this section using the VE-9800S scanning electron microscope of the State key laboratory of electrical insulation and power equipment of Xi’an Jiaotong University.

The MOV valve sheet is made by sintering different metal oxides mixed by the solid-state ball milling method, and its main component is zinc oxide (ZnO) [23]. The basic structure unit of the MOV valve sheet is the ZnO grain, which also includes the grain boundary wrapped around the ZnO grain, the spinel dispersed within the grain boundary, and the pores distributed within the ZnO grain and the grain boundary [24].

The specific experimental steps for the morphological scanning of the MOV valve sheet using a scanning electron microscope are as follows:

• Take a small piece of relatively flat cross-section from the MOV sample without impulse aging, and smash a small piece of the section from the damaged surface of an MOV sample that has undergone impulse aging tests;
• Smooth the other surfaces of the sample to facilitate smooth placement in the instrument for observation;
• Observe the microstructure morphology of the sample at appropriate magnification.

The average grain size of the MOV sample can be obtained by the line intercept method, so that the impulse aging characteristics of the MOV valve sheet can be analyzed from the microscopic perspective according to the variation in grain size. The average size of grain can be calculated using Equation (8):

$$\overline{D}=1.56\frac{C}{MN}$$

(8)

where D is the average size of the grains, C is the total length of the line segments for testing, M is the magnification of the image, and N is the number of lines for the test. The microstructure of the selected MOV sample without impulse aging is shown in Figure 12.

As shown in Figure 12a, the surface morphology of the MOV valve sheet can be observed from the overall perspective. After magnifying it 50 times, it can be found that the cross-section is relatively flat, where the black dots represent the pores, and it can be seen that the distribution of pores in the new MOV valve sheet is more uniform and the pore size is small. The gray area in Figure 12a represents the sintered MOV valve sheet. The ZnO grain structure can be clearly observed after further magnifying its image 1000 times. As shown in Figure 12b, the black porous parts and the irregular polygons represent the pores and the individual ZnO grains, respectively, and the distributions of both are relatively uniform. The grains are intact and full, except for some fractures caused by stress during sampling. The bright line between two ZnO grains is the grain boundary, which is clearly visible. According to Equation (7), the grain size of the sample is calculated to be 14.69 μm.

The pictures of the microstructure of the samples under the three impulse aging tests mentioned above are as follows.

Test 1: The amplitude of the impulse current is 300 A, one impulse per round.

The pictures of the impulse damage area of the 300 A-10 group of MOV valve sheets under SEM are displayed in Figure 13.

Figure 13a shows the image of the microstructure of the impulse break cross-section with a clear damaged area after magnifying it 50 times. Figure 13b shows the microstructure of the yellow area in Figure 13a at 200 times magnification. Figure 13c shows the microstructure of the yellow area in Figure 13b at 2000 times magnification. In Figure 13a, the breakdown pores caused by impulse damage can be clearly seen in the blue line box, and the ablation marks caused by heat release can be seen on the whole section. Comparing Figure 13a with Figure 13b, it can be found that the pores were densely distributed on the impulse aperture, and some of the pores on the aperture were connected to form a larger pore. Comparing Figure 12a with Figure 13a, it can be found that the number of pores was greatly increased after impulse. From Figure 13c, it can be seen that the crack penetrated the pores, which may be one of the reasons for the cracking of the sample after breakdown. It can be found that the longer the length of the crack, the more pores it will penetrate, and the easier it can crack. It can also be observed from Figure 13 that the denser the pores on the breakdown aperture, the easier it can crack along the aperture. The cause of the above phenomenon is that under the impulse of high current, the current easily flows through the area with a thin grain boundary of the MOV valve sheet, such as the breakdown aperture that has low resistance, resulting in uneven temperature distribution in different regions. The grain boundary melts when the temperature reaches a certain value, leading to the perforation and damage of the MOV valve sheet. The uneven temperature also causes thermal expansion stress, which generates pores and cracks during stress release, and the MOV valve sheet breaks if its tensile stress exceeds its resistance. As shown in Figure 13c, ZnO grains melted and expanded under heat stress, resulting in a radial pattern after exploding, with white noise representing electrostatic accumulation that cannot be completely eliminated.

Figure 14 shows the microstructure of the damaged surface of the test sample at a magnification of 1000 times, and no visibly damaged area was observed. Regarding the non-impulse damage area, the grain size of the sample was calculated to be 8.79 μm according to Equations (3)–(5). Compared with Figure 12b, the grain size of the impulse area was significantly reduced, and the integrity of the grains was low, with most of them being damaged, which was caused by heat stress. As shown in Figure 14, the internal area of the selected section contains clastic substances, which may be caused during sampling, by the products of ion migration during impulse aging, or by debris from grain damage, and further analysis of its composition is needed.

Test 2: The amplitude of the impulse current is 150 A, 175 A, and 200 A respectively, ten impulses per round.

After ten rounds of impulse aging tests, the picture of the impulse damage area of the 150 A-10 group of MOV valve sheets under SEM is displayed in Figure 15. When the damaged area was magnified 50 times, the visible breakdown holes and ablation marks caused by heat release were revealed, with cracks running through the entire aperture. When magnified 200 times, the same phenomenon as that of the 300 A-1 group damage area magnified 200 times was observed. When magnified 2000 times, it was found that ZnO grains lost their original shape after being heated and melted. The picture of the non-visibly damaged area of MOV valve sheets under SEM is displayed in Figure 16; it can be seen that there was a significant increase in internal pores compared to before the impulse, and a decrease in grain size and a higher integrity of grain.

After ten rounds of impulse aging tests, there was no visible damage on the group of 175 A-10 test samples. Compared with the valve sheet that did not undergo impulse aging, a small piece of the sample was taken and observed under the microscope. The pictures of the impulse damage area of the 175 A-10 group of MOV valve sheets under SEM are displayed in Figure 17. It was found that there was no significant change in the number of internal pores compared to before the impulse, the grain size was slightly reduced with lower integrity, and the grains were damaged under impulse.

After ten rounds of impulse aging tests, the pictures of the impulse damage area of the 200 A-10 group of MOV valve sheets under SEM are displayed in Figure 18. When noticeable damage area was magnified 50 times, the clear breakdown pores and burn marks caused by heat release were revealed. It can be found that the impulse damage occurred between the MOV sintered body and the insulation layer, and the number of pores on the breakdown aperture was much larger than the number of pores on the sintered body of the MOV. When magnified 200 times, multiple pores were found to be connected to form larger ones. When magnified 2000 times, it was found that the ZnO grains lost their original shape due to melting after being heated. The picture of the non-visibly damaged area of MOV valve sheets under SEM is displayed in Figure 19; it was observed that the internal pores were significantly enlarged compared to before the impulse. It can also be found that the grain size was reduced, and the grain was damaged with low integrity after the impulse.

Test 3: Impulse current amplitude of 200 A and the number of impulses per round is 10, 15 and 20.

The state of the 200 A-10 test samples was analyzed in test 2.

After fifteen rounds of impulse aging tests, the pictures of impulse damage area of the 200 A-15 group of MOV valve sheets under SEM are displayed in Figure 20. When test samples were magnified 50 times, clear breakdown pores and ablation marks caused by heat release on the entire surface were revealed. The areas magnified 200 times revealed pores distributed densely on the aperture, with some pores connected to form larger ones, and cracks running through the pores. Compared to valve sheets that did not experience impulse aging, the number of pores in the areas receiving an impulse increased significantly. Magnifying the areas to 2000 times showed that ZnO grains melted and expanded under thermal stress, resulting in a radial pattern after exploding. Additionally, the pictures of the non-visibly damaged area of MOV valve sheets under SEM are displayed in Figure 21, and it was observed that the grain size decreased and the integrity of the grains further deteriorated due to the impulse.

After the impulse aging test of 200 A per round for 20 rounds, there was no visible damage on the samples. A small piece of cross-section from the MOV sample was smashed to observe its microstructure under SEM. The pictures of the impulse damage area of the 200 A-20 group of MOV valve sheets under SEM are displayed in Figure 21. It can be found that, compared with the valve sheets that did not undergo the pulse aging test, there was no significant change in the number of internal pores, but the grain size was slightly reduced and the integrity of the grains was lower. There were small pits with a diameter of about 2 μm on the grain boundaries, where the grains were broken due to the impulse. It was observed that the extent of grain damage caused by the impulse did not change significantly with the increasing number of pulses, indicating that the effect of changes in the number of impulses per round on the thermal damage of the MOV valve sheets was relatively small when the number of impulses per round exceeded 10 under the impulse current with the amplitude of 200 A.

5. Conclusions

In this research, the fault current impulse process of the MOV of the energy-absorbing branch were simulated by a dynamic model based on the hybrid DCCB diode full-bridge topology. The accelerated impulse aging test platform for the MOV was built according to the simulation results to study the impulse aging characteristics of the MOV. The aging characteristics of the MOV from both macro- and micro-perspectives were analyzed in detail. The main contributions of this study are as follows:

• The indicators commonly used in industry, such as the change rate of U_1mA and the value of leakage current I_L, have a delay in judging the aging of the MOV, but the nonlinear coefficient α is sensitive to the initial aging state of the MOV. Therefore, the nonlinear coefficient α could be emphasized as a basis for judging the “sleeping” component states and the aging degree of the hybrid DC circuit breaker.
• The main forms of MOV valve sheet impulse damage are insulation breakdown and body breakdown. Due to the inconsistent structure of the MOV valve sheet, the breakdown fault often occurs in the edge area.
• The impulse current applied in the MOV valve sheet mainly causes grain damage. The greater the amplitude of the impulse current, the more severe the grain damage. During the impulse aging, there is little change in the grain size and grain boundary layer of the MOV valve sheet, which also explains why the residual voltage of the MOV did not change much after the impulse. At the same time, pores are generated inside the MOV valve sheet under thermal stress, and perforation damage occurs due to the unevenness of the MOV structure itself.