IET Renewable Power Generation

Research Article

Design methodology of a passive damped ISSN 1752-1416

Received on 17th April 2017
Revised 10th July 2017
modified LCL filter for leakage current Accepted on 5th August 2017
E-First on 10th October 2017
reduction in grid-connected transformerless doi: 10.1049/iet-rpg.2017.0256
www.ietdl.org

three-phase PV inverters
Julian C. Giacomini1 , Leandro Michels2, Humberto Pinheiro2, Cassiano Rech2
1Federal Institute of Education, Science and Technology Farroupilha, Panambi, Brazil
2Power Electronics and Control Research Group, Federal University of Santa Maria, Santa Maria, Brazil
E-mail: [email protected]

Abstract: Although grid-connected transformerless photovoltaic (PV) inverters present higher efficiency and power density
compared with inverters with a transformer, the leakage current caused by the inverter common-mode voltage introduces
several problems. Among the techniques to reduce the leakage current, the modified LCL (MLCL) filter with passive damping is
an effective and simple solution. However, the classical design of the filter damping resistance is not adequate for ensuring both
proper leakage current attenuation and control system stability. Therefore, this study proposes a methodology to design the
resistance in a low-loss passive damping structure applied to the MLCL filter. In addition to the conventional specifications for
LCL-type filters, this study includes the leakage current limit in the design procedure. Simulation and experimental results for a
10 kW PV inverter show the damping resistance impact on the leakage current. The results related to the efficiency and grid
inductance variation are also presented. Therefore, it is possible to conclude that the proposed design methodology is very
useful for obtaining a damping resistance that ensures control system stability and a leakage current in conformity with PV
standards.

1 Introduction high-frequency components of the voltage on the parasitic

capacitance [15–17]. However, the increased number of
Transformerless photovoltaic (PV) systems have been used due to components usually increases the complexity and cost as well as
several advantages over inverters with high or low-frequency reduces the overall efficiency of the PV inverter. Topology
isolation transformer, such as high efficiency, low cost, and modifications can also be found in single-phase topologies with
reduced size/weight [1–3]. However, the lack of galvanic isolation very low leakage current [18]. Another alternative to reduce the
causes the leakage current circulation between parasitic leakage current is to use converters with the DC bus capacitor's
capacitances of the PV grounded array and the grid [4, 5]. These middle point connected to the neutral point of the grid. However, it
parasitic capacitances are located between the PV cells and panel is crucial that the neutral line has very low inductance in order to
structure, and their values depend on several factors, such as the not increase the leakage current [19, 20].
surface area of PV array and its mounting structure, dust, and On the other hand, the use of a modified LCL (MLCL) filter in
manufacturing technology of the photovoltaic cells. The leakage the grid connection is the simplest solution compared with other
current circulation is mainly caused by the amplitude and leakage current reduction methods [21, 22]. This is because the
frequency variation of the inverter common-mode voltage, MLCL filter only requires an additional connection in the classical
resulting in electromagnetic interference, safety issues, waveform LCL filter, where the common point of the output filter capacitors
distortion of the grid currents, and higher losses [6]. Due to this, is connected directly to the DC bus central point. This connection
there are safety requirements defined in standards to protect against attenuates the high-frequency components of the voltage across the
excessive leakage current values [7, 8]. The standards DIN VDE V PV parasitic capacitance. Furthermore, the MLCL filter does not
0-126-1-1:2006 and IEC62109-2:2011 require the installation of a require modifications in the modulation technique and inverter
residual current monitoring unit that must disconnect the inverter topology, keeping the maximum inverter modulation index in the
from the grid within a predefined time when a continuous residual linear region and not increasing the number of semiconductors.
current exceeds 300 mA rms. Nevertheless, grid-connected inverters with LCL-type filters, as
Therefore, some techniques to reduce the magnitude of the the MLCL structure, require the use of a damping technique
leakage current have been proposed [9–21]. One of these (passive or active), since the filter resonance may lead the grid
techniques is to use modified modulation strategies to change the current control loop to instability. Among the damping techniques,
behaviour of the common-mode voltage of the inverter [9–14]. the passive damping is considered the simplest, most used and low
Most of these techniques are based on space vector modulation and cost alternative to reduce the filter resonance peak compared with
they consist in selecting a suitable set of switching vectors to the high cost and complex active control techniques. In this case,
reduce the high-frequency components of the common-mode the damping resistance is designed to obtain a desired attenuation
voltage. However, they penalise the inverter modulation index and of the filter resonance peak, improving the stability. Moreover,
may increase the switching and magnetic core losses due to the regarding the leakage current reduction of the PV system, the use
reduced number of vectors used to synthetise the output voltage. A of passive damping in the MLCL filter is also an alternative
space vector modulation technique was proposed in [13] using two provided that the damping resistance has an adequate design [23].
medium vectors with an adaptive neutral point voltage to allow the Nevertheless, classical design techniques of this damping
inverter operation with unitary modulation index and constant resistance [24–26] do not take into account its effect on the leakage
common-mode voltage. Nevertheless, it requires the use of a current. Therefore, the damping resistance is an important
bidirectional half-bridge DC/DC converter to generate the adaptive parameter to be considered in case of the MLCL filter and it must
neutral point voltage, increasing the complexity of the converter.
Modified inverter topologies have also been proposed to reduce the
IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777 1769
© The Institution of Engineering and Technology 2017
Fig. 1  Transformerless grid-connected PV systems
(a) With the classical LCL filter, (b) With an MLCL filter, (c) Three-level PV inverter connected to the grid through an MLCL filter with passive damping

be properly designed in order to not compromise the filter voltage in the PV array. A capacitor (Cp) is included at the DC bus
attenuation and the leakage current reduction. negative terminal to model the parasitic capacitance of the PV
Consequently, the main challenge is to determine the damping system [9, 20]. The grid impedance is considered purely inductive
resistance values in which the current control loop is stable and the and represented by the inductor Lg. The PV inverter modulation
leakage current is below the standard limits. Therefore, this strategy is the pulse width modulation (PWM) with phase
manuscript proposes a methodology to design the resistance in a disposition carriers that presents an easy implementation provided
low-loss passive damping structure applied to MLCL filter to by the carrier-based concept [28]. A common-mode signal is used
comply with main PV standards [7, 8]. Differently from most LCL- to extend the modulation index of the NPC inverter and for the DC
type filter methodologies, the proposed approach considers the bus capacitors' voltage balance (C1 and C2). A low-loss passive
leakage current limit in the design procedure in addition to the damping is used to reduce the MLCL filter's resonance peak,
usual specifications. The methodology is applied to a three-phase avoiding the control system instability. This damping scheme keeps
PV inverter and it is based on the trade-off between leakage current the high-frequency attenuation of the filter even by increasing the
reduction and control system stability. In addition, the passive damping resistance Rd [29]. Moreover, the damping effect, as well
damping losses and the impact of the grid impedance on the
as the power losses, increases with Cd/Cn ratio.
passive damping design are also analysed. The proposed design
methodology can be extended to other grid-connected
transformerless three-phase PV inverter topologies that employ the 3 Proposed passive damping design technique
MLCL filter, including another passive damping schemes and
3.1 Step 1: design of L1, L2, and Cf
control techniques.
The main contributions of this manuscript are: (i) a design The first step is the design of the main filter parameters, i.e. L1, L2,
methodology for the damping resistance in the passive damped and Cf=Cd+Cn. Nevertheless, it is not addressed in this paper,
MLCL filter to keep the current control loop stable and the leakage
since conventional techniques can be applied. For instance, L1 can
current below the standard limits; (ii) a systematic and simple
procedure to calculate the leakage current level of the PV inverter. be designed from the desired inverter output current ripple [30].
Similarly, the filter capacitance Cf can be designed from the
inverter reactive power consumption at the fundamental frequency
2 Topology description [31]. A typical range of 5–10% of the inverter reactive power
A typical grid-connected transformerless three-phase PV system consumption at the fundamental frequency is generally adopted.
using an LCL filter is shown in Fig. 1a. It is well known that the Furthermore, the Cf value is also related to the leakage current
LCL filter presents a smaller size and a better harmonic attenuation reduction in the MLCL filter, since this capacitor offers a low
than an L filter. Nevertheless, a low leakage current cannot be impedance path for the high-frequency harmonic current
achieved with this classical LCL filter without a proper leakage components. Increase in the Cf value is desirable for better leakage
current reduction technique. On the other hand, the use of the current reduction. Nevertheless, special attention should be taken
MLCL filter, as shown in Fig. 1b, is the simplest solution to reduce to avoid low-frequency resonance excitation [32]. Regarding the
the leakage current in three-phase PV inverters [21, 22]. The inductance L2, it can be designed based on the grid current total
MLCL filter performs two tasks in the conversion system: harmonic distortion (THD) [33] or based on the ripple attenuation
harmonic attenuation of the grid currents and leakage current criteria [31].
reduction. It has the same components of the classical LCL-type Consequently, a total capacitance of 25 µF was adopted to
filters, except by the connection of the filter capacitors' common balance leakage current reduction and the reactive power
point to the central point of the DC bus. This connection produces consumption, which results in fundamental reactive power
a low-pass filter and attenuates the high-frequency components of consumption around 10% at rated power. This filter capacitance
the voltage on parasitic capacitance Cp, reducing the leakage was divided in the following proportions to obtain a balance
current of the PV system. between damping and losses: Cd is equal to 0.6Cf and Cn is equal
The transformerless three-phase neutral point clamped (NPC) to 0.4Cf [29]. The inductances L1 and L2 were designed to meet a
inverter considered in this paper is shown in Fig. 1c. NPC topology
grid current THD <2% of the fundamental current at inverter rated
generates line-to-line voltages with five levels, improving the
power. The resulting inverter and filter parameters are described in
waveforms quality [27]. Moreover, the voltage across the
Table 1. A parasitic capacitance of 1.25 μF was adopted in this
semiconductors is equal to half of the DC bus, allowing a high
1770 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777
© The Institution of Engineering and Technology 2017
analysis according to the typical values for crystalline silicon k1 = CdRd
modules [34]. The next steps consider the parameters of Table 1, Cp Cp
but it can be extended for other set of parameters. k2 = L1 Cd + Cn + + L2 + Lg
3 3
3.2 Step 2: analysis of the impact of Rd on the leakage Cp L1 + L2 + Lg
k3 = CdRd CnL1 +
current 3 (6)
It is necessary to obtain the common-mode equivalent circuit of the Cd + Cn
k4 = CpL1 L2 + Lg
three-phase PV inverter to evaluate the leakage current. Many 3
works have addressed the achievement of the common-mode CdCpCnL1 L2 + Lg Rd
circuit and their respective equations [5, 9, 20]. Fig. 2 shows the k5 = .
3
common-mode equivalent circuits of the three-phase PV inverter
using a single L filter (see Fig. 2a) and using the passive damped
Consequently, the rms value of the leakage current (ip,rms) is
MLCL filter [see Fig. 2b]. The models in Fig. 2 considers that filter
elements are equal for all phases and the DC bus capacitors (C1 evaluated from (4), (5), and using
and C2) are replaced by two equivalent voltage sources with a hmax hmax
1 1
2 h∑ 2 h∑
constant value of Vdc/2. The grounding impedance and the ip, rms = ip jω1h
2
= vcmv jω1h Gcm jω1h 2
(7)
resistance of the filter elements were not included, representing the =1 =1
worst case from the point of view of the leakage current magnitude
[5]. The voltage source vcmv is the inverter common-mode voltage, where h is the harmonic order related to the angular fundamental
which is defined as [9, 20] frequency ω1 and hmax is the maximum harmonic considered. The
unknown variables in (7) are only the damping resistance Rd and
vao + vbo + vco the rms value of the leakage current ip,rms, since the filter
vcmv = . (1)
3 parameters (L1, L2, Cd and Cn) were defined previously. In
addition, the common-mode voltage spectrum vcmv (jω) shown in
The amplitude variations in the vcmv are responsible by the
Fig. 3a is obtained by the fast Fourier transform of (1) considering
generation of the leakage current [5, 9–11, 20]. From the a time interval equal to the grid period. Furthermore, the inverter
equivalents circuits of Fig. 2 and neglecting the effect of the leg voltages (vao, vbo, and vco) are obtained from the comparison of
resistive elements, it is possible to obtain the high-frequency
the modulating signals with the triangular carriers. As a
attenuation gain of the leakage current for the simple L filter and
consequence, Fig. 3b shows the behaviour of the rms value of the
for the MLCL filter, respectively:
leakage current obtained from (7) as a function of the damping
1 resistance, considering Lg = 0 µH, hmax = 1024, and the parameters
AttenL ω = (2) of Table 1.
ωLT
By increasing Rd, the capacitive branch impedance also
3 increases, which results in a higher leakage current value as can be
AttenMLCL = (3) observed in Fig. 3b. Therefore, the maximum value for the
ω3L1 L2 + Lg Cf
damping resistance (Rd = Rd,max) is given by the crossing point
between the standard line and the leakage current curve. Note that
where LT=Lf+Lg. Note that (3) has a third-order gain attenuation
Rd,max = 4.0 Ω for the parameters shown in Table 1. On the other
while (2) has a first-order gain attenuation. Therefore, it is verified
hand, the damping resistance must not be so small to preserve the
that the MLCL filter provides a better attenuation of the common-
control system stability, which implies that the minimum value for
mode voltage harmonics in comparison with a simple L filter. This
Rd must be chosen to comply with the stability criteria of the
is the reason for the use of the MLCL filter in many applications
[21–23]. Fig. 2b shows how the capacitive branch (Cn–Cd–Rd) control system, as presented in the next step.
creates a low-pass filter, thereby reducing the leakage current
circulation through the grid. 3.3 Step 3: analysis of the impact of Rd on the system
Nevertheless, the attenuation of the MLCL filter is affected by stability
the damping resistance used in the MLCL filter. Higher values for
The Rd design is also related to the control system stability, since it
Rd reduce the effectiveness of the leakage current filtering since the
affects the MLCL filter resonance peak. Moreover, the stability
impedance of the capacitive branch is modified. Generally, the
analysis is dependent on the control technique used. In this way, a
design of the damping resistance is made by adopting criteria such
digital proportional-integral (PI)-based control was adopted in a
as harmonic attenuation, power losses, and/or control system
synchronous reference frame (dq) for the grid current control [35],
stability. However, the leakage current requirement is included in
as illustrated in Fig. 4a. The DC bus voltage control loops are not
the present paper. Therefore, the leakage current behavior in
addressed here, since their dynamics are slower than the current
function of the damping resistance Rd has to be found.
control loop. The duty cycle d0 is the common-mode signal used to
The leakage current spectrum can be obtained from the
extend the modulation index of the inverter [28]. The discrete-time
common-mode voltage spectrum vcmv(jω) and from the transfer
equivalent control system for the current control loop useful for the
function Gcm(jω) of the equivalent circuit of Fig. 2b, resulting in stability analysis is represented in Fig. 4b.
The control system shown in Fig. 4b is valid for both axes (d
ip jω = vcmv jω Gcm jω (4) and q). The unit delay z−1 is included due to digital implementation,
where Ts is the sampling period. Furthermore, the zero-order-hold
where Gcm(jω) is the complete common-mode circuit transfer (ZOH) models the PWM dynamic. The dq transformation effect
function: was not taken into account in Fig. 4b since the switching frequency
is sufficiently higher than the fundamental frequency [36]. The
jωCp jωCdRd + 1 continuous-time filter model Gid is given by
Gcm jω = (5)
jω 5k5 + jω 4k4 + jω 3k3 + jω 2k2 + jω k1 + 1
V dc CdRds + 1
and the coefficients k1 to k5 are: Gid s = (8)
s4 p4 + s3 p3 + s2 p2 + sp1

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777 1771
© The Institution of Engineering and Technology 2017
where p1 = L1 + L2 + Lg, p2 = CdL1Rd + RdCd L2 + Lg , The decrease in the damping resistance tends to reduce the
p3 = L1Cf L2 + Lg , and p4 = CdCnL1Rd L2 + Lg . relative stability of the control system, as shown in Fig. 5a.
The filter discrete-time model can be obtained from the Z- Therefore, the impact of Rd on the closed-loop poles of the current
transform, taking into account the ZOH model [37] and adopting a control loop is shown in Fig. 5b. As expected, the reduction of the
sampling frequency equal to twice the switching frequency (fs =  damping resistance tends to move the closed-loop poles to outside
2fsw). In this sense, considering the filter discrete-time model, the of the unity circle (instability region). The minimum resistance (Rd 
PI controller parameters were designed using the w-plane = Rd,min) that can be used is given by the lowest value that ensures
technique [37] to meet a gain crossover frequency of fGC = 400 Hz that the closed-loop poles are within the unity circle. In Fig. 5b,
with a phase margin of γ = 60°, which results in kp = 0.0042857, a  Rd,min = 0.3 Ω in order to ensure current control loop stability. One
= 1.02441 and b = −0.97558. These performance parameters (fGC can verify from this analysis that Rd must be chosen between the
and γ) are not significantly affected by Rd, since the gain crossover minimum and maximum value given by the stability limit and by
the leakage current limit, respectively.
frequency is sufficiently lower than the filter resonance frequency.
In other words, the PI controller design could be made considering
the output filter as a purely inductive filter (L1 + L2), since the filter 3.4 Step 4: analysis of the impact of the grid impedance
capacitor Cf can be neglected for frequencies lower than the Leakage current and the control system stability are affected by the
resonance peak [31]. These statements can be verified from the grid impedance because it modifies the frequency response of the
open-loop frequency response of the control system, shown in common-mode circuit and changes the position of the closed-loop
Fig. 5a. Similar current controller is obtained if the symmetrical poles of the current control loop. Consequently, the analysis of
optimal method was applied [38]. these two requirements for distinct grid impedance values is
required. The fifth order common-mode circuit shown in Fig. 2b is
considered for the evaluation of the grid impedance impact on the
leakage current. Neglecting the influence of the damping resistance
Table 1 Inverter and filter parameters and considering Cf ≫ Cp, the two resonance peaks of this circuit
Parameter Value are located at the following frequencies (in Hertz):
inverter output power 10 kW
grid line voltage/frequency 380 V/60 Hz 1
f1 ≅ (9)
DC bus voltage 700 V 2π L1Cf
inductance L1 1100 μH
1
inductance L2 200 μH f2 ≅ .
Cp (10)
capacitance Cd 15 μF 2π L2 + Lg
3
capacitance Cn 10 μF
capacitance Cp 1.25 μF Since Cf ≫ Cp, it results that f1<f2. Note that f2 can change in
switching frequency fsw 7.68 kHz practice, since the grid inductance is uncertain. For better
understanding, the gain frequency response of the common-mode

Fig. 2  Common-mode equivalent circuits

(a) Three-phase inverter with a single L filter, (b) Three-phase inverter with the MLCL filter

Fig. 3  Impact of the damping resistance on the leakage current

(a) Spectrum of the common-mode voltage, (b) rms value of the leakage current (Lg = 0 µH)

1772 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777
© The Institution of Engineering and Technology 2017
Fig. 4  Control system of the transformerless PV inverter
(a) Grid current control in a synchronous reference frame, (b) Block diagram of the grid current control loop

Fig. 5  Impact of the damping resistance on the system stability (Lg = 0 µH)

(a) Open-loop frequency response of the control system, (b) Closed-loop poles of the current control loop

circuit Gcm(jω) for different grid inductances is shown in Fig. 6a μH, there is not a feasible range for the choice of Rd, since
considering Rd=0.3 Ω. One can observe that f2 tends to reduce with Rd,min=Rd,max = 1.3 Ω. In this case, it is necessary to return to step
the increase of the grid inductance and, as a consequence, it may 3 and to reduce fGC to improve the system stability, making the
match with the switching frequency components of vcmv, resulting reduction of Rd,min possible. However, the control system
in an increase in the leakage current. On the other hand, f1 is well performance will be deteriorated with the current loop bandwidth
defined, since it only depends on well-known filter parameters. reduction. Similarly, if, for some Lg, the leakage current remains
Nevertheless, special attention should be taken to avoid the above the standard limit regardless of Rd, such as Lg = 450 μH in
excitation of f1 if a low-frequency common-mode signal is used to Fig. 6b, it is necessary to return to step 1 for the filter redesign to
extend the inverter modulation index [32]. comply with the leakage current standard.
From this analysis, the rms value of the leakage current is
obtained in function of Rd using (1), (4)–(7) for different grid
4 Rd impact on the damping losses
inductance values, as shown in Fig. 6b. As expected, the leakage
current increase with Lg for the same Rd. As a consequence, Rd,max The proposed design process is based on leakage current and
tends to reduce, and the choice of Rd becomes more restrictive. control system stability, which results in a maximum and a
Similarly, the closed-loop poles of the control system for different minimum damping resistance value. Nevertheless, it is also
values of Rd and Lg are shown in Fig. 6c. One can observe that the important to verify the damping resistance losses, which is an
inherent concern in these schemes. The losses are composed of
increase in Lg moves the closed-loop poles to the outside of the
fundamental frequency and switching frequency harmonic
unit circle (instability region). Consequently, Rd,min increases and components [31]. From the spectrum of the damping resistance
the solution range for Rd is reduced. current, the total losses are obtained by
From Figs. 6b and c, one can note that Lg values exist wherein it
is not possible to both comply with the leakage current standard
and ensure the control system stability. For instance, if Lg = 300 

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777 1773
© The Institution of Engineering and Technology 2017
Fig. 6  Impact of the grid impedance on system performance
(a) Gain frequency response of the common-mode circuit for different grid inductances (Rd = 0.3 Ω), (b) Leakage current for different grid inductances and different damping
resistances, (c) Position of closed-loop poles for Lg varying from 0 to 450 µH with steps of 150 µH

hmax b show the grid currents and the leakage current for distinct values
3
2 d h∑
2
PR = R iRd jω1h (11) of damping resistance.
=1 Fig. 7a shows that the rms value of the leakage current is 334.9 
mA (above the standard limit) if a damping resistance equal to 7.4 
where iRd is the damping resistance peak current (one phase) at Ω is used. On the other hand, if a resistance of 1.0 Ω is used, the
harmonic order h and PR are the total damping losses. Expression leakage current level complies with the standard, since ip,rms = 
(11) considers that all phases have the same current (in module). 210.5 mA, as illustrated in Fig. 7b. These results show the impact
The current iRd can be found from: of the damping resistance on the leakage current, confirming the
theoretical analysis. In addition, Figs. 7c and d show the impact of
iRd jω = L2vao jω + L1vgrid jω Gd jω + vcmv jω Gc jω . (12) the damping resistance on the control system stability. According
to the theoretical analysis of Section 3, the minimum resistance
One can observe from (11) to (12) that the losses are affected by value to ensure control system stability, with Lg = 0 µH, is 0.3 Ω.
the harmonic spectrum of the voltage vao, grid voltage vgrid and Fig. 7c demonstrates that by using Rd = 0.3 Ω the control system is
common-mode voltage vcmv. The transfer functions Gd and Gc are really close to instability. On the other hand, using Rd = 0.35 Ω the
control system is stable, as can be verified in Fig. 7d.
Cd
Gd jω = (13)
jω 4 p4 + jω 3 p3 + jω 2 p2 + jω p1 5.2 Experimental results
Experimental results were obtained considering the parameters
L1Gd jω (CdRd(jω) + 1) presented in Table 1. The control algorithm was implemented using
Gc jω = . (14)
jω CdCnRdL1 + jω 2 Cd + Cn L1 + jω CdRd + 1
3
the digital signal processor TMS320F28335 from Texas
Instruments. The PV system was emulated with a controlled DC
Considering the parameters in Table 1 and hmax = 192, the total power source and the converter was connected to the utility grid.
power losses for Rd,max = 4.0 Ω are 43 W, which represents 0.43% The leakage current was measured with the precision power
of the inverter rated power. In fact, the power losses are reduced analyser Yokogawa WT1800. The parasitic capacitance of 1.25 μF
for lower damping resistance values, but the control system gets was modelled by the connection of a polypropylene capacitor at
closer to the instability limit, as presented in Section 3. The negative terminal of the DC bus. The prototype of the PV inverter
complete theoretical curve of the damping losses is presented in is shown in Fig. 8.
Section 5 together with the experimental losses. Fig. 9a shows the current injected into the grid (phase a) and the
multilevel line-to-line voltages generated at inverter output at rated
conditions using a damping resistance equal to 1.0 Ω. Moreover,
5 Results Fig. 9b shows the satisfactory dynamic performance of the current
5.1 Simulation results control, where a grid current step from 5 to 10 kW was applied.
The maximum efficiency obtained was 96.2% and the current THD
Some simulation results are presented in this section to validate the was around 2.5% at rated power with a power factor of 0.99.
proposed methodology. The parameters used in simulations were Experimental results for the leakage current (ip), grid voltage
the same as shown in Table 1 and a parasitic capacitance of 1.25 μF (vgrid) and grid current (igrid) of phase a are illustrated in Figs. 9c
was connected at the negative terminal of the DC bus. Figs. 7a and

1774 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777
© The Institution of Engineering and Technology 2017
Fig. 7  Simulation results
(a) Grid currents and leakage current (Lg = 0 µH, Rd = 7.4 Ω), (b) Grid currents and leakage current (Lg = 0 µH, Rd = 1.0 Ω), (c) Grid currents and modulating signals (Lg = 0 µH, Rd 
= 0.3 Ω), (d) Grid currents and modulating signals (Lg = 0 µH, Rd = 0.35 Ω)

increased to 406.3 mA, an elevation of 15.9%. Therefore, these

results confirm the influence of the grid inductance on the leakage
current.
The leakage current levels for several values of damping
resistance and grid inductance are shown in Fig. 10c. The increase
in the rms value of the leakage current according to the damping
resistance can be observed, since the MLCL filter effectiveness is
affected by it. Moreover, it is also possible to verify in Fig. 10c the
effect of the grid inductance on the leakage current levels. The
experimental results presented a good correspondence with the
theoretical analysis, validating the proposed design methodology.
Fig. 10d presents experimental and calculated damping losses
and the efficiency of the grid-connected PV inverter operating at
rated power for several values of damping resistance. As expected,
the damping loss increases with the value of the damping
resistance. However, it can be verified that the power losses
Fig. 8  Prototype of the PV inverter
represent a small portion of the overall power processed by the
inverter and the efficiency presents a small variation with the value
and d, considering a grid inductance around 250 μH. As shown in of the damping resistance.
Fig. 9c, using a damping resistance equal to 7.4 Ω the rms value of Fig. 11 shows the difference between the efficiency obtained
the leakage current is 434.3 mA, which is higher than the standard with the active damping method and the efficiency obtained with
limit. The rms value of the leakage current presents a reduction of the passive damping method for several values of damping
38.2% if a damping resistance equal to 1.0 Ω is used, as illustrated resistance. The system efficiency grows up to 96.4% when is
in Fig. 9d, where ip,rms = 268.3 mA. This value is in conformity assuming an active damping technique such as the full state
with the main standards [7, 8]. This result proves that an feedback [39]. Moreover, considering the whole range of values for
appropriate passive damping design is very important for the damping resistance, efficiency drops <1.0% when the passive
leakage current reduction in the transformerless grid-connected PV damping is used instead of active damping. This demonstrates that
inverters with an MLCL filter. using passive damping techniques may not result in a significant
The impact of the grid inductance on the leakage current can be decrease of system efficiency.
verified in Figs. 10a and b. The results of Fig. 10a were obtained
using Rd = 2.7 Ω and considering a grid inductance around 250 μH.
6 Conclusion
In this case, the rms value of the leakage current verified was
350.4 mA. The inductance was increased to 340 μH in the results This paper proposed a design methodology for the damping
of Fig. 10b. As a consequence, the rms value of the leakage current resistance in the passive damped MLCL filter applied to leakage

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777 1775
© The Institution of Engineering and Technology 2017
Fig. 9  Experimental results Fig. 10  Experimental results
(a) Grid current (20 A/div) and phase-to-phase voltages (500 V/div) generated by (a) Leakage current ip (1 A/div), grid current igrid (20 A/div), and grid voltage vgrid
inverter using Rd = 1.0 Ω, (b) Grid current step (5 kW → 10 kW) using Rd = 1.0 Ω: (200 V/div) using Rd = 2.7 Ω and Lg = 250 μH, (b) leakage current ip (1 A/div), grid
grid voltage (100 V/div) and grid current (20 A/div) in phase a, (c) Leakage current ip current igrid (20 A/div), and grid voltage vgrid (200 V/div) using Rd = 2.7 Ω and Lg = 
(1 A/div), grid current igrid (50 A/div), and grid voltage vgrid (200 V/div) using Rd =  340 μH, (c) rms value of the leakage current, (d) Damping losses and efficiency at
7.4 Ω, (d) Leakage current ip (1 A/div), grid current igrid (50 A/div), and grid voltage output rated power
vgrid (200 V/div) using Rd = 1.0 Ω
1776 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777
© The Institution of Engineering and Technology 2017
 [12] Yen-Shin, L., Po-Sheng, C., Hsiang-Kuo, L., et al.: ‘Optimal common-mode
voltage reduction PWM technique for inverter control with consideration of
the dead-time effects-part II: applications to IM drives with diode front end’,
IEEE Trans. Ind. Appl., 2004, 40, (6), pp. 1613–1620
[13] Chen, Z., Yu, W., Ni, X., et al.: ‘A new modulation technique to reduce
leakage current without compromising modulation index in PV systems’.
Proc. IEEE ECCE, 2015, pp. 460–465
[14] Freddy, T.K.S., Rahim, N.A., Hew, W.P., et al.: ‘Modulation techniques to
reduce leakage current in three-phase transformerless H7 photovoltaic
inverter’, IEEE Trans. Ind. Electron., 2015, 62, (1), pp. 322–331
[15] Vazquez, G., Kerekes, T., Rocabert, J., et al.: ‘A photovoltaic three-phase
topology to reduce common mode voltage’. Proc. IEEE ISIE, 2010, pp. 2885–
2890
[16] Guo, X., Xu, D., Wu, B.: ‘New control strategy for DCM-232 three-phase PV
inverter with constant common mode voltage and anti-islanding capability’.
Proc. IEEE ECCE, 2014, pp. 5613–5617
Fig. 11  Difference between the system efficiency with active damping and [17] Guo, X., Xu, D., Wu, B.: ‘Three-phase DC-bypass topologies with reduced
leakage current for transformerless PV systems’. Proc. IEEE ECCE, 2015, pp.
with passive damping 43–46
[18] Kerekes, T., Teodorescu, R., Rodríguez, P., et al.: ‘A new high-efficiency
current reduction in grid-connected transformerless three-phase PV single-phase transformerless PV inverter topology’, IEEE Trans. Ind.
inverters. From the simulation and experimental results it was Electron., 2011, 58, (1), pp. 184–191
[19] Kerekes, T., Teodorescu, R., Liserre, M.: ‘Common mode voltage in case of
verified that the damping resistance affects not only the control transformerless PV inverters connected to the grid’, Proc. IEEE ISIE, 2008,
system stability but also directly impacts the leakage current of the pp. 2390–2395
PV system. It was shown that an inappropriate design of this [20] Kerekes, T., Teodorescu, R., Liserre, M., et al.: ‘Evaluation of three-phase
resistance can lead to leakage current above the standard limit. transformerless photovoltaic inverter topologies’, IEEE Trans. Power
Electron., 2009, 24, (9), pp. 2202–2211
Although the reduction of the damping resistance is desired for [21] Dong, D., Fang, L., Boroyevich, D., et al.: ‘Leakage current reduction in a
better leakage current attenuation, it leads the grid current control single-phase bidirectional ac-dc full-bridge inverter’, IEEE Trans. Power
loop to instability. Therefore, it is concluded that this is a trade-off Electron., 2012, 27, (10), pp. 4281–4291
that must be taken into account in the passive damping design to [22] Rendusara, D.A., Enjeti, P.N.: ‘An improved inverter output filter
configuration reduces common and differential modes dv/dt at the motor
ensure leakage current agreement with the standards and control terminals in PWM drive systems’, IEEE Trans. Power Electron., 1998, 13,
system stability. Even when compared with the efficiency obtained (6), pp. 1135–1143
with an active damping technique, it was also verified that the [23] Giacomini, J.C., Michels, L., Schuch, L., et al.: ‘Design of a LCL filter for
passive damping losses represent only a small portion of the leakage current reduction in transformerless PV grid-connected three-level
inverter’. Proc. IEEE APEC, 2015, pp. 239–245
overall power processed by the inverter. Consequently, the overall [24] Dong, D., Luo, F., Zhang, X., et al.: ‘Grid-interface bidirectional converter for
system efficiency is affected rather little. It was also shown that the residential dc distribution systems—part 2: ac and dc interface design with
grid inductance affects the behaviour of both leakage current and passive components minimization’, IEEE Trans. Power Electron., 2013, 28,
control system stability, restricting the solution range for the filter (4), pp. 1667–1679
[25] Wu, W., He, Y., Tang, T., et al.: ‘A new design method for the passive damped
resistance. Therefore, the design methodology contributes to LCL and LLCL filter-based single-phase grid-tied inverter’, IEEE Trans. Ind.
providing an important tool for the selection of the damping Electron., 2013, 60, (10), pp. 4339–4350
resistance in applications that use the passive damped MLCL filter [26] Channegowda, P., John, V.: ‘Filter optimization for grid interactive voltage
for leakage current reduction. The approach also can be extended source inverters’, IEEE Trans. Ind. Electron., 2010, 57, (12), pp. 4106–4114
[27] Ye, Z., Xu, Y., Wu, X., et al.: ‘A simplified PWM strategy for a neutral-point-
to other transformerless PV inverter topologies and passive clamped (NPC) three-level converter with unbalanced dc links’, IEEE Trans.
damping schemes that employ the MLCL filter. Power Electron., 2016, 31, (4), pp. 3227–3238
[28] Choudhury, A., Pillay, P., Williamson, S.S.: ‘Performance comparison study
of space-vector and modified-carrier-based PWM techniques for a three-level
7 References neutral-point-clamped traction inverter drive’, IEEE J. Emerg. Sel. Top.
Power Electron., 2016, 4, (3), pp. 1064–1076
[1] Barater, D., Lorenzani, C., Franceschini, G., et al.: ‘Recent advances in [29] Beres, R.N., Xiongfei, W., Blaabjerg, F., et al.: ‘Optimal design of high-order
single-phase transformerless photovoltaic inverters’, IET Renew. Power passive-damped filters for grid-connected applications’, IEEE Trans. Power
Gener., 2016, 10, (2), pp. 260–273 Electron., 2016, 31, (3), pp. 2083–2098
[2] Cho, Y.W., Cha, W.J., Kwon, J.M., et al.: ‘Improved single-phase [30] Yang, J., Lee, F.C.: ‘LCL filter design and inductor current ripple analysis for
transformerless inverter with high power density and high efficiency for grid- a three-level NPC grid interface converter’, IEEE Trans. Power Electron.,
connected photovoltaic systems’, IET Renew. Power Gener., 2016, 10, (2), pp. 2015, 30, (9), pp. 4659–4668
166–174 [31] Liserre, M., Blaabjerg, F., Hansen, S.: ‘Design and control of an LCL-filter-
[3] Debnath, D., Chatterjee, K.: ‘Maximising power yield in a transformerless based three-phase active rectifier’, IEEE Trans. Ind. Appl., 2005, 41, (5), pp.
single-phase grid connected inverter servicing two separate photovoltaic 1281–1291
panels’, IET Renew. Power Gener., 2016, 10, (8), pp. 1087–1095 [32] Kangle, R., Xing, Z., Fusheng, W., et al.: ‘Optimized design of filter for
[4] Lopez, O., Freijedo, F.D., Yepes, A.G., et al.: ‘Eliminating ground current in a transformerless three-level photovoltaic grid-connected inverter’. Proc. IEEE
transformerless photovoltaic application’, IEEE Trans. Energy Conv., 2010, TENCON, 2013, pp. 1–5
25, (1), pp. 140–147 [33] Michels, L., Camargo, R.F., Botteron, F., et al.: ‘Generalised design
[5] Gubía, E., Sanchis, P., Ursúa, A., et al.: ‘Ground currents in single-phase methodology of second-order filters for voltage-source inverters with space-
transformerless photovoltaic systems’, Prog. Photovolt., Res. Appl., 2007, 15, vector modulation’, IEE Proc. Electr. Power Appl., 2006, 153, (2), pp. 219–
(7), pp. 629–650 226
[6] Gonzalez, R., Lopez, J., Sanchis, P., et al.: ‘Transformerless inverter for [34] Myrzik, J.M.A., Calais, M.: ‘String and module integrated inverters for
single-phase photovoltaic systems’, IEEE Trans. Power Electron., 2007, 22, single-phase grid connected photovoltaic systems - a review’. Proc. IEEE
(2), pp. 693–697 PowerTech, 2003, p. 8
[7] IEC 62109–2: ‘Safety for power converters for use in photovoltaic power [35] Teodorescu, R., Liserre, M., Rodríguez, P.: ‘Grid converters for photovoltaic
systems – part 2’, 2011 and wind power systems’ (John Wiley-IEEE, 2011)
[8] DIN VDE 0126–1–1: ‘Automatic disconnection device between a generator [36] Espi, J.M., Castello, J., García-Gil, R., et al.: ‘An adaptive robust predictive
and the public low-voltage grid’, 2005 current control for three-phase grid-connected inverters’, IEEE Trans. Ind.
[9] Cavalcanti, M.C., Oliveira, K.C., Farias, A.M., et al.: ‘Modulation techniques Electron., 2011, 58, (8), pp. 3537–3546
to eliminate leakage currents in transformerless three-phase photovoltaic [37] Ogata, K.: ‘Discrete time control systems’ (Prentice-Hall, 1995)
systems’, IEEE Trans. Ind. Electron., 2010, 57, (4), pp. 1360–1368 [38] Dannehl, J., Wessels, C., Fuchs, F.W.: ‘Limitations of voltage-oriented PI
[10] June-Seok, L., Kyo-Beum, L.: ‘New modulation techniques for a leakage current control of grid-connected PWM rectifiers with LCL filters’, IEEE
current reduction and a neutral-point voltage balance in transformerless Trans. Ind. Electron., 2009, 56, (2), pp. 380–388
photovoltaic systems using a three-level inverter’, IEEE Trans. Power [39] Maccari, L.A., Santini, C.L.A., Pinheiro, H., et al.: ‘Robust optimal current
Electron., 2014, 29, (4), pp. 1720–1732 control for grid-connected three-phase pulse-width modulated converters’,
[11] Cavalcanti, M.C., Farias, A.M., Oliveira, K.C., et al.: ‘Eliminating leakage IET Power Electron., 2015, 8, (8), pp. 1490–1499
currents in neutral point clamped inverters for photovoltaic systems’, IEEE
Trans. Ind. Electron., 2012, 59, (1), pp. 435–443

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1769-1777 1777
© The Institution of Engineering and Technology 2017