4.1. Experimental Investigation of Interconnection Coordination
Since the interconnection coordination in
Figure 12a represents the smallest wiring effort, this option was chosen as the base case for the demonstrator test circuit. However, during the commissioning tests on the basis of the base case design, automatic terminations by the test circuit control can be reproducibly observed, even at the lower end of the high-power test circuit’s limit spectrum. The causal chain that leads to these terminations can be found in the design of combining three individual PEBC modules each into PEBC module packages. A central control signal is generated for each PEBC module package, which is transmitted to the GDU of the master module. The master’s GDU processes the signal internally and relays it to the slave modules’ GDUs, which change the switching state of the IGBTs synchronously with the master module (cf.
Section 3.3.3). At the same time, the respective gate-emitter voltage
VGE of the individual modules is monitored locally at the GDU and continuously compared with the gate-emitter voltages
VGE of the other two PEBC modules involved. If the deviation leaves a predefined tolerance band, this is transmitted to the CCU as an error condition, whereupon the CCU sends a switch-off signal to all GDUs of the overall test circuit and the control is aborted. To investigate this, the comparative analysis of the
VGE voltages is conducted.
The first essential object of investigation is the formation of the Miller plateau, as described in
Section 2.2, during the turn-on of the modules’ IGBT.
Figure 13a shows the corresponding time range of the
VGE voltages (
VGE,Master,
VGE,Slave1,
VGE,Slave2) during the initial switch-on at the set parameters from
Table 2 for the base case interconnection between the PEBC modules. As can be seen, the formation of the Miller plateau, between
t = 4 µs and
t = 7 µs, does not exhibit significant deviations between the three associated
VGE voltages. However, first indications as to the subsequently growing discrepancies can be observed after
t = 13 µs.
Figure 14a acts as the continuation of the identical switch-on operation as shown in
Figure 13a for the time range of
t = 20 µs to
t = 100 µs. The target gate-emitter voltage for the applied IGBTs is
VGE,target = 15 V. As is evident from the measurements, the deviations between the three measured gate-emitter voltages (
VGE,Master,
VGE,Slave1,
VGE,Slave2) as well as from the target voltage (
VGE,target) reach values of up to
VGE,Diff = 2.43 V. These in turn lead to the automatic termination by the test circuit control.
The cause for the considerable deviations between
VGE,Master,
VGE,Slave1 and
VGE,Slave2 respectively is best described using the involved inductances, highlighted in
Figure 11. The GDUs are using the potential of the emitter-terminal (
E) on the module level as the reference potential for setting the gate-emitter voltage
VGE. Additionally, this emitter potential is taken as the reference point for the comparison between the
VGE voltages within the master–slave configuration. However, since the stray inductances between the modules and the joint point of common coupling
PCC2 differ in value (cf. Equations (7)–(9)), different voltage drops for all three involved PEBC modules result. This leads to different values of the emitter-potential, which the associated GDUs try to even out. Since the computing time of each GDU is finite, these matching processes interfere with each other, which ultimately leads to increasing deviations and an automatic termination by the test circuit control.
The choice of the presented parameters for the experimental investigations (
itest,
vtest, cf.
Table 2) arises from the fact that with the base case coordination design, a stable operation above these parameters cannot be achieved. As a consequence, the coordination design of the base case is dismissed as not feasible for the provision of a stable test circuit control. The main objective of the development of new interconnection options is therefore to harmonize the emitter inductances from the terminals of the PEBC modules to the point of common coupling of the superordinate PEBC module package (
LSlave1→PCC2,
LMaster→PCC2,
LSlave1→PCC2). The design steps are highlighted in
Figure 12b–d and the results of the associated experimental investigations are discussed hereafter.
As a first step towards homogenizing the emitter inductances within the PEBC module packages, the connection of the PEBC module internal busbars to the common busbar is modified towards a common star point connection via short cables with high-current carrying capability. The added cable inductances are
LE1′,
LE2′ and
LE3′ (
LEj’), resulting in the overall inductances from each PEBC module’s emitter terminal to the
PCC2, as described in Equation (10).
Considering the formation of the Miller plateau (cf.
Figure 13b), no significant difference can be observed as compared to
Figure 13a. This is, however, due to the fact, that the formation of the Miller plateau is not the critical phase during a switch-on operation. Nevertheless, it can be shown that the introduction of the star point connection also does not negatively impact this phase. Thus, the principle feasibility of the developed interconnection can be accepted for the formation of the Miller plateau. As a supplementary, it is evident from
Figure 14b, that the redesign towards option (b) significantly improves the stability and the homogeneity of the gate-emitter voltages
VGE. It is important to note, that for the base case, the ordinate-scale ranges from
VGE = 11 V to
VGE = 17 V, whereas in the other displayed measurement results (
Figure 14b–d), the voltage scale only includes
VGE = 14.5 V to
VGE = 15.5 V. A presentation with a uniform scale was not feasible, as otherwise the effects studied could not be presented sufficiently well.
However, especially in the time range until
VGE reaches the steady state value of
VGE,on, deviations between the gate-emitter voltages of the investigated PEBC modules remain for the interconnection option (b). This can be attributed to the fact that constructive tolerances in the module internal busbars, as well as the star point connection cables, cannot be ruled out across the board. In order to reduce these differences to a minimum, direct connections between the emitter terminals are introduced in concept (c). Through this modification, the emitter potential is almost perfectly equalized within the PEBC module package. Again, the effects of the extended design in the first 20 µs of the turn-on operation (cf.
Figure 13c) are negligible. For the continued switch-on operation (
t = 20 µs to
t = 100 µs, cf.
Figure 14c), on the other hand, the impact on the homogeneity of the gate-emitter voltages (
VGE) can be clearly observed. At this state of the design, the space for optimization regarding the harmonization of the emitter inductances is exhausted in constructional terms. However, minor deviations still persist between the considered
VGE voltages.
The last parameter with potential for optimization within the confines of the constructive and peripherical boundary conditions of the demonstrator test circuit is the harmonization of the charging voltages of the module capacitors. These impact the
VGE voltages, since a higher charging voltage of one module capacitor leads to a higher module output current and thus to a higher voltage drop across the busbar to the star point connection. Since PE equipment is sensitive to even small deviations, already capacitance variances of the module capacitors due to manufacturing tolerances can lead to instabilities. This challenge is resolved by the introduction of direct connections between the high voltage terminals of the module capacitors within a PEBC module package. Through this constructive extension, the level of the charging voltage can be homogenized for all three PEBC modules. The resulting
VGE measurements (
Figure 13d and
Figure 14d) prove the effectivity of the developed design for the interconnection coordination within one PEBC module package. In order to avoid propagation of the fault in the event of a failure, the high-voltage electrodes of all test circuit modules are not connected directly to each other in the practical implementation. Although this would in theory lead to a homogenized source voltage distribution across the entire test circuit, it might also lead to an uncontrolled discharge of all involved modules into one faulty module and consequently in the worst case to the destruction of the module. The uniform charging level of the test circuit is realized through the charging circuit, as depicted in
Figure 6, which is used to supply the entire test circuit charge from one source. As the results presented in
Section 4.2 demonstrate, the measure to directly connect all high-voltage electrodes of one test circuit stage is not necessary for a stable and secure operation of the developed demonstrator high-power test circuit. It is, however, recognized that this design parameter can serve as an optimization potential in future applications.
As a quantitative measure to assess the effectiveness of the interconnection developments, the absolute value of the deviation (
VGE,Diff) between the master’s
VGE and the slaves’
VGE, is determined. The results are shown in
Figure 15a–d. Already an optical inspection of the results reveals the gradual effects, starting from the base case (a) to the subsequent further development steps b–d.
A gradual reduction of the absolute value for
VGE,Diff can be observed as the complexity of the interconnection coordination increases and approaches the optimum (from a to d). The respective values are included in
Table 3. Through the presented optimized interconnection design d), a reduction of
VGE,Diff by 90.5%, as compared to the base case a, can be achieved. This improves the test circuit control stability significantly and ensures the integrated functionality of interconnected high-power PEBC modules. For (a) reliable test circuit control, this value needs to be limited to a maximum deviation of
VGE,Diff < 1.0 V.
The results presented above support the decision not to integrate control inductances at the PEBC module level. Even small deviations in the stray inductances, which are typically orders of magnitude smaller than dedicated control inductances, lead to increasing difficulties regarding the control stability of the high-power test circuit. Since production-specific tolerances of the actual inductance ratings cannot be ruled out, this uncertainty factor is eliminated from the outset in the development process of the modular high-power test circuit.
4.2. Experimental Demonstration of Presented Interconnection Method
In order to demonstrate the feasibility of the described method of interconnecting multiple high-power PEBC modules to form one overall high-power test circuit, an exemplary test current generation is analyzed. For this purpose, a rectangular test current waveform with an amplitude of
îtest,target = 10.0 kA for a duration of
t = 50 ms is specified as the target current (cf.
Table 4). The test circuit capacitances are pre-charged to
Vtest,overall = 3.0 kV, which translates to the individual modules being charged to
Vtest,module = 1.5 kV. For the presented experimental demonstration, the control inductance is set to
LControl = 0.5 mH. The underlying control algorithm applied is a two-point control, which is described in detail in [
24].
The specified target current waveform (
itest,target, dashed grey), the generated test current (
itest, continuous black) and the corresponding course of the overall test circuit voltage (
vtest, continuous grey) for the conducted experiment are depicted in
Figure 16.
At
t = 0 ms, the CCU distributes a switching signal,
HIGH, to the PEBC modules of the test circuit, which is realized according to the control flow chart in
Figure 10. This initiates the current flow. Simultaneously, the test circuit voltage decreases as a result of the capacitors being discharged to feed the test current. At
t = 1.7 ms, the upper target current envelope limit is reached and the CCU changes the switching signal to
LOW. As a result, the current flow commutates to the free-wheeling diodes of the PEBC modules and the overall test current decreases due to ohmic losses and the losses in the DuT. At
t = 1.96 ms, the lower target current limit is reached and the CCU changes the control signal back to
HIGH, leading to a renewed increase in test current due to a further discharge of the capacitors. In this time range, the test circuit voltage decreases equivalently. This process is repeated until the end of the pre-set target current is reached (
t = 50 ms) and the CCU transmits a continuous
LOW signal to end the control sequence and to switch off the IGBTs.
Two phenomena need to be addressed with regards to the adequacy of the target current mapping: the finite rate of rise, respectively, the rate of fall of the generated test current and its ripple. As a result of the slope-limiting characteristics of the combined circuit inductances, the actually generated test current does not match the rate of rise of the pre-set target current. However, this is not a restriction for the intended test operation, as the field of application of the developed high-power test circuit is the simulation of realistic fault currents in future MVDC grids. The finite rate of rise adequately represents that of realistic use cases. At the end of the control sequence (t = 50 ms), the test current subsides with a relatively low rate of fall as compared to the target current curve. This is due to the magnetic energy stored in the control inductance and does also not restrict the applicability of the developed test circuit design, since this energy is to be absorbed by the MVDC circuit breaker during current interruption operation. The resulting ripple of the test current (i.e., ∆I ≤ 320 A, 3.2% of the target current amplitude in the presented test scenario) is a result of the finite switching frequency of the applied IGBTs and corresponding GDUs. In order to operate the PE switching devices in a permissible operation state, minimum on- and respectively off-times need to be adhered to. As with the previously discussed current slopes, the current ripple does not represent a functional limitation either, since it is in a comparatively low range on the one hand and, on the other hand, fault currents in future MVDC networks also exhibit current ripples.
When considering the course of the test circuit voltage over the duration of the conducted test, voltage spikes and oscillations are clearly observable during every switching operation of the IGBTs. These can be attributed to PEBC module internal commutation processes from the IGBT to the freewheeling diode, and vice versa. These processes are component-dependent and cannot be influenced through constructive measures outside of the PE switching equipment. For the test voltage, as seen by the DuT, these phenomena are negligible.
As is evident from the presented results in
Figure 16, the applied and previously discussed measures to harmonize the PEBC module internal emitter inductances, as well as the respective modules’ charging levels, sufficiently stabilize the test circuit control and enable a stable and effective high-power test circuit operation. By adjusting the control algorithm to the desired test current waveform, the generation of near-arbitrary monopolar test currents is possible and the feasibility of interconnecting multiple high-power PEBC modules to form an overall high-power test circuit that can be virtually operated as one singular PEBC is thus demonstrated and supported by the presented results.