LCL Filters for a Grid Emulator Application

Comparative Study of Active Damping Techniques

Franois Bronchart, Yves Mollet and Johan Gyselinck

Bio Electro And Mechanical Systems department
cole Polytechnique de Bruxelles
Universit Libre de Bruxelles

AbstractThis paper focuses on the implementation of instead of the output current.

LCL filter active damping in the control of an inverter
for a three-phase grid emulator application. The studies II. G RID E MULATOR
about LCL damping in the literature focus on the case
of an inverter injecting current into the grid while a grid vi vi
emulator application is rather about establishing correct vg PR controller Inverter LCL + load vg
voltage waveforms and emulating the grid for any kind of -
load connected to it, which brings a different perspective.
After having analyzed the stability of the LCL filter
taking into account possible load variations, three damping Fig. 1. Closed-loop voltage control structure of the grid emulator.
methods are compared and assessed with respect to those
concerns.
The general structure of the emulator and its control
is schematized in figure 1. A closed-loop voltage control
I. I NTRODUCTION
with proportional and resonant controller [12] [13] in
Most inverter-based applications need LC or LCL abc frame is adopted as this technique allows a simpler
low-pass filters to remove the switching harmonics and less computationally expensive implementation than
from the output current and voltage. A natural be- traditional PI control with several dq0 abc transfor-
haviour and inherent problem of those LC circuits is mations and PLLs.
the resonance phenomenon leading to uncontrolled high-
The controller equation is
amplitude voltage and current oscillations. It is therefore
a common problem that has to be addressed. Active X 2Kh s
HPR (s) = Kp + 2 + (h )2
(1)
damping methods in the case of an inverter injecting a h=1,3,5,7
s 0
current to the grid (with quasi-constant electrical char-
acteristics) are studied in [1] - [4]. A grid emulator on where Kp is the proportional gain and Kh the gain of
the contrary operates with various loads or even with no the resonant terms associated with the fundamental and
load, and must therefore display robustness with respect the harmonics 3, 5 and 7.
to load variation (i.e. behaviour should remain constant From the three-phase voltage reference vg and the
with different kind of loads and working regimes). While actual output voltages vg , the controller outputs the
some papers already investigated about the control of grid commands for the inverter vi with the knowledge of the
emulators and the generation of fault transients [6] [8] [9] DC bus voltage. The dynamics of the PWM inverter are
[10], they do not consider the damping of the filter or use very fast compared to the filter dynamics and therefore
a passive damping [11], and do not consider the stability neglected in the next developments. PWM dead times
issues that could arise from highly varying loads. and voltage limits relative to the DC bus are also ne-
glected.
This paper compares three active damping methods
of the three-phase LCL filter used in a grid emulator ap- One phase of the equivalent circuit of the filter is
plication. It focuses on the impact of highly varying RL shown in figure 2 for defining the notations. The voltage
loads that affect the stability of the filter and considers at the output of the inverter is vi and the RL load has
the problem from the point of view of the output voltage an impedance Zg = sLg + Rg . An LCL filter is chosen
 L1 L2 ig 4
i x 10
1

iC

Imaginary Axis (seconds1)

vi uC C Zg vg
0.5
Rg

Fig. 2. Perfect LCL filter. 0

rather than a LC filter because the second inductance L2 0.5

helps to reduce the variation of the resonance frequency
due to varying load inductance Lg .
1
III. F ILTER STABILITY 5000 4000 3000 2000 1000 0 1000 2000
1
Real Axis (seconds )
With the usual assumption that the components have a
negligible Equivalent Series Resistance, the filter transfer Fig. 3. Root loci of the filter for decreasing values of Rg from 1 k
function between the output and input voltages is R
to 0.1 , keeping Lgg = 5000 s1 constant.

vg sLg + Rg Inverter-side inductor L1 0.8 mH

= 3 (2) Load-side inductor L2 2 mH
vi s L1 CL2g + s2 CL1 Rg + sL12g + Rg Filter capacitor C 50 F
Inverter switching frequency fswitch 10 kHz
with L2g = L2 + Lg and L12g = L1 + L2 + Lg . Load resistance Rg 10
Load impedance Lg 0 mH
It consists of a pair of dominant complex poles that Proportional gain Kp 1

depend mostly on the L1 C resonant circuit and a real TABLE I. PARAMETERS OF THE SYSTEM
pole and a real zero that depend on the RL load Zg =
sLg + Rg .
Figure 3 shows several root loci of the filter where to a situation close to the no-load case, where the two
the gain parameter of the loci is the proportional gain loci are superposed to the imaginary axis. For the next
Kp 1 . If a constant R g
Lg ratio (position of the zero) is developments, the parameters of the system and the load
considered and if the value of Rg is decreased (and are indicated in table I.
therefore the value of Lg as well), the root locus is
modified as shown in figure 3. An unfavourable load
IV. ACTIVE DAMPING METHODS
from a stability point of view is therefore such that the
locus branches are strongly pushed into the right half- Active damping methods can be separated in two
plane. This corresponds to the case of a small resistive categories [2]: feedback or multiloop-based methods and
only (Lg = 0) impedance for which the R g
Lg ratio tends
filter-based methods.
to infinity. The zero is then pushed to and loses its
influence on the loci. This kind of load will be used in the A. Feedback-based methods
following simulations to assess the effectiveness of the
Feedback-based methods exploit the knowledge of
damping methods. However, for a large load impedance,
state variables of the system besides the output, to alter
the complex poles go on the imaginary axis and the
its behaviour and modify its dynamics.
real pole and real zero are cancelling each other, leading
1) Full state feedback: Also called pole placement
1
The resonant part of the controller merely adds pairs of poles and technique, this method allows to move the system poles
zeros close together for each value of h considered in equation (1),
at any position (within the physical limits of the system)
which are much lower than the resonance frequency of the filter.
Therefore, it does not affect significantly the loci and has been provided that a full set of state variables are fed back.
neglected to avoid cluttering the figure. With all system poles being movable, this technique
 15000 10
Undamped
Capacitor current feedback 0
Imaginary Axis (seconds1)

Passive damping
10
10000

Magnitude (dB)
20

30

5000
40

Undamped
50
(k = RC ) Capacitor current feedback
Passive damping
0 60
2 3 4
10000 8000 6000 4000 2000 0 2000 10 10 10
1 Frequency (Hz)
Real Axis (seconds )

Fig. 4. Passive damping with real resistor and active damping with capacitor current feedback in the case of table I. Left: Root loci for
increasing values of RC = k = [1; 2.4; 4; 6; 11] (direction of the black arrow). The negative imaginary part of the locus is symmetrical to
the positive imaginary part. Right: Closed-loop bode for k = RC = 4 .

allows even more optimization than only damping the

1 Ic 1 1 Ig
dominant poles, such as reducing the actuating energy 1
v_i sL1 sC u_C sL2
s Lg
v_g
v_g
1

[7]. However, the practical implementation requires many

Rc Rg
voltage/current sensors or alternatively an observer that Real resistor
estimates the state variables on the basis of a model of
the system, at the expense of a larger control complexity. Model of the filter

2) Capacitor current feedback or Virtual resistor: Fig. 5. Block-diagram of the system with classic passive damping.
Very often, the feedback of one state variable only
is enough to stabilize the closed-loop system. Among
the state variables that can be fed back, the capacitor 1 v_i* v_i
v_i
1 Ic 1
u_C
1 Ig
s Lg
v_g
1
v_i*
sL1 sC sL2 v_g
Inverter
current is a good choice [6]. Another advantage is its
Rg
equivalent physical meaning as this method amounts to
approximately the same effect as passive damping with Model of the filter
k
a resistor in series with the capacitor, hence the name of
virtual resistor method. Capacitor current feedback gain
In the simplified case of no load, the transfer functions
Fig. 6. Block-diagram of the system with capacitor current feedback.
of the classical passive damping with resistance RC and
the capacitor current feedback with feedback gain k are
respectively
and real resistor add the same amount of damping to
vg sCRc + 1 the system poles but capacitor current feedback allows
= 2 (3) to keep the roll-off rate of -40dB/decade from the un-
vi s L1 C + sCRc + 1
damped case instead of -20dB/decade in the case of the
vg 1 real resistor, at the expense of a less damped resonance
= 2 (4) at high feedback gains. The corresponding diagrams are
vi s L1 C + sCk + 1
presented in figures 5 and 6 respectively. As for the pole
The only difference lies in the absence of the zero in placement technique, capacitor currents can be directly
equation (4). As it can be seen in figure 4, both virtual measured or estimated through a model of the system.
B. Filter-based methods
This family of methods consists of an additional block
in the control loop, generally just after the controller. 20
They share the advantage of not needing any additional
measurement but require a more precise knowledge of 0

Magnitude (dB)
the system parameters in order to be effectively tuned.
20

vg Controller GAD System vg 40

Open loop undamped
-
Closed loop undamped
60 Open loop + lead
Closed loop + lead

90

Fig. 7. Filter-based methods: an additional block GAD is added, 0

generally after the controller, to reduce the resonance.
90

Phase (deg)
180

1) Lead compensator: The lead compensator [3] is 270

the simplest stabilizing block that can be added to the 360
control: 450
1 2 3 4 5
Ts + 1 10 10 10 10 10
Glead (s) = ( < 1) (5) Frequency (Hz)
T s + 1

Its real pole and real zero define a frequency zone Fig. 8. Increase of the phase margin due to the lead compensator
with T = 4.68 104 s, = 0.01, with the parameters of table I.
in which phase lead is added. The maximization of the
system phase margin is an iterative process that must
be performed to centre the phase "bump" of the lead
compensator at the unit gain frequency of the LCL
magnitude curve (see figure 8). The amount of phase
that can be added is limited by the limited spacing of the 10
zero and the pole in the implementation (i.e. cannot be
0
infinitely small) and any modification of the system or the
Magnitude (dB)

control may reduce the stability margin or compromise 10

the stability. In addition, the lead compensator tends to 20

amplify the higher frequencies such as the switching 30 Open loop undamped
harmonics or the noise. Closed loop undamped
40 Open loop + notch
Closed loop + notch
2) Notch filter: From the point of view of its fre- 50
quency response, the Notch filter is basically a band-stop 0
filter tuned to the resonance frequency of the system and 90
placed in the forward path of the control loop to damp
Phase (deg)

the resonance peak: 180

270

s2 + 2z 0 s + 02 360
Gnotch (s) = (z < p ) (6)
s2 + 2p 0 s + 02 450
2 3 4
10 10 10
Frequency (Hz)
The notch filter effects a 45 phase lag and lead below
and above notch frequency respectively (see figure 9).
In order to keep enough phase lead at the unity gain Fig. 9. Effect of the notch filter tuned at the resonance frequency
of the filter with p = 0.4 and z = 0.07, with the parameters of
frequency, the notch frequency should be chosen equal table I.
or smaller than the system resonance frequency but never
 150
higher. The same result arises from the point of view of
100
the root locus: the zeros of the notch have to attract the
locus in the left half-plane and must therefore be placed 50

VoltagebzV7
below the system poles. Also, the distance between 0
system poles and notch zeros should be large enough
50
to avoid pole-zero flipping in case of load change or any
other variation of the system parameters that affects the 100

position of its dominant poles. 150

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Timebzs7
The notch poles aim at making the notch transfer za7bWithbvirtualbresistor,bkb=b6bOhm
function proper. They are usually chosen far on the neg-
200
ative real axis and heavily damped unless their response
impedes on the closed-loop bandwidth.
100

VoltagebzV7
V. C OMPARISON AND RESULTS 0

In order to check their behaviour in presence of 100

switching harmonics, the three active damping methods
have been implemented in simulation by modelling all 200
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
components including the control blocks and the inverter. Timebzs7
Simulation results are presented in figure 10 for the zb7bWithbnotchbfilterbtunedbatbresonancebfrequencybzpb=b0.4bandbzb=b0.077
parameters of table I. 600

The results show that the lead compensator does not 400
have a large enough stability margin and even if its 200
VoltagebzV7

poles are designed to be stable in a continuous-time

study, the simulation at a switching frequency of 10 kHz
cannot achieve stability (see figure 10)2 . The other active
damping methods are able to successfully stabilize the
system at 10 kHz.
The loci of the closed-loop system poles for different
load impedances and for each active damping method
are presented in figure 11. This figure helps to visualize
the load influence on the system stability for the three
considered damping methods. The poles of the system
with the notch filter remain poorly damped for every
considered load and even become unstable in some cases.
This damping method would then have to be re-adjusted
with each system modification. The system behaviour
with the lead compensator becomes surprisingly stable
with the help of the zero from the load inductance or
when Rg is very high (close to no load case), however
it remains unstable for the low impedance resistive load
considered as most unfavourable earlier.
Capacitor current feedback on the other side performs
very well and is robust as it maintains a damped reso-
nance for each load variation considered.
2
It is possible to reach stability but only beyond a switching
frequency of 20 kHz or with lower Kp gain.
0
200
400
600
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Timebzs7
zc7bWithbleadbcompensatorbzTb=b4.68bb104bs,bb=b0.017
Fig. 10. Simulation of the filter voltage outputs (vg ) resulting from
the three active damping methods, with the parameters of table I,
using the model of a 3-leg IGBT converter switching at 10kHz.
The simulation consist in establishing three phase voltages at 100 V
(starting from 0 V at t = 0 s) and imposing a 50 V step in the voltage
amplitude at t = 0.02 s.
VI. C ONCLUSION
This paper showed the influence of load variations
on the stability of an LCL filter in the context of a grid
emulator. The capacitor current feedback technique is the
preferred active damping method for its robustness since
it provides damped poles for all load cases considered
and its simplicity to implement during the development
of the grid emulator. On the contrary, filter-based meth-
ods may be used at lower costs for applications in which
the load remains constant. It can also be noted that the
 4 4
x 10 x 10
1 1.5

0.8

1
Rg
0.6
Lg
0.4

Imaginary Axis (s )
Imaginary Axis (s-1)

0.5

-1
0.2

0 0

0.2

0.5
0.4

0.6
1
Capacitor current feedback Capacitor current feedback
0.8
Lead Lead
Notch Notch
1 1.5
20000 15000 10000 5000 0 5000 20000 15000 10000 5000 0 5000
Real Axis (s-1) Real Axis (s-1)

Fig. 11. Locus of the closed-loop poles for the three considered damping methods, with parameters of table I and varying Rg and Lg . Left:
Rg = 10 and Lg = 0.01 mH 1 H. Right: Lg = 0 and Rg = 0.1 1 k.

robustness of the capacitor current feedback method may
be helpful in case of a modification of the filter (e.g. if
components are replaced3 ). Further works will include
tests with capacitive and non-linear loads.
R EFERENCES
[1] Wenqiang Zhao; Guozhu Chen, "Comparison of active and
passive damping methods for application in high power active
power filter with LCL-filter" Sustainable Power Generation and
Supply, 2009. SUPERGEN 09. International Conference on,
pp.1,6, 6-7 April 2009.
[2] Teodorescu, R.; Liserre, M.; Rodriguez, P., Grid Converters
for Photovoltaic and Wind Power Systems Ch.11 : Grid Filter
Design, 2011.
[3] Gullvik, W.; Norum, L.; Nilsen, R., "Active damping of reso-
nance oscillations in LCL-filters based on virtual flux and virtual
resistor" Power Electronics and Applications, 2007 European
Conference on, pp.1,10, 2-5 Sept. 2007.
[4] Wessels, C.; Dannehl, J.; Fuchs, F.W., "Active damping of LCL-
filter resonance based on virtual resistor for PWM rectifiers
stability analysis with different filter parameters" Power
Electronics Specialists Conference, 2008.
[5] Shiying Tang ; Li Peng ; and Yong Kang, "Active damping
method using grid-side current feedback for active power filters
with LCL filters" Journal of Power Electronics, Vol. 11, No. 3,
May 2011.
3
Typical tolerance on the component value for commercial power
inductances and MKP capacitors (the most used kind of capacitors
used for this application) is 5%-10%.
[6] Fei Li; Xiongfei Wang; Zhe Chen; Xing Zhang, "A laboratory
grid simulator based on three-phase four-leg inverter: Design
and implementation" Electrical Machines and Systems (ICEMS),
2011 International Conference on, pp.1,5, 20-23 Aug. 2011.
[7] Lohde, R.; Fuchs, F.W., "Laboratory type PWM grid emulator
for generating disturbed voltages for testing grid connected
devices" Power Electronics and Applications, 2009. EPE 09.
13th European Conference on, pp.1,9, 8-10 Sept. 2009.
[8] Arun B Karuppaswamy; John Vinod, "A hardware grid simu-
lator to test grid-connected inverter systems" National Power
Electronics Conference, June 2010.
[9] Namwon Kim; Sang-Yong Kim; Hyo-Guen Lee; Chulsang
Hwang; Gyeong-Hun Kim; Hyo-Ryong Seo; Minwon Park; In-
Keun Yu, "Design of a grid-simulator for a transient analysis
of grid-connected renewable energy system" Electrical Ma-
chines and Systems (ICEMS), 2010 International Conference on,
pp.633,637, 10-13 Oct. 2010.
[10] Richter, S.A.; Von Bloh, J.; Dick, C.P.; Hirschmann, D.; De
Doncker, R.W., "Control of a medium-voltage test generator"
Power Electronics Specialists Conference, 2008. PESC 2008.
IEEE, pp.3787,3793, 15-19 June 2008.
[11] Cespedes, M.; Tao Qi; Jian Sun, "Development of a grid simu-
lator" Control and Modeling for Power Electronics (COMPEL),
2012 IEEE 13th Workshop on, pp.1,8, 10-13 June 2012.
[12] Remus Teodorescu; Frede Blaabjerg, "Proportional-resonant
controllers. a new breed of controllers suitable for grid-
connected voltage-source converters" Proceedings of The 9th
International Conference on Optimization of Electrical and
Electronic Equipments, Optim 2004, 2004.
[13] Blaabjerg, F.; Teodorescu, R.; Liserre, M.; Timbus, A.V.,
"Overview of control and grid synchronization for distributed
power generation systems" Industrial Electronics, IEEE Trans-
actions on, vol.53, no.5, pp.1398,1409, Oct. 2006.