Successful fault current interruption on DC circuit breaker

: This study focus on the interruption capability of the DC circuit breaker employing a current commutation approach and evaluates the two main factors that determine the success rate for breaker current interruption, namely the current slope d i/ d t before current zero and the rate of rise of the transient recovery voltage d v/ d t across the mechanical breaker contacts after current zero. A vacuum circuit breaker is used to evaluate DC breaker characteristics. Detailed mathematical and graphical analysis are presented for the proposed circuit operation used in analysing the circuit breaker properties, with simulation and experimental results at fault current levels up to 330 A.


Introduction
Current interruption in a direct current (DC) system is more difficult than in an alternating current (AC) system due to the absence of a natural current zero (CZ). Since there is no large inductive device in the voltage source converter-based DC system, DC breakers have to interrupt the fault current quickly to avoid excessive overcurrent and to dissipate the stored magnetic energy without producing an excessive high voltage. Numerous proposals for introducing CZ have been presented in articles and patent applications [1][2][3][4][5][6][7][8][9]. The methodology can be divided into two groups. The first group can be addressed as the inverse voltage method where the classic mechanical AC interrupter creates arc voltage significantly in excess of the system voltage. The second group can be interpreted as employing a current commutation method where a virtual CZ is developed by utilising auxiliary means which include current oscillation and voltage commutation; where the current oscillation is accomplished by switching parallel commutation circuits to produce a counter-current through the breaker. Voltage commutation means the current commutation is achieved by introducing a voltage which exceeds the on-state voltage of the parallel commutation path.
Asea Brown Boveri Ltd (ABB) has proposed a hybrid DC breaker to fulfil high-voltage DC (HVDC) grid requirements [10,11]. This new hybrid configuration has negligible on-state power losses and provides current interruption capability within 4 ms, at 70 kV. There are two parallel branches: one branch contains a series combination of a semiconductor-based commutation switch and a fast mechanical disconnector, and the other branch is the semiconductor-based breaker that comprises several sections each with individual arrester banks. Each independent section in the breaker branch is designed for pro rata full voltage and full current breaking capability, whereas the load commutation switch branch is dimensioned for lower voltage and energy capability. The disconnecting circuit offers dielectric separation of the load after fault clearance thereby protecting the arrester banks of the hybrid HVDC breaker from thermal overload. The fast mechanical switch is opened with zero current thereby facilitating a lightweight contact system. When the mechanical switch reaches the open position the main HVDC breaker interrupts the current that has been commutated into it. The dielectric separation means the load commutation switch has a low voltage requirement. Proactive control of the hybrid HVDC breaker is utilised to compensate for the time delay of the fast disconnector. Another method [12][13][14][15]t o produce CZ in the mechanical switch involves current oscillation. In general, this topology comprises two mechanical switches; a main breaker and an isolation switch (as with the ABB hybrid breaker); the main breaker is parallelly connected to commutation and energy absorbing paths. The main breaker supports the continuous current flow; the isolation switch provides dielectric separation of the load after fault clearance thereby avoiding metal oxide varistor (MOV) thermal overload, while the solid-state switches in the commutation path only conduct during the interruption process. A series combination of a capacitance C and inductance L is incorporated into the commutation path; thus there will be an oscillating current between the main breaker and commutation path. The line current originally flowing through the main breaker is sinusoidally transferred into the commutation path. At this point, a CZ arises in the nominal path and the main breaker can interrupt with zero current. As the line current continues to flow through the LC commutation circuit, the voltage across the capacitor C charges to a voltage within the capability of the grid. At this voltage, the remaining energy stored in the line inductance is dissipated in the energy absorption path (MOV), forcing the line current to decrease. There are two current commutation modes; namely active commutation if C is pre-charged, otherwise passive commutation.
The current oscillation approach, especially active commutation, dominates the development of the hybrid HVDC breaker. Since the current oscillation involves an arc situation, the vacuum interrupter becomes the prefer building block for HVDC circuit breakers (CBs) due to its excellent insulating properties after the CZ. However, this is not the approach proposed by ABB in [10]. Rather than interruption with an arc based on current oscillation, the ABB approach is interruption without an arc voltage. Assuming interruption at same power rating, the two approaches (traditional and ABB approaches) compare as follows: † The mass of mechanical switch in the traditional approach is large compared with the ABB approach. Therefore, the ABB opening speed can be faster. † Since the first approach has to deal with an arc, the interruption performance is uncertain, unlike the ABB approach. For interruption at high power rating, the hybrid HVDC breaker proposed by ABB appears the only way to meet the requirement of HVDC grids. However, from an economical view, the hybrid HVDC breaker adopting current commutation may be more suitable for the low-and medium-voltage applications, such as electric traction, various drives and converter systems. This paper focus on the interruption capability of the DC CB under a current commutation approach and evaluates the two main factors that determine the success rate for breaker interruption, namely the current slope di/dt before CZ and the rate of rise of the transient recovery voltage dv/dt across the mechanical breaker after CZ. As such, this paper is applicable to any LC resonant approach, including the self-sustaining, multi resonant cycle approach in [12][13][14][15]. Section 2 in this paper evaluates the proposed test circuit used to investigate the vacuum CB (VCB) characteristics. Sections 3 and 4 include the simulation and experimental results that evaluate the factors that contribute to unsuccessful and successful breaker interruption.

Proposed test circuit
An accurate VCB model is complicated, since it involves many parameters including contact material, gap distance and electrodes dimension. A small change in any parameter can result in a significant performance change. There is no complete mathematic model describing all VCB properties in terms of its internal parameters. Here, the VCB will be considered as a black box, ignoring all the internal parameters, but this black box will attempt to retain the features and regular pattern of successful VCB interruptions based on external conditions. There are two specific external parameters that determine this functionality; namely di/dt and dv VCB /dt, where the former represents the rate of decrease of current through the VCB immediately before a CZ and the later represents the rate of increase of voltage across the VCB contacts immediately after a CZ. For example, in obtaining a 100% successful interruption rate with fast commutation, an increase in di/dt introduces a decreased allowable dv VCB /dt.F o rfixed di/dt, dv VCB /dt and arcing time, the interruption probability reduces as the interruption current increases. The di/dt and dv VCB /dt for successful interruption reduce with decreasing electrode spacing [16][17][18]. Thus, in order to investigate VCB characteristics, a unipolar test circuit is proposed based on a VCB with active mode commutation. (Note the test circuit is not necessarily the topology that would be used for a practical hybrid DC CBrather this proposed circuit is used in this paper to evaluate the VCB characteristics only). Fig. 1 shows the proposed unipolar test circuit and the timing intervals for commencing a fault current and subsequent breaker interruption (respective test interruption sequences are shown in Fig. 2).
There are three operating sequences during the cycle. † Pre-charging of the commutation capacitor (resetting the commutation circuit). † Creation of the fault by load switching T 4 (fault current introduction). † Commutation of the VCB by switching T 3 (VCB current interruption).
The sequential timing operations for the test circuit shown in Fig. 2 are tabulated in Table 1. . To investigate VCB properties in high-current situations, the capacitor bank (C bank ) is charged to V DC in order to supply a fault current determined by the fault resistor R FAULT , which is connected in parallel with the load resistor R LOAD by switch T 4 . When the VCB is in an on-state (zero contact gap), a low current V DC /R LOAD is supplied by V DC. The switches T 2 , T 3 and T 4 are off and T 1 is on, charging the commutation capacitor to V DC with an exponential growth via the resistor R 1 . In this interval, the equation relating currents i DC , i VCB′ and i 1 is

Pre-charging of the commutation capacitor
which has the following differential equation with the initial condition and where t 1 = R 1 C C When the current i 1 reduces to zero, the voltage across the commutation capacitor C C reaches V DC , retaining this voltage provided T 1 remains on. The with the initial condition where t 2 = L LOAD /R LOAD When (6) stabilises, the current i VCB′ through the VCB is equal to the source current i DC . With large load resistance, this 'wetting' current is small during this interval and can be neglected in the following analysis.
Although the voltage V C C across the commutation capacitor C C has been charged to the DC source, it is not ready to introduce a counter-current through the VCB due to its voltage polarity, V C C . The pre-charge cycle needs to be completed before the fault current is applied. The voltage polarity of the commutation capacitor has to be reversed before t 0 at t 0− . With T 1 off and T 2 on, the circuit loop with the initial conditions  6 off off off on on t nor represents a normal working period; t 0− is the moment immediately following the occurrence of the peak in the coil current; t 0 represents the moment that the fault current starts to arise in the circuit; t 1 represents the moment that the VCB electrodes start to separation; and t 2 −t 6 is the commutation and interruption of the fault current.

IET
and where At ω 0 t = π, the current i T 2 cannot reverse as the diode in series with T 2 is reverse bias and the voltage V C C across the commutation capacitor retains a charge of −V DC . Fig. 2b illustrates how the fault current rises. When T 4 is switched on at t 0 (with the others switches off), the energy stored in the capacitor bank C bank is released through the load inductor L C and fault resistor R FAULT to produce a high current through the VCB before the electrodes separate. The commutation circuit is a second-order L-C-R circuit with a capacitor initial voltage of V DC . The load resistor R LOAD is ignored due to its large resistance compared with the fault resistor R FAULT . The circuit loop

Activation of the fault by switching T 4
with the initial conditions Equation (10) can be divided into three different models in term of the relationship between R FAULT and 2 L LOAD /C bank . When R FAULT . 2 L LOAD /C bank , it is an over-damped circuit, such that there are two unequally negative real numbers in its latent root. Thus, it yields and where The capacitor bank C bank always discharges its stored energy into the fault path. When t =0 + , i FAULT (0 + ) = 0; as t → ∞, i FAULT (∞)=0. Thus, the fault current must rise from zero and then decrease to zero; and time t m to reach the maximum current, is determined from di FAULT /dt = 0, and is given by Since the opening speed of the VCB is found to be about 1 m/s and the occurrence of the arc voltage indicates the contacts of the VCB starts to separate [9,10], the time difference between t 1 and t 2 determines how far the internal electrodes have separated. Independent of the discharge mode, the commutation circuit current (having triggered T 3 ) has to rise to (or exceed) the fault current i FAULT (t) level at t 2 , in order to be able to investigate VCB interruption properties. When the electrodes start to open at t 1 , the VCB experiences an arc voltage, the polarity of which depends on the direction of the VCB current, until a successful commutation is achieved.

Commutation of the VCB by switching T 3
In Fig. 2c, the commutation period t 2 − t 3 is far shorter than the fault path time constant, and it is assumed that the fault current i FAULT (t)is constant within this period. T 3 is fired to introduce the counter-current flow through the VCB, forcing the fault current through the VCB to zero, once the electrodes have separated to a specific distance. During this interval, the relationship between i FAULT , i VCB and i T 3 is given by where the differential equation for i T 3 is the same but with opposite direction to i T 2 during commutation and is given by with the initial conditions and where v 0 = 1/ L C C C (rad/s) and Z = L C /C C (V) With the appropriate choice of L C and C C , that is, with an appropriate selection of di/dt, a successful interruption should occur as a result of the first CZ. The second CZ is utilised if the first interruption fails. As the auxiliary switches (T 3 and T 4 ) are uni-directional due to the series blocking diodes thereby preventing the current reversing, the counter-current introduced by the resonant L C C C circuit can produce at most two CZs as its amplitude exceeds the VCB current at the time of commutation, thus providing two opportunities for an interruption. The interruption process is considered a failure if the VCB current continues after the second gap CZ (this is similar to AC breakers which can be specified based on two mains cycles for commutation). It is assumed that interruption is achieved at first CZ, which means i T 3 (t 3 ) = i FAULT (t 2 , t 3 ). The time t CZ1 to the first VCB CZ is The timesingledollar singledollar to the second VCB CZ is The VCB voltage falls to and is clamped to the residual voltage of the commutation capacitor. Thus, the fault current passing through the VCB is commutated into the L C − C C path; and the current i T ′ where the voltage across C bank can be considered a DC source due to a large C bank with the initial conditions Practically, R FAULT , 2 L LOAD /C C . Thus, (19) yields where and During interval t 3 −t 5 , the current i T ′ 3 initially increases due to the stored magnetic energy transfer and the residual voltage on the commutation capacitor. Then it starts to reduce, allowing diode D FW to conduct; whence the voltage across the VCB reaches its maximum voltage V DC thereby forward biasing the freewheel diode.
With diode D FW forward bias, the load (the fault) is bypassed, effectively decoupling the fault from the commutation circuit.
discharges through the circuit loop C bank −C C −T 3 −L C −D FW , transferring magnetic energy to an electric field, causing V C C to increase according to with the initial conditions

(t 5 ) (A) and V C ′′
where v 0 = 1/ L C C C (rad/s) and Z = L C /C C (V) The voltage V C C across the commutation capacitor terminates with the opposite polarity, when current i T ′′ 3 reduces to zero. The transfer of the stored magnetic energy in L C causes this voltage rise. The time t d , t 5 to t 6 ,i s As the current i T ′′ 3 reduces to zero at t 6 , the current i FW conducts through the circuit loop L LOAD −T 4 −R FAULT −D FW to dissipate the magnetic energy stored in the load (fault) inductor L LOAD , which obeys the following differential equation with the initial conditions where t 3 = L LOAD /R FAULT The analytical equations corresponding to each interval of the test circuit have now been derived. They can be used to calculate the required peak voltage, peak currents and the CZ time when the component values and conditions are known; such as when the commutation circuit is triggered to achieve a CZ.

Simulation results
A typical interruption failure occurs when the counter-current produced by the commutation circuit is less than the fault current due to slow detection and triggering of the solid-state switches. That is, there is no CZ in the VCB. The commutation parameters are initially deliberately designed to result in an unsuccessful interruption. Fig. 3a gives the simulated results overview. Fig. 3b depicts the simulation in more detail, and the preparation and interruption of the test circuit. The figures on the left describe the voltage across capacitor bank V C bank , the commutation capacitor V C C and the main switch V VCB . The figures on the right plot the current passing through the main switch i VCB , the fault path i FAULT , the resistor R 1 path i 1 , the commutation T 2 path i T 2 and T 3 path i T 3 . A failed interruption induced by large di/dt and dv VCB /dt can be simulated, as shown in Figs. 3c and d. With high di/dt and dv VCB /dt, the current passing through the VCB still conducts even if CZ points occur. The model of the VCB is seen as a short circuit in this situation [19,20].
Test circuit performance during a successful interruption at the first CZ is shown in Fig. 4. Fig. 4a shows the overall view and Fig. 4b depicts the results in detail. The waveforms on the left are voltage profile, while current profile is on the right.
During successful interruption, the energy stored in the inductance is transferred so as to contribute to the V C C increase. In Fig. 4, V VCB immediately equals V C C when the first CZ occurs and then tracks its change until the free-wheel diode D FW conducts. Test circuit experimentation is necessary to validate the analysis and simulation results, and also to explore VCB interruption in terms of external conditions, including di/dt and dv VCB /dt. Interruption probability is based on the successful interruption at either of first two CZ points. To avoid system damage due to the effects of excessive currents, an MOV was connected across the VCB and a series diode was located before capacitor bank (C bank = 7 mF) to block currents flowing back into the DC source. 600 Ω load resistance was utilised in all the experimentation. The VCB characteristics are shown in Table 2.
The performance of the test circuit interrupting a 330 A fault current at the first CZ is shown in Fig. 5a. After the voltage V C C across commutation capacitor reversed and retained as described in Section 2.1, the fault current i FAULT produced by the capacitor bank voltage V C bank is actuated and flows through the VCB before electrode separation, so i FAULT = i VCB . The fault current through the VCB is sinusoidally displaced by the commutation circuit current i T 3 , until the VCB current is zero; when i FAULT = i T 3 and i VCB = 0. During this period, the voltage across the VCB (V VCB )i s zero initially due to the electrodes being closed and when parting, a constant 12 V arc forms as i FAULT rises. As long as the di/dt (rate of change of the VCB current before CZ) was appropriately designed or the gap distance of the VCB reaches to a specific value to support a large di/dt, the commutation will be successful and the gap voltage clamps to the capacitor residual voltage V C C , until i T 3 starts to decrease (D FW conduction). V VCB has been charged to V DC but V C C continues to increase due to the energy stored in the commutation inductor L C .
The corresponding parts of Fig. 5b show interruption at the second gap CZ with the same interrupting current level. In this case, i VCB continues to flow through the VCB in the opposite direction. The arc voltage V VCB 12 V is reverse. Finally, the interruption is successful at the second CZ. With successive interruption attempts the resonance voltage in V VCB progressively decays, since resistance is introduced as the plasma starts to recover. Oscillation with the commutation inductor occurs after a successful interruption because of a VCB gap capacitive effect [21].
The voltage performance during an unsuccessful interruption is shown in Fig. 6a, with the corresponding current waveforms in the expanded view. Interruption failure occurrence is random; sometimes, even with successful interruption conditions at the first or second CZ, failure occurs without any electrical visual indication, as shown in Figs. 6a and b. The reason for re-ignition is that there is no vacuum state in the vacuum interrupter after CZ or parts of the contacts are still able to emit vapour because of a hot spot on electrode surface after experiencing the arc. In other words, the arc starts to recover as its current is reduced, but recovery speed is determined by a recovery rate. For a given gap, if di/dt is slightly more than the recovery rate, interruption is possible. However, if di/dt significantly exceeds the recovery rate,   interruption becomes impossible. Failed interruption due to a peak i T 3 that does not realise a CZ in i VCB is shown in Fig. 6c. After electrode separation, the VCB maintains stable movement with an opening speed of about 1 m/s. The occurrence of the arc voltage could be considered as the separation point of the VCB internal electrodes if the interruption current is low for a given VCB. This means the period of the arc voltage is related to the how far the electrodes have parted. With adjustment of the control signals as shown in Fig. 1, the delay is controlled by varying time between t 1 and t 2 . That is, this determines the vacuum gap at point that commutation starts. Thus, the VCB can be interrupt at the same current level for different gap distances, by adjustment of the control signals as shown in Fig. 7.Inth isfigure, the time between V C c been reversed and the occurrence of arc voltage is fixed at 2 ms, but the time to trigger i FAULT is variable; hence interruption at gap distances 1and2mmareshowninFig.7a and Fig. 7b, respectively.
The VCB characteristics in terms of the relationship between di/dt and dv VCB /dt were investigated with interruption currents of 90 and 330 A at arcing times of 1 and 2 ms. Based on the fixed delay time as shown in Fig. 7, Fig. 8a illustrates the results with interruption currents of 90 and 330 A at a gap distance of 1 mm. The results for interruption at gap distances 1 and 2 mm with interruption current 90 A are shown in Fig. 8b. Interruption success and failure are represented by the symbol circles and crosses, respectively. The lines indicate the boundary between successful and failed interruption (the area above each lines is the failed interruption zone). The experimental conditions are shown in Table 3. The experiments were repeated at least 30 times for every value. i FAULT is the interruption current; gis gap length; di/dt is rate of change of the VCB current before CZ; dv VCB /dt is the VCB voltage immediately after CZ.
With the same di/dt, the dv VCB /dt in interrupting 330 A is small than when interrupting 90 A. At 1 mm, although some dv VCB /dt are large compared with those at 2 mm, the di/dt generally presents a small value. Regardless of interrupting 90 and 330 A or with a gap distance of 1 and 2 mm, in obtaining 100% interruption probability, an increase in di/dt requires a decrease in dv VCB /dt. However, di/dt cannot increase even if it is aided by a decrease in dv VCB /dt. An experiment interrupting 330 A at a 1 mm gap was carried out. The result indicated that the interruption probability can reach 100% when di/dt is reduced to 2.53 A/μs. The VCB interruption characteristics were investigated with varying interruption current, di/dt, and arcing time, with results shown in Figs. 8c and d. The interruption success and failure are indicated by the symbol circles and crosses, respectively. The boundaries between successful and failed interruption are denoted by solid lines (above the line is the failed interruption zone). The experimental conditions are tabulated in Table 4 and the experiments were repeated at least 30 times for the test.
As shown in Fig. 8c, with a small arcing time, the di/dt ′ s asymptote regardless of the interruption current level. For a large arcing time, the interruption limit current increases with decreasing di/dt. The reason seems that the dielectric strength increases as the gap volume is increased [22]. Fig. 8d shows that a failed interruption could occur with a low probability (about 1 in 30 times) even if the di/dtis low. The successful interruption of the VCB is very sensitive to di/dt,d v VCB /dt and interruption current level. In addition, all these factors vary among VCBs. With LC commutation circuit, the peak value of the counter-current produced is larger than the fault current, therefore introducing at most two CZs with different interruption current level. The interruption current is equal to the fault current during the first CZ. The interruption current level is smaller at second CZ. When the di/dt at first CZ is too large resulting in failure interruption, successful interruption will occurs at second CZ as current level has reduced.

Conclusion
An active commutation test circuit is proposed in this paper to investigate VCB properties in terms of varied interruption current, di/dt,d v VCB /dt and arcing time. The experimental results confirmed that VCB interruption performance at low power rating is similar to that at high power rating. A decrease in di/dt or dv VCB /dt increases the interruption probability. Based on a fixed di/dt,d v VCB /dt and arcing time, the interruption probability is inversely proportional to the interruption current. di/dt and dv VCB / dt for successful interruption reduce with decreased electrode spacing. However, the key parameter determining interruption probability is di/dt, where even if dv VCB /dt is low, successful interruption become impossible if the di/dt is above certain level. In addition, the performance of the proposed active commutation test circuit offers the following features: † There is no need for a continuous external voltage to charge the commutation capacitor C C : it is charged from the system voltage. † There is no need for large capacitance C C , since the interruption time is reduced. † With an optimally designed commutation circuit, the interruption probability is improved and the extra time for commutation can be compensated by early triggering. Electrode surface erosion is reduced, thus the useful life of the VCB is extended. An accurate VCB model is not presented in this paper, whereas the VCB model used in this paper attempt to retain the features and regular pattern of successful VCB interruptions based on external conditions. Although this approach is useful in the fault current interruption analysis, an accurate VCB model will be useful in system level analysis and the simulation of the vacuum arc interruption. Further investigation on the VCB model is recommended. Current investigation to improve the VCB interruption during t 2 -t 6 intervals are ongoing with the proposed test circuit topology.