Christoph Haederli - Les thèses en ligne de l'INP - Institut National ...

164 Application and Verification
The unconstrained DC CM injection achieves a dv/dt of 130V/s in the NP voltage. This
corresponds with a current of 530 mA (C effective = 4 mF). This is 10% of the phase peak current of
5.3 A in this operating point. The unconstrained DC CM injection controller relies on a minimum
of active current to operate, in the case of pure reactive power; it could not control the NP voltage
at all. The stated value could be obtained by coincidence and cannot be relied on. Application of
the real time NP current function on the other hand makes systematic use of the peak currents
available for the NP, independently of the active part of the current. The result is a dv/dt of
200V/s. This corresponds with a current of 800mA or 15% of the phase peak current, which is
above the values expected from simulation (~10%, see Figure 68). This mismatch can be attributed
to the ohmic part of the load (copper and magnetic losses, forward voltage drop in
semiconductors) in the setup, which helps increasing the peak NP current. Also the measurements
with the oscilloscope are not very precise, so that the match can be considered sufficiently well.
The use of modified vectors allows for 30V / 50ms equal to 600V / s. This corresponds with a
current of 2.4A or 45% of the phase peak current. This is a value, which is not expected in reality
and the controller can be considered very powerful and reliable. Note that the modified vectors
applied at low modulation depth generate CM step visible in the individual line-neutral voltages
(yellow, cyan) but do not generate any distortion in the line to line voltages. The current thus keeps
the same THD as with the other NP controllers. The unbalance that can be observed in the current
of the figure in TABLE 73 ([a] and [c], green) is caused by the NP voltage step (open loop
operation).
TABLE 74 does not contain results with unconstrained DC CM injection, because this type of
controller was not able to stabilize the NP in the experiment. This has been expected from theory
and simulation and underlines that unconstrained DC CM injection should not be used in reactive
power operation at any modulation depth. The real time NP current function and the 6 th harmonic
injection are very much equivalent, resulting in a dv/dt of 600V/s. The resulting 2.4A correspond
with 8.5% of the 28A phase peak current. The calculated values shown in Figure 68 indicate a
maximum NP current of around 7%, which is a good match.
The use of vector sequences with virtual vectors type 2 reduces the duration of the transient to
20ms. The resulting 6A of NP current are 21% of the phase peak current, which is less than at low
modulation depth but significantly more powerful than any other control concept available. Note
that the virtual vectors are only applied during the transient and the real time NP current scheme is
used in steady state. This is also visible in TABLE 74 I both in line to neutral and line to line
voltages. When going to over modulation (m>1), real time NP current scheme and 6 th harmonic
injection do not provide sufficient NP current range to keep the system stable due to the very small
range of CM injection that can be used in this mode. The virtual vectors type 2 on the other hand
are available even for full trapezoidal modulation. The virtual vector NP controller remains very
powerful with an experimental maximum of 21% NP current injection as can be seen from TABLE
74 (d).
It can be concluded that the three proposed NP control schemes and modulators are very
powerful, in fact even essential for the reliable operation of the 5-L ANPC type 1 in pure reactive
power operation. Where 6 th harmonic injection and the real time NP current scheme have
equivalent performance as some SVM schemes proposed in literature, the virtual vectors extend the
available NP current range significantly and allow for more robust operation.