IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO.

5, OCTOBER 2006 1461

Wireless-Control Strategy for Parallel Operation of

Distributed-Generation Inverters
Josep M. Guerrero, Member, IEEE, José Matas, Luis García de Vicuña, Miguel Castilla, and
Jaume Miret, Member, IEEE

Abstract—In this paper, a method for the parallel operation current loop ensures that the inverter acts as a current source
of inverters in an ac-distributed system is proposed. This paper [7], [8]. Currently, when the grid is not present, the inverters
explores the control of active and reactive power flow through are normally disconnected from the ac line in order to avoid
the analysis of the output impedance of the inverters and its
impact on the power sharing. As a result, adaptive virtual output islanding operation. In the coming years, inverters should be
impedance is proposed in order to achieve a proper reactive power able to operate in island mode due the high penetration of
sharing, regardless of the line-impedance unbalances. A soft-start DG [9]. In addition, in certain zones where a stiff grid is not
operation is also included, avoiding the initial current peak, which accessible, e.g., some physical islands, rural or remote areas,
results in a seamless hot-swap operation. Active power sharing is islanding operation mode is necessary. In this situation, the
achieved by adjusting the frequency in load transient situations
only, owing to which the proposed method obtains a constant output-voltage reference should be provided internally by the
steady-state frequency and amplitude. As opposed to the conven- DG units, which operate independently without mutual inter-
tional droop method, the transient response can be modified by communication due to the long distance between them [10].
acting on the main control parameters. Linear and nonlinear loads Several control techniques based on the droop method have
can be properly shared due to the addition of a current harmonic been proposed in order to avoid using communication between
loop in the control strategy. Experimental results are presented
from a two-6-kVA parallel-connected inverter system, showing the DG units [11]–[16]. These control techniques consist in making
feasibility of the proposed approach. tight adjustments over the output-voltage frequency and am-
plitude of the inverter in order to compensate for the active
Index Terms—Distributed generation (DG), droop control
method, microgrids, nonlinear loads. and reactive power unbalances [17]. Nevertheless, the standard
approach only works well when linear loads are shared, since
the amount of distorted power demanded by the nonlinear
I. I NTRODUCTION loads is not taken into account. In this sense, an additional
control loop was proposed in order to share nonlinear loads by
D ISTRIBUTED generation (DG) is emerging as a new
paradigm to produce onsite highly reliable and good-
quality electrical power [1]. DG becomes a viable alterna-
adjusting the output-voltage bandwidth according to the amount
of delivered harmonic power [18]. In previous work, most har-
tive when renewable or nonconventional energy resources are monic current values were used to produce a proportional droop
available, such as photovoltaic arrays, fuel cells, cogeneration in the corresponding harmonic voltage term [19]. However,
plants, combined heat and power microturbines, or small wind all these control approaches have an inherent tradeoff between
turbines [2]–[4]. These resources can be connected to local voltage regulation and power sharing. Furthermore, the conven-
low-voltage electric power networks, also called mini- or mi- tional droop method exhibits slow dynamic response, since it
crogrids, through power conditioning ac units, i.e., inverters requires low-pass filters with reduced bandwidth to calculate
or ac–ac converters [5], [6], which can operate either in grid- the average values of the active and reactive powers [20]. In
connected mode or in island mode. [21], a wireless controller was proposed in order to enhance
Grid-connected operation consists in delivering power to the the dynamic performance of the paralleled inverters by adding
local loads and to the utility grid. In such a case, the output- integral-derivative power terms to the droop-control method.
voltage reference is often taken from the grid voltage sensing, Nevertheless, when an inverter is connected suddenly to the
and is using a phase-locked-loop (PLL) circuit, while an inner common ac bus, a current peak appears due to the initial phase
error.
Manuscript received December 29, 2004; revised May 9, 2005. Abstract Another drawback of the standard droop method is that, the
published on the Internet July 14, 2006. This paper was presented in part at power sharing is degraded if the sum of the output impedance
the IEEE International Symposium on Industrial Electronics (IEEE-ISIE’05), and the line impedance is unbalanced. To solve this, interface
Dubrovnik, Croatia, June 20–23, 2005. This work was supported by the Spanish
Ministry of Science and Technology under Grant CICYT DPI 2003-06508- inductors can be included between the inverter and the load
C02-01. bus, but they are heavy and bulky [22]. As an alternative,
J. M. Guerrero is with the Department d’Enginyeria de Sistemes, Automàtica novel control loops that fix the output impedance of the units
i Informàtica Industrial (ESAII), Escola Universitària d’Enginyeria Tècnica In-
dustrial de Barcelona (EUETIB), Universitat Politècnica de Catalunya (UPC), by emulating lossless resistors or reactors have been proposed
08036 Barcelona, Spain (e-mail: josep.m.guerrero@upc.edu). [23], [24]. However, although the output impedance of the
J. Matas, L. García de Vicuña, M. Castilla, and J. Miret are with the Departa- inverter can be well established, the line impedance is unknown,
ment d’Enginyeria de Electrònica, Universitat Politècnica de Catalunya, 08036
Barcelona, Spain. which can result in an unbalance of reactive power flow [18].
Digital Object Identifier 10.1109/TIE.2006.882015 This problem can be overcome by injecting high-frequency

0278-0046/$20.00 © 2006 IEEE

1462 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006

Fig. 2. Equivalent circuit of j inverters connected to a common ac bus.

utility grid through a single point of common coupling. When

the utility grid is not present, the DG units should be able to
share the total power demanded by the local loads, adjusting its
output-voltage references as a function of the dispatched power.
Fig. 2 shows the equivalent circuit of the DG units as
inverters connected to a common ac bus through decoupling
output impedances. The active and reactive power injected to
the bus by every unit can be expressed as follows [27]:


EV V2 EV
P = cos φ − cos θ + sin φ sin θ (1)
Z Z Z


EV V2 EV
Q= cos φ − sin θ − sin φ cos θ (2)
Z Z Z

where E and V are the amplitudes of the inverter output voltage

and the common bus voltage, φ is the power angle, and Z and
θ are the magnitude and the phase of the output impedance,
respectively. Note that both P and Q depend simultaneously on
Fig. 1. Diagram of a microgrid. the output-voltage parameters ω (i.e., ω = dφ/dt) and E.
Therefore, the droop-control method can be expressed in a
signals through the power lines [25] or by adding external data general form as
communication signals, as in [26]. Unfortunately, such com-
munication among DG units increases complexity and reduces ω = ω ∗ − m(P sin θ − Q cos θ) (3)
reliability, since the power balance and the system stability rely E = E ∗ − n(P cos θ + Q sin θ). (4)
on these signals.
In this paper, we propose a novel control scheme that is able The output impedance angle θ determines to a large extent
to further improve the steady state and the transient response of the droop-control law, as shown in Table I. Traditionally, the
the parallel-connected inverters without using any communica- inverter output impedance is considered to be inductive due to
tion signals. The controller uses an adaptive output impedance, the high inductive component of the line impedance and the
which allows a good reactive power sharing with low sensitivity large inductor filter. However, this is not always true, since
to the line-impedance unbalances, while keeping constant the the closed-loop output impedance also depends on the control
output-voltage amplitude and the frequency in steady state. A strategy [28].
soft-start operation of the output impedance is also proposed in By using fast-droop algorithms, different output impedances
order to achieve a seamless connection between the inverter and can be implemented, such as, (a) purely resistive, (b) purely in-
the common bus. Finally, the most significant output-current ductive, (c) inductive and resistive connected in series, or (d) in
harmonics are treated separately using a bank of bandpass parallel, as shown in Fig. 3. Resistive output impedance can be a
filters in order to share these harmonics properly without exces- good solution to share linear and nonlinear loads in applications
sively increasing the output-voltage total harmonic distortion such as uninterruptible-power-supply systems [23]. However,
(THD). Experimental results from a two-6-kVA-inverter system when the distance between the inverters is considerable, an
are provided, showing the outstanding features of the proposed inductive impedance component appears which impoverishes
controller. active and reactive power sharing. Inductive output impedance
seems to be the most natural output impedance [24]. However,
it degrades the output-voltage THD too much when supply-
II. O UTPUT -I MPEDANCE I MPACT O VER P OWER S HARING
ing nonlinear loads due to the large impedance value seen
Fig. 1 shows a general scheme of a microgrid which consists by the current harmonics. In order to solve these problems,
of a combination of multiple microgenerator DG units, distrib- two combined resistive–inductive output impedances can be
uted loads, and electric power interfaces that transfer energy considered: series and parallel. The first one [29], although it
to the local ac bus. The microgrid can be connected to the can be controlled by using (3) and (4), is still unsatisfactory,
GUERRERO et al.: WIRELESS-CONTROL STRATEGY FOR PARALLEL OPERATION OF DG INVERTERS 1463

TABLE I
DROOP-METHOD STRATEGY DEPENDENCE WITH OUTPUT IMPEDANCE

voltage reference. Thus, the proposed output-voltage reference

can be expressed as


11
∗
νref = νref − sLD io1 − Rh ioh (5)
h=3
odd

where LD is the virtual output impedance, and Rh is the

resistive coefficient of every harmonic term ioh . Using this loop,
the output impedance presented to the fundamental and the
harmonic components can be fixed independently.
Another practical issue is the desirable hot-swap capability,
which consists in a seamless operation of the DG inverter when
it is connected suddenly to the common ac bus. As explained in
[21], the output current peak in such a situation is expressed as
Fig. 3. Different output impedance situations.
E
Ip ≈ · ∆φ (6)
ωLD
since the inductive component increases the output-voltage
THD. The second one is able to share linear and nonlinear where ∆φ is the initial phase error.
loads, since the output impedance combines inductive behavior In order to reduce this initial current peak, a soft-star opera-
around the output-voltage frequency and resistive behavior for tion of the output impedance is also proposed which achieves
high-order current harmonics [30]. However, harmonic power a seamless connection of the inverter with the common bus
sharing cannot be independently controlled using a constant (hot-swap operation). The soft-start operation consists in this
nominal value for the virtual inductance. connection using high output impedance L∗Do , which is then
slowly reduced toward the nominal value L∗Df
 
III. P ROPOSED W IRELESS P OWER -S HARING C ONTROLLER L∗D = L∗Df + L∗Do − L∗Df · e−t/TST (7)
In this section, we propose a wireless control for parallel where L∗Do and L∗Df are the initial and the final value of the
DG inverters that provides a proper output impedance, good output impedance, respectively, and TST is the constant time of
active and reactive power-sharing capability, constant steady- the soft-start operation. This way, the initial current peak can be
state output-voltage frequency and amplitude, and fast transient avoided regardless of the initial phase error.
response.

B. Active and Reactive Power-Sharing Loop

A. Virtual Output Impedance and Harmonic Current Sharing
In order to properly share both active and reactive pow-
A fast control loop known as virtual-output-impedance loop ers, we propose a control strategy which avoids the use of
can be used to fix the output impedance of the inverter. Induc- any control wire interconnections between the modules. As
tive output impedance around the output-voltage frequency can we stated before, the conventional droop method presents an
be implemented by drooping the output-voltage reference pro- inherent tradeoff between the P/Q sharing accuracy and the
portionally to the time derivative of the fundamental component frequency/amplitude output-voltage regulation. From Table I, it
of the output current io1 . Resistive output impedance for high- can be observed that when the output impedance is inductive,
order current harmonics is obtained by subtracting a voltage, the reactive power can be controlled adequately by the output-
which is proportional to the current harmonics from the output- voltage amplitude, as in a conventional droop method, or by
1464 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006

the output impedance, as in variable active–passive reactances where ∧ denotes perturbed values, capital letters mean equi-
[31]. Therefore, taking into account that Q is reduced when X librium point values, X is the output impedance at the fun-
increases, we propose the following adaptive output impedance damental frequency, and ωc is the cutoff angular frequency of
in order to regulate the Q balance [28] the low-pass filters, which is fixed over one decade below the
line frequency. For the sake of simplicity, the high-frequency
LD = L∗D + kL Q (8) impedance values are not considered in this analysis since they
have little effect over the system dynamics.
where kL is an adjustment constant of the inductive output- Subsequently, by linearizing (8) and (10), and using (11) and
impedance gain, and L∗D is the reference output inductance. The (12), we obtain
proposed output impedance compensates for reactive-power
differences between the modules due to output-voltage mis- ωc 1
φ̂ = −(mp + md s) −1
matches, component tolerances, or line-impedance unbalances, (s + ωc )(s + τ ) X
without deviations of the output-voltage reference amplitude.
× [EV cos Φ · φ̂ − P · x̂] (13)
Notice that even though the active power is also affected by the
output impedance, it can be controlled by adjusting the output- ωc EV sin Φ
x̂ = −kL · φ̂. (14)
voltage frequency. X(s + ωc ) + kL ωc Q
In order to avoid the steady-state frequency deviation of the
conventional droop method, the following control scheme is Finally, substituting (14) into (13), we can find the small-
proposed: signal dynamics of the closed-loop system

dP s3 φ̂ + As2 φ̂ + Bsφ̂ + C φ̂ = 0 (15)

ω = ω ∗ − mp P − md (9)
dt
where
where mp and md are, respectively, the proportional and deriv-
ative coefficients of the active power with no dc component, P, A=
1 −1
[τ +ωc +md EV cos Φ + Xωc + kL ωc Q]
which can be obtained by using a high-pass filter such as X

1
s B = 2 P md kL ωc2 EV sin Φ+Xmp ωc EV cos Φ
P = P (10) X
s + τ −1 
and τ is the time constant of the transient droop action. The + (Xτ −1 +Xωc +md EV cos Φ)(Xωc +kL ωc Q)
derivative term of P is added to the conventional scheme in 
order to improve the transient response of the system, as will 1
C= (Xωc τ −1 +mp ωc EV cos Φ)(Xωc +kL ωc Q)
be shown in the following Section. Note that the P − ω func- X2
tion has no frequency deviation in steady state. The transient 
2
droop function ensures a constant frequency regulation under + P mp kL ωc EV sin Φ .
steady-state conditions and, at the same time, achieves active
power balance by adjusting the inverter frequency during load
By using (15), the stability of the closed-loop system can
transients. The proposed control scheme allows us to modify
be evaluated, and a desired transient response can be selected
the transient response by acting on the control coefficients and,
following a linear third-order dynamics. After studying the
at the same time, keeping constant the steady-state frequency
eigenvalues of the system (λ1 , λ2 , and λ3 ) through a series
and amplitude of the output-voltage reference [30].
of root locus diagrams, as in [20] and [21], and using the
values listed in Table II, we can conclude that both the output
IV. S MALL -S IGNAL M ODELING AND impedance X and the coefficient kL have little effect on the
C ONTROL D ESIGN R ULES location of the roots in comparison with mp and md coeffi-
cients. Fig. 4(a) depicts the family of root locus diagrams for
A small-signal analysis is proposed in order to investigate different values of these coefficients. It can be observed that the
the stability and the transient response of the system. Thus, the poles can be adjusted to fix the desired dynamics. On the one
closed-loop system dynamics is derived, taking into account the hand, increasing mp , we can obtain a near first-order dynamics
well-known stiff load-bus approximation [20], [21]. The small- (dominance of pole λ1 ), while decreasing this coefficient, the
signal dynamics of the active and reactive power (p̂ and q̂) are two complex poles (λ2 and λ3 ) become dominant, obtaining
obtained by linearizing (1) and (2), with Z∠θ = X∠90◦ , and a second-order behavior. On the other hand, increasing md
modeling the low-pass filters with a first-order description attracts the complex conjugate poles (λ2 and λ3 ) to the real
axis, making the system dynamics more damped. Finally, the
ωc 1
p̂ = [EV cos Φ · φ̂ − P · x̂] (11) coefficient τ which is the time constant of the transient P − ω
s + ωc X droop, makes the system dynamics slower since it attracts the
ωc 1 pole toward the origin. In our practical case, its value ranges
q̂ = [−EV sin Φ · φ̂ − Q · x̂] (12)
s + ωc X between 0 and ωc , and it can be used for finetuning purposes,
GUERRERO et al.: WIRELESS-CONTROL STRATEGY FOR PARALLEL OPERATION OF DG INVERTERS 1465

TABLE II
PARAMETERS OF THE POWER-SHARING CONTROLLER

Fig. 4. Family of root locus diagram: (a) md = 0, 6 · 10−5 , and 12 · 10−5 for 0 ≤ mp ≤ 4 · 10−4 . (b) 0 ≤ τ −1 ≤ 10.

since the influence over the dominant pole location is small, as

shown in Fig. 4(b).
The nominal output inductance of the inverter X should
be selected by taking into account the following tradeoff.
Increasing the output impedance reduces the circulating current
produced by the differences between the power lines or between
the unit parameters, such as the output filter values or the
frequency and amplitude set points. Nevertheless, as Fig. 5
depicts, increasing X also reduces the maximum active power Fig. 5. Active power versus power angle: (a) Stability limits. (b) Effect of
that can be delivered to the common bus increasing output impedance.

EV which leads to a good decoupling between P/Q and the fre-

Pmax = . (16)
X quency/amplitude output voltage.
Thus, there is a tradeoff in the output-impedance design. In
On the contrary, decreasing the output impedance forces the our case, we must make sure that the output impedance has a
power angle to become smaller proper value in a range of full load conditions (Qmax ) from

 capacitive to inductive. This value also affects the selection of
−1 PX
φ = sin . (17) the kL coefficient, since
EV
X ∗ − kL Qmax < X < X ∗ + kL Qmax . (20)
In that case, the approximations sin θ ≈ θ and cos θ ≈ 1 are
good enough, and consequently, the active and reactive power
expressions become
V. DSP-C ONTROLLER I MPLEMENTATION
EV
P ∼
= ·φ (18) Fig. 6 depicts the block diagram of the proposed power-
X sharing controller, which includes the transient P − ω droop
V and the adaptive output impedance loop. The average active
Q∼= · (E − V ) (19)
X power P, without the dc component, can be obtained by
1466 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006

Fig. 6. Block diagram of the proposed wireless power-sharing controller.

multiplying the output voltage νo by the output current io and

filtering this product using a bandpass filter [30]
ωc s
P̃ = −1
Pi (21)
(s + τ )(s + ωc )

where τ −1 must range between 0 and ωc . This expression is

derived by combining the high-pass filter (10), which avoids
the dc component, and the low-pass filter (11), which averages
the instantaneous active power. In order to adjust the output-
voltage frequency, (9) is implemented, which corresponds to a
PD controller applied to the transient active power signal P.
The average reactive power is obtained by delaying the
output voltage by 90◦ through a circular buffer, multiplying
by io , and using a low-pass filter. These filters were dis-
cretized through the bilinear transformation, obtaining infinite-
impulse response solutions. The virtual output impedance is
implemented by using the adaptive output impedance, which Fig. 7. Bode diagram of the virtual output impedance.
is regulated by the reactive power, as (8) shows.
The fundamental and harmonic components of the output
THD, in contrast with the conventional techniques. In addition,
current can be extracted by using a bank of bandpass filters.
this sum of filters can be easily implemented in a digital signal
Each filter extracts the selected component, following the pro-
processor (DSP) by using finite-impulse response digital filters.
cedure described in [32]
2ki s
Hi (s) = (22) VI. S IMULATION AND E XPERIMENTAL R ESULTS
s2 + 2ki s + ωi2
The proposed controller was simulated for a two-parallel-
where ki = ζi ωi is the coefficient of the filter, ζi is the damp- inverter system sharing a load in order to show the feasibility of
ing factor, and ωi is the frequency of the i-harmonic (i = the proposed controller. Fig. 8(a) and (b) depicts the transient
3, 5, . . . , 11). Note that each filter implementation includes the response of a two-parallel-inverter system at the startup, and
resistive gain coefficient Ri . Using (5) and (22), the following for load step changes. It is shown that when the system stays
virtual impedance expression can be found: synchronized, the load changes do not affect power-sharing
 n accuracy. Fig. 8(c) and (d) shows the reduced circulating cur-
2k1 s2 2ki s
ZV (s) = LD 2 2 + Ri 2 2.
rent between the inverters, and the capability to restore the
s + 2k1 s + ω1 i=3
s + 2ki s + ωi frequency in front of the conventional droop method. Notice
odd
(23) that when the startup transient ends, the power sharing is
unaffected by the load step changes.
Fig. 7 shows a Bode-diagram representation of this impedance. Two 6-kVA single-phase inverter units were built and tested,
By properly adjusting Ri coefficients, we can obtain a good confirming experimentally the validity of the proposed ap-
tradeoff between harmonic current sharing and output-voltage proach. Each inverter consisted of a single-phase IGBT full
GUERRERO et al.: WIRELESS-CONTROL STRATEGY FOR PARALLEL OPERATION OF DG INVERTERS 1467

Fig. 8. Transient response at the startup and for load step changes using the proposed wireless controller. (a) Active and reactive power waveforms. (b) Virtual
output impedances. (c) Output currents and circulating current. (d) Frequency restoration of the wireless controller.

bridge with a switching frequency of 20 kHz and an LC

output filter, with the following parameters: L = 500 µH, C =
100 µF, Vin = 400 V, νo = 220 Vrms /50 Hz. The impedances
of the lines connected between the inverters and the load were
intentionally unbalanced, ZL1 = 0.12 + j0.028 Ω, and ZL2 =
0.24 + j0.046 Ω. The controllers of these inverters were based
on three loops: an inner current loop, an outer PI controller that
Fig. 9. Equivalent circuit of two units supplying a common load.
ensures voltage regulation [21], and the load-sharing controller,
based on (8)–(10), using the parameters shown in Table II. The
first two loops were implemented into a TMS320LF2407A, static bypass switch is turned ON, and the soft-start operation
fixed-point 40-MHz DSP from Texas Instruments, and the and the droop-based control are initiated.
wireless load-sharing controller was performed by using a The outstanding features of the parallel system are experi-
TMS320C6711, floating-point 200-MHz DSP. The connec- mentally evaluated in the case of the two-unit system sharing
tion between the two-DSPs was made through the host-port- a linear load, as depicted in Fig. 9. Fig. 10 shows the output
interface characteristic of the floating-point processor. The currents of each unit and the circulating current (i1 –i2 ). These
voltage sampling frequency was rated at 10 kHz, and the current results show an excellent dynamic response of the proposed
sampling frequency was at 20 kHz. The DSP controller also controller for load step changes. As it can be seen, the circu-
includes a PLL block in order to synchronize the output voltage lating current remains very small, even for no-load conditions.
of the inverter in frequency and phase with the common bus. Fig. 11 depicts a detail of the steady state and the dynamic
When the output-voltage synchronization is completed, the performances of the system, revealing the equal current sharing.
1468 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006

Fig. 10. Dynamic performance of the output currents and the circulating current (X-axis: 100 ms/div, Y -axis: 20 A/div).

Fig. 11. Waveforms of the output currents and the circulating current. (a) Steady state. (b) Transient response.(X-axis: 20 ms/div, Y -axis: 20 A/div).

Another experimental test consists in connecting one inverter

while the other is supplying the load all the time. As can be seen
in Fig. 12, transient output currents are not higher than inverter
nominal current, and the transient circulating current stays low.
Finally, the behavior of the two-parallel-inverter system was
evaluated when a nonlinear load with a crest factor of three is
supplied. Fig. 13 shows the load voltage and current, and the
output current of the two units. The measured output-voltage
THD was 2.1%. As it can be seen, the load sharing capability is
also very good when supplying nonlinear loads.

VII. C ONCLUSION
In this paper, a novel power-sharing controller for parallel
inverters has been proposed. Based on the droop method, the
controller avoids the use of communication signals among
the units. In a clear-cut contrast with the conventional droop
method, the presented controller is able to modify the dynamic Fig. 12. Transient response of the output currents and the circulating current
response of the paralleled system by tuning the control gain when one inverter is connected. (X-axis: 200 ms/div, Y -axis: 20 A/div).
parameters. The controller consists of a transient frequency
droop loop and an adaptive output-impedance loop which al- In addition, the proposed adaptive virtual output-impedance
low sharing active and reactive power, respectively, without further improves hot-swap operation and provides nonlinear
sacrificing frequency or amplitude regulation in steady state. load sharing in parallel-connected DG inverters. Furthermore,
GUERRERO et al.: WIRELESS-CONTROL STRATEGY FOR PARALLEL OPERATION OF DG INVERTERS 1469

Fig. 13. Waveforms of the parallel system sharing a nonlinear load. (a) Output voltage and load current (X-axis: 5 ms/div, Y -axis: 40 A/div). (b) Output currents
and circulating current (X-axis: 10 ms/div, Y -axis: 20 A/div).

it is interesting to note that the proposed control can be extended [13] C.-C. Hua, K.-A. Liao, and J.-R. Lin, “Parallel operation of inverters
to other output impedances if necessary. for distributed photovoltaic power supply system,” in Proc. IEEE PESC,
2002, pp. 1979–1983.
Experimental results have been reported to validate the pro- [14] J. Sachau and A. Engler, “Static and rotating grid formation for modularly
posed control approach, showing good power sharing when and expandable island grids,” in Proc. EPE Conf., Laussane, Switzerland,
supplying linear and nonlinear loads, even when the line im- 1999, pp. 1–10.
[15] H. Matthias and S. Helmut, “Control of a three phase inverter feeding
pedances are different. The excellent features of this wireless an unbalance load and operating in parallel with other power sources,”
controller show its applicability to parallel-connected inverters in Proc. EPE-PEMC Conf., 2002, pp. 1–10.
in distributed power systems, such as DG systems or micro- [16] M. C. Chandorkar and D. M. Divan, “Control of parallel connected in-
verters in standalone ac supply system,” IEEE Trans. Ind. Appl., vol. 29,
grids. no. 1, pp. 136–143, Jan./Feb. 1993.
[17] T. Kawabata and S. Higashino, “Parallel operation voltage source invert-
ers,” IEEE Trans. Ind. Appl., vol. 24, no. 2, pp. 281–287, Mar./Apr. 1988.
[18] A. Tuladhar, H. Jin, T. Unger, and K. Mauch, “Parallel operation of single
R EFERENCES phase inverter modules with no control interconnections,” in Proc IEEE
[1] R. H. Lasseter et al., “White paper on integration of distributed energy APEC, 1997, pp. 94–100.
resources. The CERTS microgrid concept,” Consortium for Electric Re- [19] U. Borup, F. Blaabjerg, and P. N. Enjeti, “Sharing of nonlinear load
liability Technology Solutions, Berkeley, CA, Tech. Rep. LBNL-50 829, in parallel-connected three-phase converters,” IEEE Trans. Ind Appl.,
2002. vol. 37, no. 6, pp. 1817–1823, Nov./Dec. 2001.
[2] K. Ro and S. Rahman, “Two-loop controller for maximizing performance [20] E. A. A. Coelho, P. Cabaleiro, and P. F. Donoso, “Small signal stability
of a grid-connnected photovoltaic-fuel cell hybrid power plant,” IEEE for single phase inverter connected to stiff ac system,” in Proc. IEEE-IAS
Trans. Energy Convers., vol. 13, no. 3, pp. 276–281, Sep. 1998. Annu. Meeting, 1999, pp. 2180–2187.
[3] K. Sedghisigarchi and A. Feliachi, “Dynamic and transient analysis of [21] J. M. Guerrero, L. García de Vicuña, J. Matas, M. Castilla, and
power distribution systems with fuel cells—Part II: Control and stability J. Miret, “A wireless controller to enhance dynamic performance of
enhancement,” IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 429–434, parallel inverters in distributed generation systems,” IEEE Trans. Power
Jun. 2004. Electron., vol. 19, no. 5, pp. 1205–1213, Sep. 2004.
[4] G. N. Kariniotakis and G. S. Stavrakakis, “A general simulation [22] M. C. Chandorkar, D. M. Divan, Y. Hu, and B. Barnajee, “Novel archi-
for the accurate assessment of isolated diesel-wind turbines systems tectures and control for distributed UPS systems,” in Proc. IEEE APEC,
interactions—Part I: A general multimachine power system model,” IEEE 1994, pp. 683–689.
Trans. Energy Convers., vol. 10, no. 3, pp. 584–590, Sep. 1995. [23] S. J. Chiang, C. Y. Yen, and K. T. Chang, “A multimodule parallelable
[5] C. Wekesa and T. Ohnishi, “Utility interactive ac module photovoltaic series-connected PWM voltage regulator,” IEEE Trans. Ind. Electron.,
system with frequency tracking and active power filter capabilities,” vol. 48, no. 3, pp. 506–516, Jun. 2001.
in Proc. IEEE-PCC Conf., 2002, pp. 316–321. [24] A. Engler, “Control of parallel operating battery inverters,” in Proc. PV
[6] G. Weiss, Q.-C. Zhong, T. C. Green, and J. Liang, “H∞ repetitive control Hybrid Power Syst. Conf., 2000, CD-ROM.
of dc-ac converters in microgrids,” IEEE Trans. Power Electron., vol. 19, [25] A. Tuladhar, H. Jin, T. Unger, and K. Mauch, “Control of parallel inverters
no. 1, pp. 219–230, Jan. 2004. in distributed ac power systems with consideration of line impedance,”
[7] S. R. Wall, “Performance of inverter interfaced distributed generation,” IEEE Trans. Ind. Appl., vol. 36, no. 1, pp. 131–138, Jan./Feb. 2000.
in Proc. IEEE/PES Transmiss. Distrib. Conf. Expo., 2001, pp. 945–950. [26] M. N. Marwali, J.-W. Jung, and A. Keyhani, “Control of distributed
[8] M. Karimi-Ghartemani and R. Iravani, “A method for synchronization generation systems—Part II: Load sharing control,” IEEE Trans. Power
of power electronic converters polluted and variable-frequency envi- Electron., vol. 19, no. 6, pp. 1551–1561, Nov. 2004.
ronments,” IEEE Trans. Power Syst., vol. 19, no. 3, pp. 263–2004, [27] A. R. Bergen, Power Systems Analysis. Englewood Cliffs, NJ: Prentice-
Aug. 2004. Hall, 1986.
[9] P. L. Villeneuve, “Concerns generated by islanding,” IEEE Power Energy [28] J. M. Guerrero, L. García de Vicuña, J. Matas, M. Castilla, and J. Miret,
Mag., vol. 2, no. 3, pp. 49–53, May/Jun. 2004. “Output impedance design of parallel-connected UPS inverters with wire-
[10] R. H. Lasseter and P. Piagi, “Providing premium power through distrib- less load-sharing control,” IEEE Trans. Ind. Electron., vol. 52, no. 4,
uted resources,” in Proc. IEEE 33rd HICSS, 2000, pp. 1–9. pp. 1126–1135, Aug. 2005.
[11] S. Barsali, M. Ceraolo, P. Pelachi, and D. Poli, “Control techniques of [29] L. Chen, X. Xiao, C. Gong, and Y. Yan, “Circulating current’s character-
dispersed generators to improve the continuity of electricity supply,” istics analysis and the control strategy of parallel system based on double
in Proc. IEEE PES Winter Meeting, 2002, pp. 789–794. close loop controlled VSI,” in Proc. IEEE PESC, 2004, pp. 4791–4797.
[12] S.-W. Park, I.-Y. Chung, J.-H. Cho, S.-I. Moon, and J.-E. Kim, “Control [30] J. M. Guerrero, L. García de Vicuña, J. Matas, and J. Miret, “Steady-state
schemes of the inverter-interfaced multi-functional dispersed generation,” invariant-frequency control of parallel redundant uninterruptible power
in Proc. IEEE PES Winter Meeting, 2003, pp. 1924–1929. supplies,” in Proc. IEEE IECON, 2002, pp. 274–277.
1470 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006

[31] T. Ishikawa, H. Funato, T. Ohtaki, and K. Kamiyama, “Transmission Luis García de Vicuña received the M.S. and Ph.D.
power control using variable inductance with feedforward- and feedback- degrees in telecommunications engineering from
based power controllers,” in Proc. IEEE PEDS, 1999, pp. 46–51. the Universitat Politècnica de Catalunya, Barcelona,
[32] P. Mattavelli and F. Marafao, “Repetitive-based control for selective har- Spain, in 1980 and 1990, respectively, and the Dr.Sci.
monic compensation in active power filters,” IEEE Trans. Ind. Electron., degree from the Université Paul Sabatier, Toulouse,
vol. 51, no. 5, pp. 1018–1024, Oct. 2004. in France, 1992.
From 1980 to 1982, he was an Engineer with
Control Applications. He is currently an Associate
Josep M. Guerrero (S’01–M’03) received the Professor with the Department of Electronic Engi-
B.S. degree in telecommunications engineering, the neering, Universitat Politècnica de Catalunya, where
M.S. degree in electronics engineering, and the he teaches power electronics. His research interests
Ph.D.degree in power electronics from the Univer- include power electronics modeling, simulation, and control, active power
sitat Politècnica de Catalunya, Barcelona, Spain, in filtering, and high-power-factor ac–dc conversion.
1997, 2000, and 2003, respectively.
From 1998 to 2004, he was an Assistant Professor
in the Department of Automatic Control Systems
and Computer Engineering, Universitat Politècnica
de Catalunya. In 2004, he became a Senior Lecturer
at the same university, where he teaches digital signal Miguel Castilla received the M.S. and Ph.D. degrees
processing, control theory, and microprocessors. Since 2004, he has been
in telecommunications engineering from the Univer-
the Responsible of the Sustainable Distributed Generation and Renewable
sitat Politècnica de Catalunya, Barcelona, Spain, in
Energy research group at the Escola Universita d’Enginyeria Tècnica Industrial
1995 and 1998, respectively.
de Barcelona (EUETIB). His research interests include DSP-/FPGA-based Since 2002, he has been an Associate Professor
control, uninterruptible power systems, inverters for photovoltaic applications,
with the Department of Electronic Engineering, Uni-
and wind energy conversion in microgrids.
versitat Politècnica de Catalunya, where he teaches
Dr. Guerrero is an Associate Editor of the IEEE TRANSACTIONS ON
analog circuits and power electronics. His research
INDUSTRIAL ELECTRONICS. He is a Guest Editor of the Special Issue
interests are in the areas of modeling, simulation,
of the IEEE TRANSACTIONS ON POWER ELECTRONICS “Power Electron-
and control of dc-to-dc power converters and high-
ics for Wind Energy Conversion” and the Special Section of the IEEE
power-factor rectifiers.
TRANSACTIONS ON INDUSTRIAL ELECTRONICS “Uniniterruptible Power
Supply (UPS) Systems.” He has organized and chaired sessions at several IEEE
IECON, APEC, and PESC Conferences. He is listed in Marquis’ Who’s Who in
the World and Marquis’ Who’s Who in Science and Engineering.

José Matas received the B.S., M.S., and Ph.D. Jaume Miret (M’98) received the B.S. degree in
degrees in telecommunications engineering from telecommunications, the M.S. degree in electronics,
the Universitat Politècnica de Catalunya, Barcelona, and the Ph.D. degree from the Universitat Politècnica
Spain, in 1988, 1996, and 2003, respectively. de Catalunya, Barcelona, Spain, in 1992, 1999, and
Since 1997, he has been an Associate Profes- 2005, respectively.
sor with the Department of Electronic Engineering, Since 1993, he has been an Assistant Professor
Universitat Politècnica de Catalunya. His research with the Department of Electronic Engineering, Uni-
interests include power-factor-correction circuits, versitat Politècnica de Catalunya. His research inter-
distributed power systems, and nonlinear control. ests include dc–ac converters, active power filters,
and digital control.