5.8 Experimental results of the active filtering
5.8.1 Steady-state measurements
“Today’s scientists have substituted mathematics for experiments, and they wander off through equation after equation, and even-tually build a structure which has no relation to reality.”
– Nikola Tesla in Modern Mechanix and Inventions, July, 1934
Steady-state measurements were carried out with the test setup shown in Fig. 5.5. All three control approaches, the frequency-domain, voltage feedback and time-domain-based meth-ods were measured. The dc-link voltage reference of 630 V was used except in two volt-age feedback measurements (Figs. 5.24 and 5.25), in which the dc-link voltvolt-age setpoint was 680 V. In the voltage feedback measurements the background distortion was compensated in addition to the nonlinear load originating distortion. This required increasing of the magni-tudes of the injected harmonic currents, which, in turn, required a higher dc-link voltage. All cases were measured with and without the dc-load of the active filter. The cases without the dc-load correspond to the conventional use of the active filters, in which the only purpose of using the active filter is to compensate deficiencies of nonlinear load current. The cases measured with the dc-load represent the application where the active filtering is an add-on feature to the line converter. The primary task is to regulate the dc-link voltage and supply dc-loads but, additionally, harmonic currents can be injected to compensate nonlinear load harmonics. The dc-load used was a 20Ωresistor, which with 630 V dc-link voltage corre-spond to a 20 kW load power. In cases where the dc-link voltage was 680 V, the dc-load resistor was increased to 30Ωto prevent severe overloading of the line converter. The 30Ω resistor imposed a 15 kW load to the line converter. An average switching frequency of 4 kHz was used, except in the time-domain active filtering where the switching frequency was about 3 kHz even though the hysteresis bands were set to have zero widths. In the cases of the frequency-domain method and the time-domain method the fundamental wave reactive power of the nonlinear load was compensated as well. With the voltage feedback method this is impossible because the load current is not measured. In the case of the voltage feedback method the active filter fundamental wave reactive power was controlled to zero.
The following 11 measurements were performed:
1. Grid voltage with both nonlinear load and active filter disconnected: Fig. 5.17 2. Grid voltage with nonlinear load connected and active filter disconnected: Fig. 5.18 3. Frequency-domain active filtering, LCL-filter: Fig. 5.19
4. Frequency-domain active filtering with dc-load, LCL-filter: Fig. 5.20 5. Frequency-domain active filtering, L-filter: Fig. 5.21
6. Frequency-domain active filtering with dc-load, L-filter: Fig. 5.22
7. Voltage feedback active filtering, compensation of the background distortion, LCL-filter: Fig. 5.23
8. Voltage feedback active filtering, compensation of the nonlinear load and background distortion, LCL-filter: Fig. 5.24
9. Voltage feedback active filtering with dc-load, compensation of the nonlinear load and background distortion, LCL-filter: Fig. 5.25
5.8 Experimental results of the active filtering 113
10. Time-domain active filtering, L-filter: Fig. 5.26
11. Time-domain active filtering with dc-load, L-filter: Fig. 5.27
Measurements 1 and 2 serve as reference points. Measurement 1 shows the background dis-tortion, which is transmitted through the 380/380 V transformer and is present in the primary side voltage. The measured case represents a very typical background distortion in the lab-oratory establishment in question. The feeding 1 MVA transformer was only lightly loaded and, evidently, the background distortion comes from the medium voltage network. Mea-surement 2 presents the case where no active filtering actions were taken. Note, that in the measurements 1 and 2 the LCL-filter was also disconnected from the grid.
Measurements 3–6 may be used to evaluate the effect of the active filter LCL-line filter.
Measurements 5, 6, 10 and 11 may be compared to evaluate the steady-state performance of the frequency-domain and the time-domain control approaches. Measurements 3 and 8 may be compared to evaluate the effects of the voltage feedback approach. Also, measurements 4 and 9 may be compared noting that the dc-loads are different (20 kW in measurement 4 versus 15 kW in measurement 9).
The data were sampled with a 250 kHz sample rate and a record length of 400 ms (i.e.
20 fundamental cycles) using a Yokogawa PZ4000 power analyzer. Current transformers (Goerz Electro GE 4461, 100 A/5 A) were used in the current measurements. The current transformers used have found to perform acceptably in measurements involving inverter-fed motors. Manula (1999) measured the accuracy of current transformers used in LUT with 1 A primary current in frequencies 50–5000 Hz and with 10 A primary current in frequencies 50 Hz and 1000 Hz. The results indicate very small errors (≤0.1%). The measured current transformers were not the same that was used in the measurements of this dissertation, but the measured accuracy indicates that current transformers may be accurate in frequencies higher than the power frequency.
The frequency-domain representations were calculated by applying DFT to the sample of 20 cycles. A rectangular window was used. As a result, frequency components from 0 to 125 kHz with 2.5 Hz separation were obtained. The components corresponding to the harmonic frequencies were picked up from the DFT output. Each measured waveform is portrayed in the time-domain and in the frequency-domain. In the time-domain presentations two fundamental cycles are shown. The frequency-domain figures show harmonics up to the 40thcomponent.
Five numerical parameters were calculated from the data samples. The total harmonic distor-tions up to the 40thharmonic (THD40) and up to the 200thharmonic (THD200) were evalu-ated using (2.19) on page 24. The fundamental wave amplitude was obtained directly from the DFT output and the fundamental wave RMS (FW RMS) was calculated by dividing it by√
2.
The true RMS value was calculated using (2.25) withTend =0.4 s. From the fundamental wave RMS value and the true RMS value the THD0was evaluated using (2.24). Several THD indices help in assessing where the distortion is located in the frequency-domain. The THD40
and the THD200represent the harmonic distortion up to 2 kHz and 10 kHz, respectively. The THD0covers all harmonic distortion and all interharmonic distortion. The difference between THD40and THD200reveals that there is harmonic distortion in the frequency band 2–10 kHz.
If the THD0is almost similar to the THD200there is no harmonic distortion over 10 kHz and no interharmonic distortion at all.
0 0.02 0.04
Figure 5.17: Grid phase voltage with both nonlinear load and active filter disconnected. (Measurement 1)
Figure 5.18: Nonlinear load connected, active filter disconnected. Top: Grid voltage, Bottom: Nonlin-ear load current. (Measurement 2)
0 0.02 0.04
Figure 5.19: Frequency-domain active filtering, LCL-filter. Top to Bottom: Grid voltage, Supply cur-rent, Nonlinear load curcur-rent, Active filter current. (Measurement 3)
5.8 Experimental results of the active filtering 115
Figure 5.20: Frequency-domain active filtering, LCL-filter, Dc-load. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 4)
0 0.02 0.04
Figure 5.21: Frequency-domain active filtering, L-filter. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 5)
0 0.02 0.04
Figure 5.22: Frequency-domain active filtering, L-filter, Dc-load. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 6)
0 0.02 0.04
Figure 5.23: Voltage feedback active filtering, compensation of background distortion, nonlinear load is disconnected, LCL-filter. Top: Grid voltage, Bottom: Supply current (which is also active filter current). (Measurement 7)
5.8 Experimental results of the active filtering 117
Figure 5.24: Voltage feedback active filtering, compensation of nonlinear load distortion and back-ground distortion, LCL-filter. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 8)
0 0.02 0.04
Figure 5.25: Voltage feedback active filtering, compensation of nonlinear load distortion and back-ground distortion, LCL-filter, Dc-load. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 9)
0 0.02 0.04
Figure 5.26: Time-domain active filtering, L-filter. Top to Bottom: Grid voltage, Supply current, Non-linear load current, Active filter current. (Measurement 10)
0 0.02 0.04
Figure 5.27: Time-domain active filtering, L-filter, Dc-load. Top to Bottom: Grid voltage, Supply current, Nonlinear load current, Active filter current. (Measurement 11)
5.8 Experimental results of the active filtering 119