Electric Power Systems Research 80 (2010) 1215–1221

Contents lists available at ScienceDirect

Electric Power Systems Research

journa l homepage: www.e lsev ier .com/ locate /epsr

An improved control scheme based in droop characteristic for microgridconverters

P. Arboleya ∗, D. Diaz, J.M. Guerrero, P. Garcia, F. Briz, C. Gonzalez-Moran, J. Gomez AleixandreUniversity of Oviedo, Electrical Engineering Department, 33204 Gijon, Asturias, Spain

a r t i c l e i n f o

Article history:Received 1 June 2009Received in revised form 3 March 2010Accepted 20 April 2010Available online 9 June 2010

a b s t r a c t

In the present work, an improved version of the conventional-droop control for microgrid converter ispresented. The modifications added to the control are based on a feed-forward current control that allowsthe converter to work in several modes, both when it is grid connected or in island. The use of this controlrepresents the main contribution of this paper, permitting the inverter to work as a grid supportingsource or ancillary services provider when it works grid connected. In this mode the converter varies the

Distributed generationMicrogridsIP

injected active and reactive power with the variation of voltage module and frequency using the samemain control loop as when it is working in island mode.

1

tdtdacistehetrpsasctaa

mA

0d

nvertersower-quality

. Introduction

Distributed generation (DG) technologies have achieved a dras-ic increase during the last years derived from recent technologicalevelopments [1]. The influence of this type of generation onhe distribution network stability can be positive or negativeepending on the distribution system and the DG system oper-ting characteristics [3]. The massive installation of DG systemsan produce an important reduction of the electrical losses bothn transmission and distribution networks, as well as CO2 emis-ions. Another consequence would be a significant reduction inhe investment on electrical facilities. Additionally, production ofnergy from waste heat through co-generation or combined coolingeat and power (CCHP) can give rise to an integrated high efficiencynergy system. However, an increased use of DG systems in elec-rical networks without correct addressing coordination issues canesult in a harmful influence in the electrical network, includingroblems in voltage regulation, voltage flicker generation due toudden changes in generation levels of DG, increase of harmonics,nd variations in short circuits levels, affecting the reliability andafety of the distribution system [4]. Fortunately, those problems

an be avoided with an organized introduction of these resources inhe electrical networks [5]. Additionally, the DG system can be useds ancillary services provider for voltage control, load regulationnd spinning reserve [6].

∗ Corresponding author at:. University of Oviedo, Electrical Engineering Depart-ent, Campus de Viesques s/n, Edificio Dept. 4. Despacho 4.02.09, 33204 Gijn,sturias, Spain. Tel.: +34 985182283; fax: +34 985182068.

E-mail address: arboleyapablo@uniovi.es (P. Arboleya).

378-7796/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.epsr.2010.04.003

© 2010 Elsevier B.V. All rights reserved.

The most suitable way to insert DG systems into the electri-cal network is through the use of microgrids. A microgrid, can bedefined as a cluster of loads and microsources operating as a singlecontrollable system providing both power and heat to its local area[7]. There exist different microgrid management philosophies thatcan be roughly categorized into three different groups [9]. The firstgroup consists on a set of microgrids with a physical prime movermanagement in which a large unit absorbs all transient active andreactive power imbalances to maintain the voltage magnitude andfrequency. The concept is very similar to the one used in conven-tional centralized generation systems. The cost of the central unitand the loss of stability when a fault occurs in that unit are the mainproblems of this approach. In the second group, the control systemis based on a virtual prime mover. In this case a central control unitmeasures the microgrid state variables, and dispatches orders tomicrosources using a fast telecommunication system. This controlscheme avoids the high cost of the central physical prime mover butthe communication system bandwidth limits the expansion of themicrogrid and additionally, a back-up system is needed in case ofcommunication failure. The third approach is based on a distributedcontrol. In this case, each unit responds automatically to variationsin the local state variables. A number of researchers consider thistype of control the most appropriate because neither a communica-tion system nor a large central unit is needed [7,10,11]. Nowadays,there are some important projects on microgrids launched aroundthe world [8,9] using the different microgrid management philoso-

phies abovementioned.

Control of local state variables is commonly implemented inmicrogrid converters using a so called droop characteristic control.This type of control was first introduced for parallel connectedinverters in a standalone system [12]. Recently, droop control has

1216 P. Arboleya et al. / Electric Power Systems Research 80 (2010) 1215–1221

baptmimep

diimbdumactotsd

2

scuavcclItwpaaaaqom

Fig. 1. Main control scheme.

een extended to microgrid distributed control [7,2]. A detailednalysis of the behavior of droop control based generators wasresented in [13]. However, some researchers combine a dis-ributed control with some kind of communication between the

icrosources [14,15]. In those cases, the microgrid primary controls distributed but secondary control loops are based on telecom-

unications. These control loops improve the power-quality andconomic efficiency. When a telecommunication fault occurs, therimary controller acts as a back-up system.

This paper proposes an improved control scheme based onroop characteristic control. The proposed control system uses an

nner current control loop in grid connected mode that modifies thenjected active and reactive power as a function of the grid voltage,

agnitude and frequency, therefore providing a grid support capa-ility. In the island mode the power converter can operate in threeifferent working submodes: (1)conventional-droop mode, whichses a conventional-droop characteristic control, (2) power-qualityode, which adapts the droops to provide the voltage magnitude

nd frequency nominal values and (3)sync mode, in which the droopharacteristics are changed while the phase and voltage magni-ude of the microgrid voltage are synchronized with the grid inrder to get a smooth connection transient. The proposed controlopology allows the inverter to work on several modes and makeoft changes between a droop characteristic control and an inverseroop control.

. Control strategies

Different control schemes compose the overall proposed controlystem. In this section, the different working modes of the proposedontrol are analyzed. The block diagram of the main control loopsed for both working conditions is presented in Fig. 1, where Udref

nd Uqref are the voltage references, Ud and Uq are the measuredoltages after the filter (see Fig. 2), and Id and Iq are the measuredurrents before the filter (see Fig. 2). It can be observed that thisontrol loop is based on a traditional droop characteristic controloop improved with the introduction of feed-forward bias currentsdbias and Iqbias. The use of these currents is the main contribution ofhis work and allows the converter to make a grid supporting laborhen it is working in grid connected mode. In this situation, theroposed feed-forward control will make the converter to work asn inverse droop characteristic control, varying the injected activend reactive power as a function of measured voltage magnitude

nd frequency. For all island modes Id bias and Iq bias are disablednd set to zero. It should be noted that the voltage reference of the-axis, Uqref , is set to zero for both situations while the calculationf Udref and frequency reference fref will depend on the workingode as it will be described as follows.

Fig. 2. Injection unit scheme.

2.1. Island mode

Three possible working conditions are considered in islandmode: conventional-droop mode, power-quality mode and syncmode. In either situation it is necessary to calculate the voltage Eq.(1) and frequency Eq. (2) references of the microgrid as it can beobserved in Fig. 1, where P and Q are the measured active and reac-tive power respectively, P0 and Q0 are the rated active and reactivepower U∗

0 and f ∗0 are voltage and frequency commands that depends

on the selected island mode.

Udref = U∗0 − Kp(Q − Q0) (1)

fref = f ∗0 − Kq(P − P0) (2)

The droop characteristic constants, Kp and Kq, are calculatedusing Eqs. (3) and (4), where fmax and Umax are the maximum per-mitted frequency and voltage in island mode and Pmax and Qmax arethe maximum active power and reactive power that can be injectedby the converter.

The choice of the droop constants Kp and Kq affects to the net-work stability. In general terms, we can assert that the higherthe values of the droop constants, the lower the stability marginof the system. Some methods based on trial and error procedure[8,20], have been proposed to obtain the adequate values for theseconstants. However, to date, there is not too much work relatedwith the analytical selection of this values considering microgriddynamic. In [19] a methodology based on bifurcation theory is pre-sented and discussed. The iterative methodology to obtain the bestvalues depends not only on the studied generator parameters butalso on the network parameters and other generator parameters.

Kp = fmax − f0Pmax − P0

(3)

Kd = Umax − U0

Qmax − Q0(4)

2.1.1. Conventional-droop modeThis control strategy allows the inverter to work as a classical

droop mode where the values of voltage and frequency are fixedaccording to Eqs. (6) and (5).

f ∗0 = f0 ≡ rated frequency (5)

U∗0 = U0 ≡ rated voltage (6)

2.1.2. Power-quality modeThe Power-quality mode changes the position of the droop char-

acteristic in order to recover the rated frequency and voltage when

a change in the load occurs. As can be observed in Fig. 3, when theconventional-droop mode is activated and the microgrid reactiveload is reduced from Q0 to Q2, the operating point is moved from Ato B, increasing the voltage of the microgrid to U2. If the power-quality mode is activated at that point, the droop characteristic

P. Arboleya et al. / Electric Power System

iUaa

U

f

KrKp

piooqtiftgvotampgmirs

2

gtf

U

f

ita

Fig. 3. Power-quality mode.

s modified and the voltage reference in the inverter changes to0. The same behavior can be observed in the frequency when thective load varies. In this mode the voltage, U∗

0, and frequency, f ∗0 ,

re calculated using Eqs. (7) and (8) respectively.

∗0 = U0 −

(Kpq + Kiq

s

)(Ud − U0) (7)

∗0 = f0 −

(Kpp + Kip

s

)(f − f0) (8)

where f is the measured frequency, Ud is the measured voltage,pq and Kiq are respectively the proportional and integral gain of aeactive power PI regulator, s is the Laplace operator, and Kpp andip are respectively the proportional and integral gain of the activeower PI regulator.

When a variation in the load occurs, the values of the PI regulatorarameters, determine the speed of the system to return to its nom-

nal values of voltage magnitude and frequency. In cases where onlyne generator is connected working in island model a fast responsef this mode can produce negligible voltage magnitude and fre-uency variations. In this case, if Fig. 3 is analysed, this means thathe correction of the droop characteristic position is faster thannner conventional-droop control. The effect is that the movementrom A to C is nearly horizontal and the voltage is not affected. Ifhis configuration is used in a multi-unit scenario where one of theenerators is working in power-quality mode and the rest of con-erters are working in grid supporting mode a very fast correctionf the droop characteristic position will reduce the support labor ofhe rest of the generators. An elevated percentage of the load vari-tion will be assumed by the generator working in power-qualityode. We could end up with an overloaded converter working in

ower-quality mode and other underloaded converters working inrid supporting mode without making a real grid support labor. Inulti-unit cases where all converters are connected through high

mpedance lines working in power-quality mode the speed of cor-ection of the droop position has a very high influence in the loadharing and must be studied in each particular scenario.

.1.3. Sync modeThis method is used to synchronize the main grid and the micro-

rid. The magnitude and phase of the microgrid voltage are equalo the main grid values after the synchronization process. U∗

0 and∗0 are obtained according to Eqs. (9) and (10) respectively.

∗0 = U0 −

(K ′

pq +K ′

iq

s

)(|Ug | − |U�g |) (9)

∗0 = f0 −

(K ′

pp +K ′

ip

s

)(�g − ��g) (10)

where |Ug | is the magnitude of the main grid voltage vector, |U�g |s the magnitude of the microgrid voltage vector, �g is the phase ofhe main grid voltage vector, ��g is the phase of the microgrid volt-ge vector. It should be noted that the PI controller gains K ′

pp and K ′ip

s Research 80 (2010) 1215–1221 1217

are critical parameters for the power-quality in the microgrid sincea large phase displacement between the voltage, �g − ��g , with ahigh value of these parameters could give rise to a large variation ofthe microgrid frequency. In order to get the highest synchroniza-tion speed, maintaining the frequency into a determined band, theoutput of the PIs must be limited. Moreover it must be taken intoaccount that a converter can be working in sync mode as a masterwith other converter working as slaves following the voltage andfrequency imposed by the master. A very fast variation of the fre-quency can cause problems in the slaves to follow the frequencyreference and this situation can result in the loss of the stability.

2.2. Grid connected mode

In this mode, the voltage amplitude and frequency references ofthe inverter, Udref and fref , are continuously updated depending onthe frequency and voltage amplitude of the main grid. The currentsIdbias and Iqbias are enabled. Two different strategies to obtain thevalue of those variables are proposed in this paper: grid supportingmode and grid feeding mode. The first one constitutes an innovationdue to the use of the feed-forward bias currents to make a support-ing labor since the rest of the control remains similar to island modewith the exception that Udref and fref are directly measured values.

2.2.1. Grid supporting modeIn this mode the inverter works as a grid supporting source since

it varies the injected active and reactive power depending on themain grid voltage and frequency excursions from the nominal val-ues. In this mode the currents are calculated using Eqs. (11) and(12). It must be noticed that, at the time of connection of the islandwith the main grid, the inverter injects the active and reactive ratedpower values if the voltage magnitude and frequency are in theirnominal values, in this case 400 V and 50 Hz. In this situation if thevoltage magnitude or the frequency suffers any variation the gridsupporting control will respond with a variation in the injectedactive or reactive power. Another possibility is synchronizing theinverter when the grid values are not the nominal ones. In this casethe active and reactive power will differ from the rated values atthe time of connection of the island with the main grid to make thegrid support. This working mode permits the connection with themain grid in cases where problems of stability arise contributingto the mitigation of those problems. Without this kind control anyattempt of connection under those situations can contribute to theloss of the main grid stability. This is a clear example of DG systemworking as a provider of ancillary services.

Iq bias =[

(Q − Q0) − 1Kq

(Ud − U0)

][K

′′pq +

K′′iq

s

](11)

Id bias =[

(P − P0) − 1Kp

(f − f0)

][K

′′pp +

K′′ip

s

](12)

where Kp, and Kq are the droop parameters.The use of this control mode must be selected carefully when

the converter is working as a slave coupled to a master converter ina synchronization process. The master converter working in syncmode must vary the voltage magnitude, the voltage frequency orboth, in order to reach the adequate values. In this case, the gridsupport labor of the slaves can cause the master converter loss ofstability. Options to avoid this problem include

• The activation of the grid feeding mode in the slaves convertersduring the synchronization.

• The use of the grid supporting mode varying the voltage magni-tude and frequency nominal values during the sync mode.

1218 P. Arboleya et al. / Electric Power Systems Research 80 (2010) 1215–1221

study

2

taE

I

I

3

d

Fig. 4. Case of

.2.2. Grid feeding modeWhen the inverter is working in this mode, the active and reac-

ive power references are fixed. In this case, the inverter acts asgrid feeding source. The current references are calculated usingqs. (13) and (14).

q bias =[(Qref − Q0)

][K

′′′pq +

K′′′iq

s

](13)

d bias =[(Pref − P0)

][K

′′′pp +

K′′′ip

s

](14)

where Pref , Qref are the active and reactive power references.

. Simulation results

This section presents simulation results of the inverter under theifferent working conditions described in the previous section. The

(single unit).

first subsection describes the operation of a single unit and analy-ses the transitions between the different control modes. The secondsubsection constitutes an example of generators’ coordination in amulti-unit operation scenario. Under this scenario, the most suit-able working mode for each converter is selected according to thegrid conditions.

3.1. Single unit operation

This subsection presents a case study in which the inverter isinitially working in island mode using the power-quality modedescribed in the previous section, (see Fig. 4). The total load ofthe microgrid is 25 kW, the voltage magnitude is 400 V and the

microgrid frequency is 50 Hz.

• At t = 0.1 s the Sync mode is activated when the voltage mag-nitude error is zero, as it can be observed in Fig. 4d. However,there is a phase shift between the microgrid and the main grid

P. Arboleya et al. / Electric Power System

•

•

•

3

fiobmwsbfccstttter

2pa

Fig. 5. Multi-unit scenario.

(10◦) that is corrected before the interconnection by the Syncmode. The convergence behavior of the microgrid voltage phaseis a function of K ′

pp and K ′ip

, (see Eqs. (9) and (10)). During thissynchronization process, the load suddenly changes to 34 kWand 5.2 kVAr at t = 0.2 s. As it can be observed in Fig. 4c–e thisvariation is absorbed by the inverter and does not affect to thesynchronization process.At t = 0.4 s, the microgrid and the main grid get connected. Thegrid voltage and frequency are 400 V and 50 Hz respectively, thesupport is not necessary, so the inverter generates its rated values,25 kW and 0 kVAr, thus the main grid supplies 9 kW and 5.2 kVArto the load, as can be observed in Fig. 4c–e.At t = 0.6 s the load changes to 41.5 kW and 10 kVAr. As theinverter is in Grid supporting mode and the grid voltage and fre-quency remain in their nominal values, the active and reactivepower of the inverter do not change and the main grid increasesthe injected power to 16.5 kW and 10 kVAr.Finally at t = 0.8 s the grid voltage changes from 400 V to 397 V.When the voltage decreases, the inverter starts generating reac-tive power up to 21 kVAr, of which 10 kVAr is consumed bythe microgrid load, and 11 kVAr is absorbed by the main grid,contributing to the grid voltage stability. Similarly, if the fre-quency of the main grid increases, the active power generatedby the inverter would decrease and vice versa. If a conven-tional Grid feeding mode was used, the injected active andreactive power would not vary under these situations and theconverter could contribute to an increment of the main gridinstability.

.2. Multi-unit scenario

The use of the proposed control allows multiple microgrid con-gurations and control philosophies. One scenario could be the usef the converters in a multi-unit microgrid, when the impedanceetween converters is large enough, it could be possible to haveore than one converter working in conventional-droop modeithout loosing the microgrid stability [16,18]. The same control

trategy in a low impedance microgrid would produce severe sta-ility problems [17]. The adopted control depends on multipleactors as the distance between generators, X/R lines ratio, short-ircuit power in the microgrid and in the main grid when theonverters are working in grid connected mode, etc. In this sub-ection one possible scenario showing the coordination betweenwo inserted units is presented. The selected scenario was choseno show the versatility of the control and the inverters’ adapta-ion capability when the network conditions change. The selectedopology is represented in Fig. 5. In this case of study both convert-rs (converter 1 and converter 2) start up islanded and feeding theirespective loads.

The initial conditions are the following: load 1 is 25 kW and loadis 5 kW, contactors 1–3 are open, both converters are working inower-quality mode, the nominal power of converter 1 is 25 kWnd 5 kW for converter 2 and the initial phase shift between the

s Research 80 (2010) 1215–1221 1219

converters and the main grid is 20◦. In Fig. 6e it is shown the phaseshift between converter 1 and the grid and the phase shift betweenconverters 2 and 1.

• At t = 0.15 s, the Sync mode is activated in converter 1 but con-verter 2 remains in power-quality mode, this situation causes areduction in the phase shift between converter 1 and the maingrid. Therefore, the phase shift between converter 2 and con-verter 1 increases.

• At t = 0.2 s the sync mode is activated in converter 2 and the syn-chronization between converter 2 and converter 1 begins. Thevariation in converters 1 and 2 frequencies during their synchro-nization can be observed in Fig. 6c.

• At t = 0.4 s the phase shift between converter 1 and 2 is negligi-ble but there still exists and oscillation in converter 2 frequency.Under this situation contactors 1 and 2 are closed and the gridfeeding mode is activated in converter 2. The oscillation in con-verter 2 frequency generates small voltage magnitude errorbetween converters 1 and 2 causing a reactive power oscilla-tion during the first instants of the reconnection see Fig. 6g).These oscillations are absorbed by converter 1 and the load isnot affected. It must be remarked that at t = 0.4 s, when convert-ers 1 and 2 are connected to the grid, the synchronization processbetween converter 1 and the main grid has not been finished andthere still exists a small phase shift of approximately 5◦betweenthat converter and the main grid see Fig. 6e). This phase shift andthe associated frequency variation also contribute to the reactivepower oscillation when converters 1 and 2 are connected. Thisoscillation can be suppressed by waiting until converter 2 fre-quency is stabilized and limiting the maximum frequency rate ofchange.

• At t = 0.4 s the grid feeding mode is activated in con-verter 2 avoiding the grid supporting mode, as if gridsupporting mode was activated in converter 2 during con-verter 1 synchronization, converter 2 would actuate againstthe necessary changes in frequency and voltage during thesynchronization.

• At t = 0.7 s load 1 increases from 25 kW and 0 kVAr to 37.5 kWand 5 kVA while the converter 2 is still in grid feeding modeso the whole load variation is addressed by converter 1 seeFig. 6d).

• At t = 0.9 s contactor 3 is closed connecting both converters withthe main grid and the grid supporting mode is activate in bothconverters. The voltage magnitude and frequency of the maingrid are 400 V and 50 Hz. As it can be observed in Fig. 6d, bothconverters inject their nominal active power (25 kW and 5 kW)and reactive power (0 kVAr), so the main grid must inject 12.5 kWand 5 kVAr in the microgrid.

• At t = 1 s load 1 increases to 42.5 kW and 10 kVAr while the gridvoltage magnitude and frequency values remain at the nominalvalues (400 V and 50 Hz) so the load variation is assumed by themain grid.

• At t = 1.2 s the grid voltage decreases from 400 V to 397 V andthe injected reactive power of the converters varies from 0 kVArto 25 kVAr in both converters see Fig. 6g).

• At t = 1.4 s the grid voltage rises from 397 V to 402 V makingeach converter to absorb 16.6 kVAr and the main grid to inject43.3 kVAr in the microgrid.

• Finally, at t = 1.6 s, the grid voltage returns to its nominal valueand the reactive power injected by the converters returns to0 kVAr.

4. Conclusions

The use of this control scheme in all converters inside a micro-grid allows to implement several management philosophies. One

1220 P. Arboleya et al. / Electric Power Systems Research 80 (2010) 1215–1221

study

oqwqwpfapa

Fig. 6. Case of

f them could be the configuration of one converter in power-uality mode as a master converter. The rest of the convertersill have a grid supporting configuration so the voltage and fre-

uency references will be set by the master converter and the slavesill make supporting labors. Similarly to the physical prime mover

hilosophy, only the master generates the voltage magnitude andrequency command but the active and reactive power imbalancesre absorbed by all the inverters, contributing to the microgrid sup-orting without a telecommunication system. This method wouldvoid the overrating of the master generator with respect to the

(Multi-Unit).

slaves. Another possibility is the use of all inverters working inpower-quality mode inserting impedances between converters toavoid over currents in the lines during transients.

The main benefits of these control algorithms arise when themicrogrid is connected to the grid. In this situation all the inverters

change their mode of operation to grid supporting mode and aninner current control varies the injected active and reactive poweras a function of the grid voltage magnitude and frequency makinga grid supporting labor and contributing to increase the stability ofthe whole system.

ystem

R

[

[

[

[

[

[

[

[

[

[Coefficients for Frequency and Voltage Regulation in Isolated Microgrids, IEEETransactions on Power Systems, in press.

[20] J.M. Guerrero, J. Matas, V. LuisGarciade, M. Castilla, J. Miret, Decentralized con-

P. Arboleya et al. / Electric Power S

eferences

[1] T. Ackermann, G. Andersson, L. Soder, Distributed generation: a definition,Electric Power Systems Research. 57 (3) (2001) 195–204.

[2] T.C. Green, M. Prodanovc, Control of inverter-based micro-grids, Electric PowerSystems Research. 77 (2007) 1204–1213.

[3] F. Bastiao, P. Cruz, R. Fiteiro, Impact of distributed generation on distributionnetworks, in: 2008 5th International conference on the European electricitymarket, vols. 1 and 2, 2008, pp. 315–320.

[4] R.A. Walling, R. Saint, R.C. Dugan, J. Burke, L.A. Kojovic, Summary of distributedresources impact on power delivery systems, IEEE Transactions on Power Deliv-ery 23 (3) (2008) 1636–1644.

[5] P. Barker, R. de Mello, Determining the impact of distributed generation onpower systems: Part 1—radial distribution systems, in: 2000 IEEE Power Engi-neering Society summer meeting, vols. 1–4, 2000, pp. 1645–1656.

[6] G. Joos, B. Ooi, D. McGillis, F. Galiana, R. Marceau, The potential of distributedgeneration to provide ancillary services, in: 2000 IEEE Power Engineering Soci-ety Summer Meeting, vols. 1–4, 2000, pp. 1762–1767.

[7] R. Lasseter, MicroGrids, in: IEEE Power Engineering Society Winter Meeting,vols. 1 and 2, 2002, pp. 305–308.

[8] N. Hatziargyriou, H. Asano, R. Iravani, C. Marnay, Microgrids, IEEE Power &Energy Magazine 5 (4) (2007) 78–94.

[9] M. Barnes, G. Ventakaramanan, J. Kondoh, R. Lasseter, H. Asano, N. Hatziar-gyriou, J. Oyarzabal, T. Green, Real-world microgrids—an overview, in: IEEEInternational Conference on Systems Engineering, vols. 1 and 2, 2007, pp.

37–44.

10] N. Jayawarna, X. Wu, N. Jenkins, M. Barnes, Stability of a microgrid, in the 3rd IETinternational conference on power electronics, Machines and Drives 1 (2006)316–320.

11] E. Barklund, N. Pogaku, M. Prodanovic, C. Hernandez-Aramburo, T. Green,Energy management system with stability constraints for stand-alone

s Research 80 (2010) 1215–1221 1221

autonomous microgrid, in: IEEE International Conference of Systems Engineer-ing, 2007, pp. 1–6.

12] M. Chandorkar, D. Divan, R. Adapa, Control of parallel connected inverters instandalone AC supply-systems, IEEE Transactions on Industry Applications 29(1) (1993) 136–143.

13] P. Piagi, R.H. Lasseter, Autonomous control of microgrids, in: Power EngineeringSociety General Meeting, vols. 1–9, 2006, pp. 496–503.

14] K. De Brabandere, K. Vanthoumout, J. Driesen, G. Deconinck, R. Belmans, Controlof microgrids, in: IEEE Power Engineering Society General Meeting, vols. 1–10,2007, pp. 3502–3508.

15] K. De Brabandere, B. Bolsens, J. Van den Keybus, A. Woyte, J. Driesen, R. Bel-mans, A voltage and frequency droop control method for parallel inverters,IEEE Transactions on Power Electronics 22 (4) (2007) 1107–1115.

16] A. Arulampalam, M. Barnes, A. Engler, Control of power electronic interfaces indistributed generation Microgrids, International Journal of Electronics 91 (9)(2004) 503–523.

17] A. Engler, N. Soultanis, Droop control in LV-Grids, in: International Conferenceon Future Power Systems, 2005, pp. 142–147.

18] N. Pogaku, M. Prodanovic, T.C. Green, Modeling, analysis and testing ofautonomous operation of an inverter-based microgrid, IEEE Transactions onPower Electronics 22 (2) (2007) 613–625.

19] G. Diaz, C. Gonzalez-Moran, J. Gomez-Aleixandre, A. Diez, Scheduling of Droop

trol for parallel operation of distributed generation inverters using resistiveoutputimpedance, IEEE Transactions on Industrial Electronics 54 (2) (2007)994–1004.