Symmetric-component decoupled control of grid-connected inverters for voltage unbalance correction and harmonic compensation

In this paper, a strategy is investigated for control of distributed generators to additionally support the local grid on top of conventional electricity generation. The strategy decouples unbalance and harmonic compensation in the phase sequences and the frequency domain for a grid-connected inverter with local unbalanced and distorting loads. Unlike conventional control schemes for grid-connected converters, the proposed strategy is designed to be sequence-asymmetric for the purpose of unbalanced and harmonic local voltage correction. A frequency-domain Norton equivalent model is derived to illustrate the working principle of the strategy. Accordingly, it is shown that following a frequency-domain decoupled method the fundamental positive-se-quence, the harmonic symmetrical sequence, the fundamental zero-sequence and the fundamental negative-sequence components can be regulated independently. Consistent to the model analysis, experiments validate the reduction of the local voltage total harmonic distortion and unbalanced factor when the converter is programmed for both harmonic sinking and voltage unbalance correction.


Introduction
Unbalanced voltages cause adverse effects on electrical loads and power distribution networks, such as transformer overloading and capacity limitation of distribution lines [1], temperature rise and overloading of induction motors, and efficiency reduction of power electronic rectifiers [2][3][4][5].Generators suffer from voltage imbalance as well; the effects include reduced stability margins, deteriorated transient performance of distributed generators [6], and negative impacts on wind generators such as reduced output power capacity and impaired lifetime [7][8][9].The causes of the local voltage problem include unbalanced loading, unequal line impedance and grid faults; on top of that, another main cause of voltage imbalance is the high penetration of distributed generators, e.g.single-phase photovoltaic inverters [10,11].However, this voltage unbalance problem can be mitigated by optimal control of grid-interactive power converters.
The control strategies of grid-interactive power converters dealing with unbalanced network conditions can be separated into two main categories.The first is devoted to defensively improving the power converter's immunity to the imbalance [12][13][14], and the second is dedicated to actively absorbing or sharing the unbalance.Compensation for voltage unbalance is usually implemented using an active power filter acting as a voltage source in series with the power distribution line [15].Another solution is to apply a shunt converter behaving as a current source to absorb or share the unbalanced currents [16,17]; therefore, the grid line current unbalance is reduced.However, these local voltage support strategies based on measuring the polluting load current are not always straightforward especially when multiple loads are present.Therefore, using shunt converters to enhance the local voltage quality based on only local measurements has received increasing attention in recent years [18][19][20][21][22][23][24][25].These approaches exploits the disadvantage of a weak grid by inducing an unbalanced voltage over the grid (line) impedance to compensate the local voltage unbalance.As a result, the local voltage quality is enhanced.However, limited research has been carried out on how to correct the negativesequence component and harmonics at the same time by means of three-phase grid-connected converters using only local measurements.For instance, compensation of fundamental symmetrical components is addressed in [20], zero-sequence correction in [25,21], or mainly on harmonic compensation in [24,23].More importantly, discussion on the three-phase local voltage support is mostly carried out in the synchronous reference frame, making it difficult to build a natural analogy between single-phase and three-phase converters when talking about harmonic compensation,particularly in the context of Norton equivalent model derivation.
Starting from an earlier proposed control architecture in [26], in this paper we extend the work with comprehensive model analysis and experimental investigation.Firstly, a frequency-domain analysis methodology covering positive and negative frequencies is presented in this paper to discuss the current and voltage symmetrical-components, which establishes an equivalence between the sequence components and the harmonics.Therefore, the conventional frequency response is extended from sing-side to double-side, making asymmetrical bode plots possible.A direct benefit is that, analogously to conventionally decoupling the fundamental from harmonics, the negative-sequence component can be regulated independently of the positive-sequence content using a resonant filter centred at the corresponding negativefrequency.Secondly, following the methodology a straightforward Norton equivalent model can be derived easily for the closed-loop converter system, which is not always the case for conventionally control approaches implemented in the synchronous reference frame.
Since complex coefficients present the model, asymmetrical bode plots of the equivalent impedance become possible, allowing for independent regulation of sequence components and harmonics, e.g.low-impedance for the negative-sequence component and harmonics while high-impedance for fundamental positive component.As a result, a low-impedance path can be accordingly provided for the negative-sequence component and harmonics to achieve local voltage support.The control approach developed in this paper is straightforward and can achieve harmonic compensation, local voltage zero-sequence correction and negative-sequence correction independently.The organization of the paper is as follows: the grid-connected converter system and the architecture of the proposed control strategy are briefly shown in Section 2. In Section 3, a frequency-domain analysis approach is presented to model the closed-loop three-phase converter system.Experimental results are described and the corresponding discussions are shown in Section 4.Last but not least, conclusions are drawn and recommendations are given in Section 5.

System architecture
In this paragraph the structure of the considered grid-connected inverter system is introduced.As shown in Fig. 1, the system consists of six main parts: two direct-current (DC) supplies providing a neutral line, a three-phase (3ϕ) voltage source inverter (VSI), an LCL filter for switching-frequency harmonic attenuation, local 3ϕ loads, a 3ϕ grid, and a controller providing the pulse-width-modulation (PWM) scheme for the switches.As can be seen from Fig. 1, only the local voltage at the point of common connection (pcc) and the local converter output current are measured by the controller, which distinguishes the proposed control strategy from conventional ones measuring the load current for voltage support.
Here the converter output current, the local pcc voltage, and the desired VSI output voltage in the abc reference frame are denoted as Accordingly, the corresponding signals in the αβγ reference frame are written as Before the elaboration on the control strategy, it is worthwhile to address that in this paper besides vectors and matrices in bold, symbols with a underline, e.g.(.) ̲ , represent a complex quantity, complex func- tion, or polynomial with complex coefficients.Unless mentioned otherwise, a symbol without a underline denotes a real-valued quantity or a transfer function with only real-valued polynomial coefficients.

Control strategy
This section elaborates on the design of the controller in Fig. 1 for add-on voltage support.Note that the two objectives of the control strategy are to regulate the active power injection and to support the local voltage simultaneously.Since the average (not the instantaneous) active power is dominantly determined by the fundamental positivesequence component of the converter output current, as demonstrated in [27], the active power injection can be controlled by regulating the fundamental positive-sequence component of the converter output current.At the same time, because the fundamental negative-and zerosequence components and the harmonics hardly contribute to the average active power, they can be regulated to correct the local voltage for unbalanced and harmonic compensation.Therefore, the controller is composed of two loops: one for the control of the output current fundamental positive-sequence component and the other for the regulation of the pcc voltage fundamental negative-and zero-sequence components and the harmonics.
Accordingly, a parallel current-voltage control architecture is considered in this paper, because it can well generalize a current controller [16,17], a hybrid current-voltage control scheme [19,20] and a parallel scheme [28].Fig. 2 shows the block diagram of the current-voltage Fig. 1.Diagram of a three-phase four-wire grid-connected voltage source inverter system with local loads.control architecture.In the proposed control strategy, the current reference is composed of only fundamental positive-sequence component.A conventional phase-lock-loop (PLL) [21,22] is applied to derive the phase (θ g ) of the pcc voltage fundamental positive-sequence component for grid synchronization, and the current reference amplitude ( + * I 1 ) is computed to regulate the average active power, as shown in Fig. 2. A stationary reference frame is applied in order to describe the threephase four-wire system, where the well-known Clarke transformation matrix is applied to transform three-phase quantities from the abc-coordinate system to the αβγ-reference frame.As pointed out in Fig. 2, the control algorithm is composed of a current controller in parallel with a voltage controller.The current control regulates the three-phase system output current fundamental positive-sequence component.The voltage controller is implemented by two sub-controllers, as highlighted by grey dash-dotted blocks, which correspond to voltage unbalance factor correction and pcc voltage harmonic compensation respectively.The proposed decoupled current and voltage controllers for regulating the αβ-quantities are the following [26]: The controllers for the γ-quantities are summarized as and K i are the series gain of the fundamental positive- sequence resonant (FPSR) filter, the proportional gain, and the integrating gain, respectively; − K res,1 is the series gain of the fundamental negative-sequence resonant (FNSR) filter, K res,1 0 is the series gain of the fundamental zero-sequence resonant (FZSR) filter, and K res h , is the series gain of the harmonic resonant filter for the three symmetrical components.are written as where ω 1 is the grid fundamental angular frequency, ω b1 is the band- width of the corresponding filter [29] considering the tolerance to grid frequency variation, and δ 1 and δ 2 are the damping factors for the corresponding resonant filters to avoid wind-up.The fundamental zerosequence and harmonic resonant filters are ) where k  is the set of harmonics to be compensated, and δ 3 is the corresponding damping factor for the harmonic resonant filter.It can be seen from Eq. ( 7) that in comparison to the FPSR filter which is applied to the current control loop, the FNSR filter has an additional series fundamental positive-sequence damping filter.It is done to make sure that the positive-sequence component is only controlled by the current control loop, not by the voltage support loop.

Sequence component modelling and analysis of the closed-loop converter system
In this section a frequency-domain equivalent model is derived for the closed-loop 3ϕ converter system with the control architecture in Fig. 2. The model groups the positive-and negative-sequence components into a complex quantity in the frequency-domain, making frequency-decoupling methods applicable for controlling of them.

Frequency-domain modelling of positive-and negative-sequences
Complex transfer functions are commonly used for the discussion of positive-and negative-sequence components [30,29,14].An arbitrary proper and stable complex filter H s ̲ ( ) with distinct poles can be ex- pressed as i n i 1 (10) where p ̲ i are distinct complex poles and N s ̲ ( ) is the complex numerator polynomial of the transfer function.As illustrated in [29], a set of positive-or negative-sequence sinusoidal αβ-quantities with an arbitrary frequency ω can be grouped into a complex variable where A̲ is a complex constant.So, x t ̲ ( ) αβ is a negative-sequence component when ω is negative, a DC component when ω is zero and a positive-sequence component when ω is positive.Its Laplace transform is αβ (12) Therefore, the output signal by feeding x t ̲ ( ) .
, the polynomial in Eq. ( 13) can be decomposed by (n + 1) partial fractions expressed as where the coefficient k̲ ω is The reverse Laplace transformation of the output is Therefore, in steady state the output be- comes Hence, it can be seen that analogously to a real-valued transfer function, a (stable) complex transfer function does not change the sequence and the frequency of the input signal, e.g. the output of a positive-sequence component is still sequence-positive with the same frequency.Moreover, as is clear from Eq. ( 17), the output y t ̲ ( ) αβ can be predicted by the characteristics of the complex transfer function H s ̲ ( ) and the input signal x t ̲ ( ) . Therefore, the analysis of the positive-and negative-sequence components can be merged in the frequency-domain.A positive frequency corresponds to a positive-sequence component, a zero frequency to a DC component and a negative frequency to a negative-sequence component.
It is worthwhile to mention that all complex transfer functions can be implemented as a two-by-two transfer function matrix whose entries are transfer functions with only real-valued polynomial coefficients.This is illustrated by the implementation of the complex transfer function H s ̲ ( ), as follows where Re{.} and Im{.} represent the respective operators to retrieve the real and imaginary parts of a complex transfer function.So, H s Re{ ̲ ( )} and H s Im{ ̲ ( )} are transfer functions with real-valued coefficients and thus can be implemented in a regular method.As is clear in Eq. ( 18), not all two-by-two transfer function matrix can be expressed by a complex transfer function but certain conditions should be met.

Open-loop system representation and sequence model
The illustration of a 3ϕ open-loop system starts from the per-phase representation.Neglecting the PWM switching-frequency harmonics, the two-level VSI is modelled as a controlled voltage source, as shown in Fig. 3.The LCL filter is assumed to be linear and therefore, the openloop converter in Fig. 3 is equivalent to a controlled voltage source with an internal impedance in Fig. 4, where we have The model in Fig. 4 is extended to each phase.Conventionally, the VSI and the LCL filter parameters are designed to be symmetrical, the per-phase natural gain and impedance are considered equal.Correspondingly, the 3ϕ open-loop converter can be described by abc o s abc o abc , (21) Applying the Clarke transformation to Eq. (21)we have in the αβγ re- ference frame Hereafter, the αβγ-quantities are divided into two groups for the symmetrical component discussion: γ-for zero-sequence and αβ-for positive-& negative-sequence.

Open-loop zero-sequence model
According to Eq. ( 22), the open-loop model for the zero-sequence component is

Open-loop positive-and negative-sequence model
The αβ-quantities are grouped into a single complex variable for the sake of simplified computation.The grouped αβ-quantities of the con- verter output current, the local voltage and the VSI output voltage are Therefore, according to the open-loop converter system model in Eq. ( 22), we have

Closed-loop system representation and sequence model
The equivalent model of the closed-loop converter system with the control architecture in Fig. 2 is derived in this subsection.

Closed-loop zero-sequence model
The control law for the γ-quantities in Fig. 2 is rewritten as with * i s ( ) being the γ-quantity of the current reference.Combining the controller in Eq. ( 26) and the plant model in Eq. ( 23) yields where k s ( ) is the corresponding equivalent gain of the controlled current source and Z s ( ) cv γ , the equivalent internal impedance ( Fig. 5), and we have Equivalently, the closed-loop model Eq. ( 27) can be represented by a Norton circuit, as shown in Fig. 5.It can be seen from Fig. 5 that the model is comprised of a controlled current source in parallel with an internal impedance.As indicated by Eq. ( 28), the gain is solely determined by the current controller C s ( ) while the impedance is codetermined by the current controller and the voltage controller C s ( ) . As is clear from Eq. ( 28), the current controller appears in the numerator of the impedance polynomial and the voltage controller in the denominator, which gives much freedom in designing the frequency response of the closed-loop converter.

Closed-loop positive-& negative-sequence model
Analogously to the zero-sequence, the closed-loop positive-& negative-sequence model can be derived by analysing the open-loop model in Eq. ( 25) and the control low for the αβ-quantities in Fig. 2. Therefore, we have where the coefficients are found to be It can be seen from Eq. ( 30) that the αβ-equivalent gain and impedance have a similar dependency on the current and voltage controllers as the γ-quantity., depending on the desired control objectives.Moreover, it is clear from Eq. ( 30) that the equivalent impedance equals the natural impedance Z s ( ) when both current and voltage controllers are set to zero.

Controller design based on gains and virtual impedances
In this section the performance of the proposed strategy is addressed using the derived model.Using the LCL filter parameters in Table 1 and the controller parameters in Table 2, the Bode plot of the equivalent gain k̲ cv αβ , is shown in Fig. 6; the plot for the γ-quantity is omitted because the γ-current reference is zero, as shown in Fig. 2. The Bode plots of the equivalent impedances for the γand the αβ-quantities are shown in Figs. 8 and 7 respectively, together with the natural impedance Z s ( ) o .As mentioned earlier, when the current and voltage controller are disabled, the converter equivalent impedance equals the natural impedance; therefore, the closed-loop controller design aims to overcome the converter's natural impedance (determined by the LCL filter parameters) and to modify the impedance in a desired way.For a harmonicunbalance-rejection control strategy, it is conventional to change the equivalent impedance to a high value for the entire frequency range, which is not the case in this paper because the converter is designed to operate in grid-supporting mode.

Fundamental positive-sequence current regulation
Since the primary objective of a grid-connected converter is to inject (absorb) desired active power into (from) the grid, the current control should dominate and a precise regulation of the fundamental positivesequence output current is expected.It can be seen from Fig. 6 that the αβ-gain has almost a unity gain and zero phase shift at the fundamental positive frequency.Moreover, as indicated in Fig. 7, the αβ-equivalent impedance at the corresponding frequency is boosted, compared to the converter natural impedance.This implies that the fundamental positive-sequence component of the converter output current will be predominately determined by the current reference, and is insensitive to the voltage applied at the local pcc.As is also shown in Figs.7 and 8, the magnitude of the γ-and αβ-equivalent impedance at the zero fre- quency is deliberately increased, compared to the converter natural impedance.This is done to prevent DC current flowing between the inverter and the grid.

Local harmonic sequence component compensation
As shown in Figs.7 and 8, the magnitude of the closed-loop equivalent impedance for both αβ-and γ-quantities is decreased at the selected harmonic frequencies in k  .By doing so a low-impedance path is created for the positive-, negative-, and zero-sequence harmonic currents generated by the local distorting loads.and 8, the γ-equivalent impedance is de- creased at the fundamental frequency and the αβ-equivalent impedance is reduced at the fundamental negative frequency.This creates a lowimpedance path for the fundamental zero-and negative-sequence components for the purpose of local voltage unbalance correction.

Discussion on the controller design
The control concept in this section aims at the reduction of the fundamental negative-and zero-sequence components and harmonics at the pcc voltage by decreasing the equivalent impedance of the gridconnected converter.By doing so, a low-impedance path (in relation to the grid impedance) is provided to sink harmonics or unbalanced components in the grid network.The fundamental positive-sequence, harmonic and fundamental negative-and zero-sequence components all contribute to the RMS current, and therefore to power loss, in the converter system.Thus, when applying the proposed strategy in a practical case, e.g.PV inverters, the available power/current rating of the grid-connected converter system should be taken into account.More specific, attention should be paid to the available power/current rating surplus that can be used for compensation.On-line adaptation of the control parameters, e.g. the resonant filter gains and thus the magnitude of the decreased equivalent impedance (for the target unbalanced/harmonic component), to the available power/current rating surplus is perceived as a solution.Optimal harmonic/unbalanced compensation subject to limited available power/current rating or other limitations is a relevant practical topic for future research.In this way, the compensation ability of a single GCC system is limited; however, the ability can be enhanced by coordinated local voltage support using multiple distributed grid-connected converter systems.

Setup description
A laboratory setup was realized to investigate the performance of the aforementioned local voltage support scheme.Its configuration is shown in Fig. 9. Its seven main parts are: two DC power supplies providing a neutral point, a ϕ 3 two-level VSI, an LCL filter, a ϕ 3 unbalanced resistive load, a programmable ϕ 3 current source acting as a harmonic load, a Spitzenberger type DM 3000 grid emulator and a dSPACE DS1104 real-time control system.A LeCroy WaveRunner-44Xi oscilloscope was used to capture the grid current waveforms.The controller was implemented on the DS1104 system, which was also sued to log the local pcc voltage and the converter output current measured by the digital controller.The nominal parameters of the converter system and current-voltage controller are in accordance Tables 1 and 2, respectively.The rest of the system parameters are summarized in Table 3.

Impact of unbalanced distorting load on the local voltage quality
The impact of the non-linear unbalanced load on the local voltage quality is assessed when the VSI is not operational (NO) by disabling all the inverter switches.In this case only the filter capacitor and the gridside inductor passively interact with the grid and the local loads.The corresponding measured waveforms of the grid current, local pcc voltage and converter output current are shown in Figs. 10 and 11.The per-phase harmonic spectra of the local pcc voltage and the converter output current are shown in Fig. 12.
Visible distortion of the local pcc voltage can be observed in Fig. 10, which is further confirmed by the per-phase harmonic spectrum in Fig. 12.On top of that it can be seen that the grid current is significantly distorted and asymmetrical, which can be explained by the unbalanced and harmonic loads connected to the pcc.The peak value of the phase-A current is higher than that of the other two phases, which is a consequence of a heavier load connected to phase-A (in Fig. 9 (b)), as mentioned in Table 3.Moreover, both fundamental and harmonic components are present in the converter output current, as shown in Figs.11 and 12.This is expected because the passive LC filter is connected to local pcc despite of the fact that the VSI is not operational.

Current control loop activated only
This subsection shows the experimental results when only the current control (CC) loop in Fig. 2 is activated.The waveforms are omitted because limited information can be retrieved.The respective per-phase harmonic spectra of the local pcc voltage and the converter output current in this case are displayed in Fig. 13.
It can be seen that compared to Fig. 12, the per-phase voltage amplitude of the low-order harmonic content ( ⋯ {3, , 13}) in Fig. 13 is slightly reduced.However, the high-order harmonic content ({15, 17, 19}) increases, and the converter current has an increased low- order and reduced high-order harmonic content.This can be explained by the fact that when the converter operates in 'CC' mode the converter equivalent impedance is only altered for the fundamental positive-sequence component but kept aligned with the natural impedance for the other frequencies.As shown in Figs.7 and 8 the equivalent impedance at high-order harmonic frequencies is relatively high, especially around order {17}, and low at low-order harmonics, thus the converter does not interact with high-order harmonics while it sinks low-order harmonics.

Enabled harmonic sequence component sinking
This subsection shows the experimental results when harmonic sequence compensation is enabled.In this mode, the converter is programmed for fundamental positive-sequence current regulation and harmonic sinking (CC + HS).The corresponding measured per-phase local pcc voltage and converter output current spectra are Fig.14; waveforms are omitted because of a similar consideration as that in 'CC' Fig. 11.Measured waveforms of the per-phase local voltages and the converter output currents when the converter is not operational.In comparison to Fig. 13 the per-phase local pcc voltage {3, 5, 7, 9, 11, 13} harmonic amplitudes are reduced in Fig. 14.This is because the equivalent impedance at these harmonic frequencies is made smaller, providing a low-impedance path for the selected harmonics, as indicated in Figs.7 and 8. Correspondingly, a significant increase of 3 rd harmonic content can be observed in the converter output current in Fig. 14 with comparison to the spectrum in Fig. 13.

Enabled fundamental-zero and fundamental-negative sequence correction
Now the correction of unbalanced voltage is enabled.In this mode, the converter is programmed for current control, harmonic sinking, and voltage unbalance correction (CC + HS + VUC).Here, the voltage unbalance correction includes the reduction of the fundamental negativesequence and zero-sequence components.The measured waveforms and per-phase harmonic spectra are shown in Figs.15-17.Compared to the waveform in Fig. 10, significant reduction in the current asymmetry can be observed in Fig. 15, which is due to the increased phase-A current, as shown in Fig. 16.Moreover, negligible difference can be observed in the current harmonic content between Figs. 17 and 14.However, significant current unbalance is observed in Fig. 17 because now the converter is programmed for unbalance correction and thus more current is drawn from the converter phase-A to compensate for the heavier load that is connected.
Table 4 summarizes the per-phase voltage total harmonic distortion (THD) and the (negative-sequence) local voltage unbalanced factor (VUF) in the aforementioned cases.It can be seen that the per-phase local pcc voltage THD is reduced when harmonic-sinking is enabled, and VUF decreased when the voltage-unbalance correction is activated.

Sequence harmonic spectrum analysis
In this section, the zero-sequence and the grouped positive-& negative-sequence harmonic spectra of the local pcc voltage, and the converter output current in the aforementioned three modes are presented.The zero-sequence harmonic spectrum can be obtained by using Fast-Fourier-analysis of the γ-quantity.The harmonic spectrum of the grouped positive-& negative-sequence components can be analysed by expressing the αβ-local voltage quantities in the time-domain [29] as where V n is the amplitude and φ n the phase angle of the n th harmonic positive-or negative-sequence component when = t 0.
where d̲ h and q ̲ h are orthogonal negative-sequence components when h is negative, DC components when when h is zero, and positive-sequence components when h is positive.Associating Eqs. ( 31) and (32) yields that where " • " denotes the dot product of two complex quantities, and = T π ω 1 2 1 is the fundamental period.Accordingly, the amplitude of the h th harmonic positive-or negative-sequence component can be found from Following the method presented above the harmonic spectra of the grouped positive-& negative-sequence components of the local pcc voltage and the converter output current in the 'CC', 'CC + HS' and 'CC + HS + VUC' cases are presented in Fig. 18.The corresponding γ-harmonic spectra are shown in Fig. 19.
It can be seen from Fig. 18 that, except for the order + { 11} har- monic, the amplitudes of the αβ-quantity voltage harmonics of the order ± ± ± ± − ± { 3, 5, 7, 9, 11, 13} are reduced in 'CC + HS' and 'CC + HS + VUC', in comparison to the mode 'CC'.When the voltage unbalance correction is enabled, significant reduction of the local voltage fundamental negative-sequence component in is observed by comparing 'CC + HS' and 'CC + HS + VUC', which is in accordance with Fig. 7 because a low-impedance fundamental negative-sequence path becomes available.Note that the converter output current has a significant increase in the fundamental negative-sequence component when 'VUC' is activated, as shown in Fig. 18.
Consistent to the voltage unbalance correction trend in Fig. 18, the fundamental zero-sequence component in the local voltage is significantly reduced and that in the converter output current is increased when 'VUC' is activated, as shown in Fig. 8. Slightly different from the harmonic trend in Fig. 18, only the 3 rd voltage harmonic content in the γ-quantity is significantly reduced when the 'HS' is turned on, as can be seen from Fig. 19.This could be explained by the non-linearity of the VSI and the switching noise present in the measurement network.
During the modelling of the closed-loop converter system, the VSI is considered as a linear controlled voltage source, which is not always the case considering dead-time distortion.

Conclusion
A local voltage support control architecture for a three-phase fourwire voltage-source-inverter system is discussed in this paper.On top of conventional active power transfer, the designed converter system can be programmed for local voltage symmetrical harmonic sinking and fundamental negative-and zero-sequence component correction.A sequence impedance model is derived in this paper for intuitive controller design and to illustrate the working principle of the control architecture.Firstly, the αβγ-components are divided into αβ-and γ-quantities.By doing so the regulation of the zero-sequence component is separated from the control of the positive-& negative-sequence component, making asymmetrical control possible.Secondly, the αβ-quantities are grouped into a complex quantity, thus a straightforward frequency-domain-based analytical method can be applied to the analysis of these components.Consequently, the regulation of the fundamental positive-sequence component and the compensation of the other components can be correspondingly decoupled in the frequency domain.The voltage support functionality of the closed-loop converter can thus be achieved by altering the corresponding sequences' equivalent impedance at specific frequencies, e.g. an increased equivalent impedance for the fundamental positive-sequence, a reduced one for the harmonic sequence, and a decreased one again for the fundamental negative-and zero-sequence components.
Next to the demonstration where the voltage-source-inverter is disabled, three experiments were carried out with different settings of the controller.In the first case the converter is programmed to operate in current control 'CC' only, which corresponds to a boosted fundamental positive-sequence equivalent impedance while the impedance for the other component remains untouched.Consistent to the model analysis, the local pcc voltage and the grid current are still significantly distorted and unbalanced in this case.Subsequently, in the second case harmonic sinking is enabled for the symmetrical components, which results in reduced per-phase local voltage harmonic content at the selected frequencies with comparison to the 'CC' case.Thereafter, the local voltage unbalance correction 'VUC' is activated as well, which corresponds to a reduced equivalent impedance for the fundamental zero-and negative-sequence components.Experiments show that the asymmetry in the per-phase grid current is significantly reduced, which agrees well with the model analysis.Consistently, the sequence harmonic analysis shows that the voltage fundamental zero-and negativesequence components are reduced when 'VUC' is activated.In conclusion, model analysis and experiments show that the proposed voltagesupport strategy helps to reduce the per-phase voltage THD and the voltage unbalanced factor, including fundamental zero-and negativesequence components.Since the control architecture is based on only local measurements, it is feasible to upgrade software of existing gridconnected converters for add-on local voltage support.

Declaration of Competing Interest
The author(s) declare(s) that there is no conflict of interest.

Fig. 2 .
Fig. 2. Block diagram of the parallel current-voltage control architecture in the stationary reference frame.

1 and Z 2
are the impedances of the inverter-side inductor, the filter capacitor and the grid-side inductor, respectively.It can be seen that the gain k s ( ) and the equivalent open-loop impedance Z s ( ) o are determined by the LCL filter parameters and hence, they are called the natural gain and the natural impedance, respectively.

Fig. 5 .
Fig. 5. Circuitry equivalent representation of the closed-loop converter for the γ -quantities.
Due to the limited computation power of the micro-controller the sampling rate and the switching frequency are chosen as = f

Fig. 8 .
Fig. 8. Bode plot of the closed-loop γ -equivalent impedance together with the natural impedance of the LCL filter.

Fig. 9 .
Fig. 9. Experimental implementation diagram of a three-phase grid-connected VSI system with local harmonic unbalanced loads: (a) connection block diagram (b) photograph of the laboratory system.

Fig. 10 .
Fig. 10.Measured waveforms of the phase-A local voltage and the per-phase grid currents when the VSI is not operational.

Fig. 12 .
Fig. 12. Measured per-phase harmonic spectra when the VSI is not operational (NO): (a) the local pcc voltage, (b) the converter output current.

Fig. 13 .
Fig. 13.Measured per-phase harmonic spectra when the VSI is in 'CC' mode: (a) the local pcc voltage, (b) the converter output current.

Fig. 14 .
Fig. 14.Measured per-phase harmonic spectra when the controller is in 'CC + HS' mode: (a) the local pcc voltage, (b) the converter output current. case.

Fig. 15 .Fig. 16 .
Fig. 15.Measured waveforms of the phase-A local voltage and the per-phase grid currents when the controller is in 'CC + HS + VUC' mode.

Fig. 17 .
Fig. 17.Per-phase harmonic spectra when the controller is in 'CC + HS + VUC' mode: (a) the local pcc voltage, (b) the converter output current.

Fig. 18 .
Fig.18.Grouped positive-& negative-sequence harmonic spectrum (from the αβ-quantities) in the three cases: (a) the local pcc voltage, (b) the converter output current.

Fig. 19 .
Fig. 19.Zero-sequence harmonic spectrum (from the γ -quantities) in the three cases: (a) the local pcc voltage, (b) the converter output current.

Table 1
Three-phase LCL filter parameters.

Table 2
Control parameters.

Table 3
Three-phase grid and load parameters.

Table 4
Summary of the local voltage THD and unbalance factor in the aforementioned three cases.