*B*. *Comparative performance analysis between SRF-based PA estimation and conventional PQbased PA estimation under non-sinusoidal voltage conditions*

Two different PA estimation approaches were realized and a comparative analysis was presented; one was a conventional PA estimation based on the PQ concept and the second was the SRF-based PA estimation method [11,12]. A transient load condition of a sudden reactive load change was also considered. The reactive load demand was increased from 2.5 kVAr to 5 kVAr at 0.5 s. Figure 12a,b illustrates the PA estimation pattern for equal reactive power sharing based on the conventional PQ concept and the SRF concept, respectively. It was clearly observable that the PA with the PQ estimation method fluctuated and was also less than the reference value whereas the PA estimation with the SRF method followed the reference with fewer fluctuations. Both APFs with the SRF were found to be highly responsive to this sudden load change. These inconsistencies in the PA resulted in unequal reactive power sharing with the PQ concept, thus affecting the set criteria (Figure 12c). However, the PA with the SRF parameters resulted in more precise equal reactive power sharing, as depicted in Figure 12d. Table 3 illustrates a comparative analysis between the two approaches. Thus, it was clear that the SRF-based estimation of the PA involved fewer parameters for the estimation; the PA estimation errors and fluctuations were much fewer compared with the PQ method. Equal reactive power sharing between the shunt and series APF was more precise with fewer deviations from the reference (1.25 kVAR until 0.5 s and 2.5 kVAR after 0.5 s).

*Energies* **2022**, *15*, 3769


Harmonics (10% each 3rd, 5th and 7th)

Normal <sup>228</sup> <sup>227</sup> 0.5 4.0 9.9 3.4 QSH <sup>=</sup> <sup>3015</sup>

Sag (20%) <sup>180</sup> <sup>225</sup> 0.5 3.7 9.8 3.8 QSH <sup>=</sup> <sup>3045</sup>

Swell (20%) <sup>275</sup> <sup>229</sup> 0.5 2.2 9.0 3.3 QSH <sup>=</sup> <sup>3023</sup>

<sup>230</sup> <sup>226</sup> 17.4 3.0 9.7 4.6 QSH <sup>=</sup> <sup>3010</sup>

reduced from around 15% to 3%.

*Energies* **2022**, *15*, x FOR PEER REVIEW 11 of 19

**Voltage Condition I II III IV V VI VII I II III IV V VI VII I II III IV V VI VII**

QSR <sup>=</sup> <sup>2962</sup> <sup>229</sup> <sup>227</sup> 0.4 4.4 0.4 3.3 QSH <sup>=</sup> <sup>2563</sup>

QSR <sup>=</sup> <sup>2923</sup> <sup>182</sup> <sup>226</sup> 0.5 3.4 0.5 3.2 QSH <sup>=</sup> <sup>2542</sup>

QSR <sup>=</sup> <sup>2943</sup> <sup>273</sup> <sup>232</sup> 0.3 3.2 0.4 3.4 QSH <sup>=</sup> <sup>2558</sup>

QSR <sup>=</sup> <sup>2941</sup> <sup>227</sup> <sup>225</sup> 17.2 4.7 0.9 3.5 QSH <sup>=</sup> <sup>2574</sup>

*6.5. Comparative Performance Analysis under Non‐Sinusoidal Voltage Conditions*

A simulation case was considered again with a non‐linear load, but with a non‐sinusoidal supply voltage condition only. The focus was on the current compensa‐ tion by the UPC under non‐sinusoidal voltage conditions. Figure 10a,b indicates the source voltage with harmonics (10% of 3rd, 5th and 7th order harmonic components) and the load voltage with compensation from the UPC system. Figure 10c,d illustrates the FFT analysis of the source voltage and load voltage, respectively. The load voltage was

**Table 2.** Performance parameters under different source voltage and loading conditions. **Composite Load Linear Load Only Non‐Linear Load Only**

QSR <sup>=</sup> <sup>2495</sup> <sup>228</sup> <sup>228</sup> 0.5 2.22 18.5 3.6 ‐

QSR <sup>=</sup> <sup>2478</sup> <sup>180</sup> <sup>226</sup> 0.5 3.26 18.63 3.6 ‐

QSR <sup>=</sup> <sup>2423</sup> <sup>275</sup> <sup>227</sup> 0.5 3.55 16.78 3.7 ‐

QSR <sup>=</sup> <sup>2456</sup> <sup>230</sup> <sup>226</sup> 17.26 4.08 20.65 3.9 ‐

**Figure 10.** Voltage analysis with UPC system under harmonic conditions: (**a**) source voltage; (**b**) load voltage; (**c**) FFT analysis of source voltage; (**d**) FFT analysis of load voltage. **Figure 10.** Voltage analysis with UPC system under harmonic conditions: (**a**) source voltage; (**b**) load voltage; (**c**) FFT analysis of source voltage; (**d**) FFT analysis of load voltage.

The efficacy of the SRF method of compensation over the conventional PQ method under non‐sinusoidal voltage conditions in terms of current compensation and equal reactive power sharing, as discussed in Section 4, is exhibited in a further analysis. *A. THD Analysis of Source Current between SRF and Conventional PQ Methods*

Figure 11a shows the source current waveform obtained from the PQ method. The source current waveform after compensation from the SRF method is illustrated in Figure 11b. It was clearly observable that the source current waveform obtained with the SRF

presented in Figure 11c,d for the PQ and SRF methods, respectively. It was clear that the THD of the source current with the SRF method was below 5% whereas with the PQ

method, it was above the allowable limit with non‐sinusoidal voltage conditions.

(**a**)

*Energies* **2022**, *15*, x FOR PEER REVIEW 12 of 19

(**c**)

(**d**)

method, it was above the allowable limit with non‐sinusoidal voltage conditions.

load voltage; (**c**) FFT analysis of source voltage; (**d**) FFT analysis of load voltage.

**Figure 10.** Voltage analysis with UPC system under harmonic conditions: (**a**) source voltage; (**b**)

The efficacy of the SRF method of compensation over the conventional PQ method under non‐sinusoidal voltage conditions in terms of current compensation and equal reactive power sharing, as discussed in Section 4, is exhibited in a further analysis. *A. THD Analysis of Source Current between SRF and Conventional PQ Methods*

Figure 11a shows the source current waveform obtained from the PQ method. The source current waveform after compensation from the SRF method is illustrated in Figure 11b. It was clearly observable that the source current waveform obtained with the SRF method was less contaminated than with the PQ method. The current FFT analysis is presented in Figure 11c,d for the PQ and SRF methods, respectively. It was clear that the THD of the source current with the SRF method was below 5% whereas with the PQ

**Figure 11.** Source current analysis with UPC system under non‐sinusoidal voltage conditions: (**a**) source current waveform with PQ method; (**b**) source current waveform with SRF method; (**c**) FFT analysis of source current with PQ method; (**d**) FFT analysis of source current with SRF method. **Figure 11.** Source current analysis with UPC system under non-sinusoidal voltage conditions: (**a**) source current waveform with PQ method; (**b**) source current waveform with SRF method; (**c**) FFT analysis of source current with PQ method; (**d**) FFT analysis of source current with SRF method.

Two different PA estimation approaches were realized and a comparative analysis was presented; one was a conventional PA estimation based on the PQ concept and the second was the SRF‐based PA estimation method [11,12]. A transient load condition of a sudden reactive load change was also considered. The reactive load demand was in‐

*B. Comparative performance analysis between SRF‐based PA estimation and conventional*

tern for equal reactive power sharing based on the conventional PQ concept and the SRF concept, respectively. It was clearly observable that the PA with the PQ estimation method fluctuated and was also less than the reference value whereas the PA estimation with the SRF method followed the reference with fewer fluctuations. Both APFs with the SRF were found to be highly responsive to this sudden load change. These inconsisten‐ cies in the PA resulted in unequal reactive power sharing with the PQ concept, thus af‐

*PQ‐based PA estimation under non‐sinusoidal voltage conditions*

*Energies* **2022**, *15*, x FOR PEER REVIEW 14 of 19

from the reference (1.25 kVAR until 0.5 s and 2.5 kVAR after 0.5 s).

fecting the set criteria (Figure 12c). However, the PA with the SRF parameters resulted in more precise equal reactive power sharing, as depicted in Figure 12d. Table 3 illustrates a comparative analysis between the two approaches. Thus, it was clear that the SRF‐based estimation of the PA involved fewer parameters for the estimation; the PA estimation errors and fluctuations were much fewer compared with the PQ method. Equal reactive power sharing between the shunt and series APF was more precise with fewer deviations

**Figure 12.** Comparative performance analysis between PQ and SRF methods: (**a**) PA estimation by PQ method; (**b**) PA estimation by SRF method; (**c**) reactive power share by PQ method; (**d**) reactive power share by SRF method. **Figure 12.** Comparative performance analysis between PQ and SRF methods: (**a**) PA estimation by PQ method; (**b**) PA estimation by SRF method; (**c**) reactive power share by PQ method; (**d**) reactive power share by SRF method.

Simulators are useful in practice to protect equipment from being damaged. With the reduction in cost and increments in the performance of virtual and real‐time simula‐ tive technology, its availability and applicability has been widespread [14]. The real‐time simulator adapted in our work was RT‐LAB, which is based on FPGA; its flexibility and scalability can be used for virtually any simulation or control system application.

A computer system with installed SIMULINK software was connected to the simu‐ lator setup via an ethernet, as shown in Figure 13. Instead of methods such as HIL (hardware in the loop), RCP (rapid control prototyping) was adapted for analyzing the

**Figure 13.** OPAL‐RT simulator connected to the host PC via ethernet and to the DSO via connecting

For QL = 2.5 kVAR: QSh = 1.5%, QSr = −2.3% For QL = 5 kVAR: QSh = 1.1%, QSr = −2.6%

behavior of the UPC system with the proposed controller.

Performance factors PQ‐based estimation of PA SRF‐based estimation of PA Parameters involved Load voltage and load current Only load current

PA fluctuation with average 6.5% approximately 1.5% approximately

PA estimation error with reference 8% 2%

**7. Real‐Time Simulator Analysis**

probes.

Percent error in equal reactive power sharing between shunt and series APF

**Table 3.** Comparative analysis between PQ and SRF methods for PA estimation.

sharing between shunt and series APF


For QL = 5 kVAR: Qsh = 18%, QSr = −20%

*Energies* **2022**, *15*, x FOR PEER REVIEW 15 of 19

**Table 3.** Comparative analysis between PQ and SRF methods for PA estimation. (**d**)

#### **7. Real-Time Simulator Analysis 7. Real‐Time Simulator Analysis**

Simulators are useful in practice to protect equipment from being damaged. With the reduction in cost and increments in the performance of virtual and real-time simulative technology, its availability and applicability has been widespread [14]. The real-time simulator adapted in our work was RT-LAB, which is based on FPGA; its flexibility and scalability can be used for virtually any simulation or control system application. Simulators are useful in practice to protect equipment from being damaged. With the reduction in cost and increments in the performance of virtual and real‐time simula‐ tive technology, its availability and applicability has been widespread [14]. The real‐time simulator adapted in our work was RT‐LAB, which is based on FPGA; its flexibility and scalability can be used for virtually any simulation or control system application.

For QL = 5 kVAR: QSh = 1.1%, QSr = −2.6%

A computer system with installed SIMULINK software was connected to the simulator setup via an ethernet, as shown in Figure 13. Instead of methods such as HIL (hardware in the loop), RCP (rapid control prototyping) was adapted for analyzing the behavior of the UPC system with the proposed controller. A computer system with installed SIMULINK software was connected to the simu‐ lator setup via an ethernet, as shown in Figure 13. Instead of methods such as HIL (hardware in the loop), RCP (rapid control prototyping) was adapted for analyzing the behavior of the UPC system with the proposed controller.

**Figure 13.** OPAL‐RT simulator connected to the host PC via ethernet and to the DSO via connecting probes. **Figure 13.** OPAL-RT simulator connected to the host PC via ethernet and to the DSO via connecting probes.

Figure 14 illustrates the waveforms of the source voltage, load voltage and series injected voltage for three different transient conditions of source voltage such as normal to sag (Figure 14a), sag to swell (Figure 14b) and swell condition to normal voltage with harmonics (Figure 14c). The series injected voltage compensated for the source voltage and the load voltage was found to be maintained at its rated value and free from harmonics irrespective of the disturbances. Figure 15 shows the waveforms of the load current, source current and compensating current for different load conditions of the linear load (Figure 15a), non-linear load (Figure 15b) and composite load (Figure 15c), respectively. The source current was found to be distortion-free in the case of the non-linear and composite loads. With the linear load condition, the presence of a compensating current was due to the partial load reactive current demand from the shunt APF; the remainder of the reactive current demand was fulfilled by the series APF.

**Figure 14.** OPAL‐RT results of source voltage (VS), load voltage (VL) and series injected voltage (VSr) for different disturbance conditions: (**a**) normal to sag; (**b**) sag to swell; (**c**) swell to normal with **Figure 14.** OPAL-RT results of source voltage (VS), load voltage (VL) and series injected voltage (VSr) for different disturbance conditions: (**a**) normal to sag; (**b**) sag to swell; (**c**) swell to normal with harmonics.

*Energies* **2022**, *15*, x FOR PEER REVIEW 16 of 19

of the reactive current demand was fulfilled by the series APF.

Figure 14 illustrates the waveforms of the source voltage, load voltage and series

injected voltage for three different transient conditions of source voltage such as normal

to sag (Figure14a), sag to swell (Figure14b) and swell condition to normal voltage with

harmonics (Figure14c). The series injected voltage compensated for the source voltage

and the load voltage was found to be maintained at its rated value and free from har‐

monics irrespective of the disturbances. Figure 15 shows the waveforms of the load cur‐

rent, source current and compensating current for different load conditions of the linear

load (Figure 15a), non‐linear load (Figure 15b) and composite load (Figure 15c), respec‐

tively. The source current was found to be distortion‐free in the case of the non‐linear and

composite loads. With the linear load condition, the presence of a compensating current

was due to the partial load reactive current demand from the shunt APF; the remainder

harmonics.

**Figure 15.** OPAL‐RT results of load current (IL), source current (IS) and compensating current (IC) for different load conditions: (**a**) linear load only; (**b**) non‐linear load only; (**c**) composite load (both **Figure 15.** OPAL-RT results of load current (IL), source current (IS) and compensating current (IC) for different load conditions: (**a**) linear load only; (**b**) non-linear load only; (**c**) composite load (both linear and non-linear).

An equal reactive power sharing strategy for a single‐phase UPC system was pro‐

posed in this work as a more efficient and practical way of adapting identical APFs with

the same rating. The SRF‐based controller employed in this system for the shunt APF

controller also proved to be helpful under non‐sinusoidal conditions of supply voltage

for a PA estimation to implement equal reactive power sharing. In this paper, a detailed

performance analysis was presented for a UPC system under different supply voltages

(i.e., voltage sag, swell and harmonics) and loading conditions (i.e., a non‐linear load,

linear load and composite load). The performance indices considered for this analysis

were current harmonic compensation, load voltage compensation and reactive power

compensation with equal sharing criteria between the shunt and series APF. It was

clearly observed from the result analysis and tabular data compilation that the UPC sys‐

tem offered a significant compensation performance under different conditions of volt‐

age and load. A comparative analysis was presented between the PQ method and the

SRF method for current harmonic compensation, a PA estimation and equal reactive

power sharing under a non‐sinusoidal supply voltage. As deduced from this compara‐

tive analysis in terms of the results and comparative data tabulation, the SRF method

proved to be superior than the PQ method under non‐sinusoidal supply conditions.

Thus, the SRF method is not only suitable for shunt APF control implementation, but also

offers better reactive power sharing support under non‐sinusoidal supply conditions

between a shunt and series APF, utilizing only SRF parameters.

*Energies* **2022**, *15*, x FOR PEER REVIEW 17 of 19

linear and non‐linear).

**8. Conclusions**
