An Improved Droop Control Strategy for Reactive Power Sharing in Islanded Microgrid Han, Hua; Liu, Yao; Sun, Yao; Su, Mei; Guerrero, Josep M.

Transcription

1 Aalborg Unverstet An Improved Droop Control Strategy for Reactve Power Sharng n Islanded Mcrogrd Han, Hua; Lu, Yao; Sun, Yao; Su, Me; Guerrero, Josep M. Publshed n: I E E E Transactons on Power Electroncs DOI (ln to publcaton from Publsher): /TPEL Publcaton date: 2015 Ln to publcaton from Aalborg Unversty Ctaton for publshed verson (APA): Han, H., Lu, Y., Sun, Y., Su, M., & Guerrero, J. M. (2015). An Improved Droop Control Strategy for Reactve Power Sharng n Islanded Mcrogrd. I E E E Transactons on Power Electroncs, 30(6), /TPEL General rghts Copyrght and moral rghts for the publcatons made accessble n the publc portal are retaned by the authors and/or other copyrght owners and t s a condton of accessng publcatons that users recognse and abde by the legal requrements assocated wth these rghts.? Users may download and prnt one copy of any publcaton from the publc portal for the purpose of prvate study or research.? You may not further dstrbute the materal or use t for any proft-mang actvty or commercal gan? You may freely dstrbute the URL dentfyng the publcaton n the publc portal? Tae down polcy If you beleve that ths document breaches copyrght please contact us at vbn@aub.aau.d provdng detals, and we wll remove access to the wor mmedately and nvestgate your clam. Downloaded from vbn.aau.d on: May 01, 2015

2 Ths document downloaded from s the preprnt verson of the fnal paper: H. Han, Y. Lu, Y. Sun, M. Su, and J. M. Guerrero, An Improved Droop Control Strategy for Reactve Power Sharng n Islanded Mcrogrd, IEEE Transactons on Power Electroncs, 2014 An Improved Droop Control Strategy for Reactve Power Sharng n Islanded Mcrogrd Hua Han, Yao Lu, Yao Sun, Me Su, and Josep M. Guerrero School of Informaton Scence and Engneerng, Central South Unversty, Changsha , Hunan Provnce, Chna J. M. Guerrero s wth the Department of Energy Technology, Aalborg Unversty, 9220 Aalborg East, Denmar (Tel: ; Fax: ; e-mal: joz@et.aau.d). Abstract For mcrogrd n slanded operaton, due to the effects of msmatched lne mpedance, the reactve power could not be shared accurately wth the conventonal droop method. To mprove the reactve power sharng accuracy, ths paper proposes an mproved droop control method. The proposed method manly ncludes two mportant operatons: error reducton operaton and voltage recovery operaton. The sharng accuracy s mproved by the sharng error reducton operaton, whch s actvated by the low-bandwdth synchronzaton sgnals. However, the error reducton operaton wll result n a decrease n output voltage ampltude. Therefore, the voltage recovery operaton s proposed to compensate the decrease. The needed communcaton n ths method s very smple, and the plug-and-play s reserved. Smulatons and expermental results show that the mproved droop controller can share load actve and reactve power, mprove the power qualty of the mcrogrd, and also have a good dynamc performance. KEY WORDS:Mcrogrd; droop control; reactve power sharng; low-bandwdth synchronzaton sgnals; voltage recovery mechansm I INTRODUCTION The applcaton of dstrbuted generaton (DG) has been ncreasng rapdly n the past decades. Compared to the conventonal centralzed power generaton, DG unts have advantages of less polluton, hgher effcency of energy utlzaton, flexble nstallaton locaton, and less power transmsson losses. Most of the DG unts are connected to the grd va power electronc converters, whch ntroduces system resonance, protecton nterference, etc. To overcome these problems a mcrogrd concept was frst proposed n the US by the consortum for electrcal relablty technology solutons [1]. Compared to usng a sngle DG unt, mcrogrd could offer superor power management wthn the dstrbuton networs. Moreover, the mcrogrd can operate both n grd-connected mode and slandng mode and beneft both the utlty and customers n economy [2-7]. In slandng mode, the load power n the mcrogrd should be properly shared by multple DG unts. Usually, the droop control method whch mmcs the behavor of a synchronous generator n tradtonal power system s adopted, whch does not need the use of crtcal communcatons [8-14, 21-22]. The actve power sharng s always acheved by the droop control method easly. However, due to effects of msmatched feeder mpedance between the DGs and loads, the reactve power wll not be shared accurately. In extreme stuatons, t can even result n severe crculatng reactve power and stablty problems [11]. To overcome the reactve power sharng ssue, a few mproved methods have been proposed. Specfcally, there are manly three approaches to address the effect of the nterconnectng lne mpedance on droop-based control. The frst approach s to ntroduce the vrtual output mpedance by modfyng the output voltage reference based on output current feedbac [11,13-14,23]. Ths method can reduce the reactve power sharng error by reducng the relatve error of the output mpedances. However, the ntroducton of the vrtual mpedance may lead to degradaton of the system voltage qualty. The second approach s based on sgnal njecton technque. In [15], a certan harmonc sgnal contanng reactve power nformaton s njected nto the output voltage reference of each DG unt, and the output reactve power s regulated accordng to the harmonc power to mprove the accuracy of the reactve power sharng. However, ths method results n output voltage dstorton. In [16], n order to reduce the reactve power sharng errors, the method njects some small dsturbance sgnal contanng

3 2 reactve power nformaton nto the frequency reference of each DG unt. By usng the actve power error before and after the njectng sgnal, ths method can elmnate the reactve power error. However, ths method s a classc event-trggered control and ts stablty s not easy to be guaranteed. Addtonally, the thrd approach s usually based on constructed and compensated method. In [17], the method constructs an ntegral control concernng the common bus voltage to ensure the reactve power sharng. However, n practcal stuaton, the common bus voltage nformaton s dffcult to get. In ths paper, a new reactve power sharng method s proposed. The method mproves the reactve power sharng by changng the voltage bas on the bass of the conventonal droop control, whch s actvated by a sequence of synchronzatonn event through the low bandwdth communcaton networ. It s a cost-effectve and practcal an approach snce only a low bandwdth communcaton networ s requred. Smulaton and expermental results are provded to verfy the effectveness and feasblty of the proposed reactve power sharng method. The paper s organzed as follows. Secton II gves the system confguraton and the reactve power sharng errors analyss wth conventonal droop control. Secton III proposes an mproved reactve power sharng control strategy, and the convergence and robustness s analyzed. Smulaton and expermental results are gven n Secton IV. Secton V gves the concluson. II ANALYSIS OF THE CONVENTIONAL DROOP CONTROL METHOD A. Confguraton and operaton of AC Mcrogrd A classc confguraton of a mcrogrd whch conssts of multple dstrbuted generaton (DG) unts and dspersed loads s shown n Fg.1. The mcrogrd s connected to the utlty through a statc transfer swtch at the PCC. Each DG unt s connected to the mcrogrd through power electronc converter and ts respectve feeder. Ths paper ams to solvng the fundamental actve and reactve power sharng n slandng mode, and the power sharng ssues on harmonc currents s out of the scope of the paper. DC Sde Local controller DC Sde Local controller LC flter Synchron -zaton sgnal LC flter DGn Synchron -zaton sgnal Common load Feeder 1 Local load Feeder n Common Bus Man Grd Statc swtch Feeder Fg.1. Illustraton of the AC mcrogrd confguraton. B. The conventonal Droop Control Low bandwdth synchronzaton sgnal Low-bandwdth communcaton Fg. 2 shows the equvalent model of a DG unt, whch s nterfaced to the common bus of the AC mcrogrd through a power nverter wth a output LCL flter. As shown n Fg.2, E δ s the voltage across the flter capactor, V 0 s the common AC bus voltage. Compared wth the nductance of the LCL flter, the lne resstance can be gnored. Then the mpedance between nverter and the common bus can be descrbed as X (X =ωl ). V dc Fg.2 Model of a DG unt. L f E C f L Power lne P jq V 0 Accordng to the equvalent crcut n Fg. 2, the nverter output apparent power s S, and t can be gven by 2 EV EV cos V S P jq sn j (1) X X From equaton (1), the output actve and reactve power of the DG unts are shown as EV P X EV Q sn cos V 2 X (2) 2

4 3 Usually, the phase shft angle δ s small. Therefore, the real power P and reactve power Q of each DG can be regulated by δ and the output voltage ampltude E, respectvely [24]. Then the conventonal droop control s gven by = m P E = E n Q (3) Where ω and E are the nomnal values of DG angular frequency and DG output voltage ampltude, m and n are the actve and reactve droop slopes, respectvely. P and Q are the measured averaged real and reactve power values through a low pass flter, respectvely. C. Reactve Power Sharng Errors Analyss For smplcty, a smplfed mcrogrd wth two DG unts s consdered n ths secton. Accordng to equatons (2) and (3), the reactve power of the -th DG unt s obtaned V E V Q X V n cos ( cos ) (4) Assume the -th and j-th DG unt are worng n parallel wth the same nomnal capacty and droop slope. Note that shft angle δ s usually vary small (snδ δ, cosδ 1), then the reactve power sharng relatve error wth respect to follows Q can be expressed as Q Qj X j X Qerr Q X V n j j (5) It s shown that, the reactve power sharng relatve error s related to some factors, whch nclude the mpedance X j, the mpedance dfference (X j X ), the voltage ampltude V of PCC and the droop slope n j. Accordng to (5), there are two man approaches to mprove the reactve power sharng accuracy: Increasng mpedance X j and the droop gan n j. Usually, ncreasng mpedance s acheved by the vrtual mpedance [11,13-14], whch requres a hgh-bandwdth control for nverters. Increasng the droop gan n j s a smpler way to reduce the sharng error. However, t may degrade the qualty of the mcrogrd bus voltage, and even affects the stablty of 3 the mcrogrd system [18-20]. III PROPOSED REACTIVE POWER SHARING ERROR COMPENSATION METHOD A. Proposed Droop Controller The proposed droop control method s gven as follows: = mp (6) 1 n n E ( t) E n Q ( t) K Q G E n1 n1 (7) where denotes the tmes of synchronzaton event untl tme t. Accordng to (7), the control s a hybrd system wth contnuous and dscrete trats. In the dgtal mplementaton of the proposed method, the contnuous varables E () t and Q() t are dscretzed wth samplng perod T, and T s greatly less than s the tme nterval between two consecutve synchronzaton events. Therefore, the droop equaton (7) at the -th synchronzaton nterval could be expressed as 1 n n n1 n1 E E n Q K Q G E s (8) where ω and E are the values of DG angular frequency and output voltage ampltude at no-load condton; m and n are the droop gan of frequency and voltage of DG- unt; G n s the voltage recovery operaton sgnal at the n-th synchronzaton nterval, G n has two possble values: 1 or 0. If G n =1, t means the voltage recovery operaton s performed. Q n represents the output reactve power of DG- unt at the n-th synchronzaton nterval. K s a compensaton coeffcent for the DG- unt, ΔE s a constant value for voltage recovery. For smplcty of descrpton, the thrd term of (8) s referred to the sharng error reducton operaton, and the last term s called the voltage recovery operaton. For smplcty, the output voltage for the DG- unt n (8) s wrtten as follows n teratve method. E E n ( Q Q ) K Q G E (9) Therefore, n ts mplementaton, only 1 Q 1 E and should be stored n DSP. To better understand

5 4 the proposed method, a specfc example s gven. If there are two DG unts wth the same capacty worng n parallel, and only the conventonal droop s used. There wll be exsts some reactve power sharng error due to some factors. If the sharng error reducton operaton for each DG unt s performed at the tme, the resultng reactve power sharng error wll decrease. The prncple behnd the sharng error reducton operaton can be understood wth the ad of Fg. 3. If the aforementoned operaton s repeated wth tme, the reactve power sharng error wll converge. However, the assocated operatons wll result n a decrease n PCC voltage. To cope wth the problem, the voltage recovery operaton wll be performed. That s to say f the output voltage of one DG unt s less than ts allowed lower lmt, then the DG unt wll trgger the voltage recovery operaton untl ts output voltage s restored to ratng value. The output voltage of all the DG unts wll be added an dentcal value ΔE to ncrease the PCC voltage. The dea for the voltage recovery operaton can be comprehended by the ad of Fg. 4. E V new V E E Q X E Q 1 Q2 Q 1new Q E E Q X E new Q err Q err new new new new new KQ E = E n Q q KQ 2 2 Fg. 3 Schematc dagram of the shang error reducton operaton Q 1 4 E new V E 0 V E = V +( X / V ) Q new new new new Q1 Q2 new Q err Q err Fg. 4 Schematc dagram of voltage recovery mechansm B. Communcaton setup A DG unt can communcate wth other DG unts by RS232 seral communcaton. Each DG unt has the opportunty to trgger a synchronzaton event on the condton that the tme nterval between two consecutve synchronzaton events s greater than a permssble mnmum value and the output voltage of each DG unt s n the reasonable range. If the output voltage of one DG unt s less than ts allowed lower lmt, t wll as for havng the prorty to trgger a synchronzaton event at once. Untl the constrant whch two consecutve synchronzaton events s greater than a permssble mnmum value s satsfed, the DG unt wth the prorty wll trgger a synchronzaton event, and n ths event, the command for voltage recovery operaton wll be sent to other DG unts. If the communcaton fals (the tme nterval between two consecutve synchronzaton events s greater than a permssble maxmum value), all the error reducton operatons and voltage recovery operatons should be dsabled and the proposed control method s revert bac to the conventonal one. Accordng to the analyss above, such a mcrogrd system only needs a low-bandwdth communcaton. And t s robust to the delay of communcaton. To llustrate ths pont, the control tmng dagram shown n Fg.5 s used. The sharng error operaton and the voltage recovery operaton are performed n update nterval. Samplng operaton occurs n samplng n q Q

6 5 nterval. There s a tme nterval, whch s long enough to guarantee the system havng been n steady state. It s obvous that proposed method s robust to the tme delay because all the necessary operatons only need to be completed n an nterval, not a crtcal pont. -1-th synchronzaton Update nterval 1 Q 1 E Samplng nterval -th synchronzaton tw Update nterval +1-th synchronzaton Q Samplng nterval E t w t Assume the PCC voltage value satsfy the followng V V V 1 (16) 1 Subtractng (13) from (14), t yelds X +1 j 1 X 1 E E = ( Q Q )+ ( Q Q ) (17) j j j j V V where X X X. j Combnng the expresson (13) and (17), then Q rq [ ] ( ) Q Q (18) +1 X +1 j j V nx j V Fg. 5 Control tmng dagram of one DG wth the two consecutve synchronzaton events. C. Convergence Analyss In ths subsecton, the convergence of the proposed method wll be proved. Wthout loss of generalty, the sharng reactve power error between DG- and DG-j wth the same capacty wll be analyzed. Accordng to (8), the reactve droop equaton for DG-j can be expressed as 1 n n j j j j j n1 n1 (10) E E n Q K Q G E Subtractng (10) from (8), then 1 n j = j j n1 E nq KQ (11) where n=n j =n, K=K j =K., and ΔE j s the voltage magntude dervaton of DG and j n the -th control perod; ΔQ j s the reactve power sharng errors. Smlarly, we can get equaton (11) n the +1-th nterval. 1 1 n j = j j n1 E nq KQ (12) Combnng (11) and (12), t yelds: E E nq nq K Q (13) +1 1 j j j j j Accordng to the feeder characterstc, as shown n (2), the followng expressons can be obtaned. E = ( Q X Q X ) (14) j 1 V j j E Q X Q X (15) 1 j = ( ) V j j X j V n K where r 1. Accordng to the contracton X n j V mappng theorem, f r 1 and X 0, then reactve power sharng error wll converge to zero. However, X 0, we should also consder the effect of the second term of (18). Accordng to the feeder characterstc, as shown n (1), we have Q 1 +1 ( E ) Q X E V (19) Because of the voltage recovery mechansm, we can ensure E mn E Emax for all. +1 V Q Q ( Emax Emn ) (20) X Therefore, the second term of (18) s bounded. Accordng to analyss above, t can be concluded that the reactve power sharng error s also bounded. IV SIMULATION AND EXPERIMENTS RESULTS A. Smulaton Results The proposed mproved reactve power sharng strategy has been verfed n MATLAB/Smuln and expermentally. In the smulatons and experments, a mcrogrd wth two DG systems, as shown n Fg. 1, s employed. The assocated parameters for Power stage and control of the DG unt are lsted n Table I. Also n the smulatons and experments, n order to facltate the observaton of the reactve power sharng, the two DG unts are desgned wth same power ratng and dfferent lne mpedances. The detaled confguraton 5

7 P/w Uo/V Q/Var 6 of the sngle DG unt s depcted n Fg. 6, where an LCL flter s placed between the IGBT brdge output and the DG feeder. The DG lne current and flter capactor voltage are measured to calculate the real and reactve powers. In addton, the commonly used double closed-loop control s employed to trac the reference voltage [5], [7], [12]. V dc Man crcut The proposed Controller SPWM E L f ref L Double loop control E ref C f u c sn ( ref dt) E ref ref L lne u c P Lne Power calculaton Q Improved droop control (Eq.8) Common Bus gradually. After t=1s, the voltage recovery operaton s performed. It can be observed that the output reactve power ncreases but the reactve power sharng performance does not degrade. Fg.8 shows the correspondng output voltages. It can be observed that the output voltages decrease durng the sharng error reducton operaton, whle the voltage recovery operaton ensures that DG output voltage ampltude can restore bac nearby to the rated value. The whole process of adjustment can be done steadly n a relatvely short perod of tme. Fg.9 llustrates actve power sharng performance of the two DG unts. It s obvous that the proposed mproved reactve power sharng strategy does not affect actve power sharng performance Fg. 6 Confguraton of one sngle-phase DG unt. Tab. I Assocated parameters for Power stage and control of the DG unt Parameters Values Parameters Values u rate (V) 220 pu 0.05 L f (mh) 1.5e-3 p 50 r f (Ω) 0.25 K p 0.2 C f (μf) 20 w c (rad/s) 31.4 L Lne1 (mh) 0.6e-3 m(rad/sec w) 5e-5 L Lne2 (mh) 0.3e-3 n(v/var) 5e-3 f s (KHz) 12.8 Ke(v/var) frate (Hz) 50 ΔE 0 (V) 5 T s (s) 1/12.8e3 T syn (mn) (s) 0.1 1) Case 1: power sharng accuracy mprovement Two dentcal DG unts operate n parallel wth the proposed voltage droop control. Fg.7 llustrates the reactve power sharng performance of the two DGs. Before t=0.5s, the sharng error reducton operaton and voltage recovery operaton are dsabled, whch s equvalent to the conventonal droop control beng n effect. There exsts an obvous reactve power sharng error due to the unequal voltage drops on the feeders. After t=0.5s, the reactve power sharng error reducton operaton s performed, t s clear that the reactve power sharng error converges to zero stage1 stage2 stage Fg. 7 Output reactve powers of two nverters wth the mproved droop control T syn Lmted voltage stage1 stage2 stage Fg. 8 Output voltage ampltude of two nverters wth the mproved droop control stage1 stage2 stage Fg. 9 Output actve powers of two nverters wth the mproved droop control. 2) Case 2: Effect of the communcaton delay To test the senstvty of the proposed mproved droop control to the synchronzed sgnal accuracy, a 0.02s delay s ntentonally added to the sgnal

8 P/w Uo/V Uo/V P/W Q/Var Q/Var 7 receved by unt at t=0.5s as shown n Fg.11, and the smulaton results are shown n Fg.10, 11 and 12. Compared to the case 1 n Fg.7 and 9, a small dsturbance appears n both the reactve and actve power, whle the voltage recovery operatons are stll able to ensure that the DG unt can delver the expected reactve power. After t=2.0s, the actve and reactve power sharng errors are almost zero. Therefore, the proposed reactve power sharng strategy s not senstve to the communcaton delay. Then t s llustrated that t s robust to some small communcaton delays. 3.0W and the reactve load decreases about 0.8Var. The correspondng smulaton results are shown n Fg.13 and 14. As can be seen, a large reactve power sharng devaton appears at t=2.5s and t=4.5s. However, the devaton becomes almost zero after a whle. Fg.15 llustrates the correspondng output voltage waveforms. It can be found that there exsts a obvous output voltage decrease and output voltage ncrease process durng each reactve power sharng error reducton process Load change Communcaton delay stage1 stage2 stage Fg. 13 Reactve power sharng performance of the mproved droop control (wth load varyng) Fg. 10 Output reactve powers of the two nverters when 0.02s tme delay occurs n synchronzaton sgnal of unt Load change 0.02s delay Fg. 14 Actve power sharng performance of the mproved droop control (wth load changng) Fg. 11 DG output voltage of the nverters when 0.02s tme delay Load change occurs n synchronzaton sgnal of unt stage1 Communcaton delay stage2 stage Fg. 12 Output actve powers of the two nverters when 0.02s tme delay occurs n synchronzaton sgnal of unt 3) Case 3: Effect of load change In order to test the effect of load change wth the proposed method, the actve load ncreases about 1.6W and the reactve load ncreases about 0.4Var at t=2.5s, and at t=4.5s the actve load decreases about 7 Fg. 15 DG output voltage of the mproved droop control (wth load changng) B. Expermental Results A mcrogrd prototype s bult n lab as shown n Fg.16. The mcrogrd conssts of two mcro-sources based on the sngle-phase nverter. The parameters for output flter are the same as those n smulaton. The load conssts of a resstor of 16Ω and a nductor of 3mH. The sample frequency s 12.8 Hz. A permssble mnmum tme nterval between two

9 8 consecutve synchronzaton events s 0.5s. The permssble mnmum output voltage does not less than the rated voltage by 90%. I 01 (2A/dv) (2A/dv) I 02 (2A/dv) U L (50v/dv) Fg.18 Steady state expermental waveforms wth the mproved droop Fg.16 Prototype of parallel nverters system setup Fg. 17 and Fg.18 shows the measured waveforms wth the conventonal and mproved droop control methods, respectvely. The waveforms from top to down are crcular current ( 0H = ), the output current of nverter 1 ( 01 ), the output current ( 02 ) of nverter 2 and PCC voltage (U L ), respectvely. As can be seen from Fg. 17, there s a qute large phase dfference between two output currents when the conventonal droop control s appled. As a result, the crcular current s pretty hgh and the pea value of crcular current s up to 1.80 A. The man reason for t s the mpedance dfference n DG feeders. Compared wth the crcular current n Fg.17, the crcular current n Fg.18 s very small, whch ndcates that the mproved method s effcent n reducng the crcular current manly caused by the output reactve power dfference between the nverters. control. Fg.19 shows the steady-state output actve and reactve power of each nverter wth the conventonal and the mproved droop control. Fg.19 (a) shows the results wth the conventonal droop. The steady-state output actve powers of the nverters are 31.4 W and 30 W, and the output reactve powers are 21.2 Var and Var. When usng conventonal P-f droop control, no actve power dvergence appear snce frequency s a global varable,.e. same frequency can be measured along the mcrogrd; however, voltage may drop along the mcrogrd power lnes, whch produces the well now reactve power dvergence. Fg. 19(b) shows the results wth the mproved droop. As can be seen, the output actve powers of the nverters are 30.6 W and 31.1 W, and the reactve powers are 3.9 Var and 4.4 Var. These results ndcates that the proposed mproved droop control has no effect on the actve power sharng performance, but maes reactve power be shared precsely P(w) / Q(Var) 20 0 P1 and P2 Q1 Q2 20 P1 and P2 Q1 and Q2 (2A/dv) I 02 I 01 (2A/dv) (2A/dv) U L (50v/dv) Fg.17 Steady state expermental waveforms wth the conventonal droop control t(s) 2 (a) (b) Fg. 19 Steady-state actve power and reactve power a) wth the conventonal droop; b) wth the mproved droop control. To verfy the effectveness of the sharng error reducton operaton and voltage recovery operaton of the proposed method, the experments wth only one operaton beng contnuously used are performed. As can be seen from Fg.20, the crcular current converges to a small value gradually when only the reactve power sharng error reducton operaton s performed. In the meanwhle, a contnuous decrease n 8

10 9 PCC voltage could be found. Fg.21 shows the results when only the voltage recovery operaton s performed. It can be seen that the PCC voltage ncreases lnearly durng ths tme, and the crcular current s always small and be almost ept constant. Fg.22 shows the results when the two operatons are combned..e. the proposed method s appled. The crcular current s controlled to be small value, and the qualty of the PCC voltage s guaranteed successfully. U L (2A/dv) U m U m (25v/dv) Fg. 20 Crculatng current and PCC voltage waveforms of DGs wth only sharng error reducton operaton performed. U L (2A/dv) synchronzaton sgnal, a 0.2 s delay s ntentonally added to the synchronzaton sgnal receved by unt every tme. The assocated expermental results are shown n Fg. 23. Compared to the normal case, there s no obvous dfference between the two cases, and the reactve power sharng error can stll reduce to a small value. Therefore, the proposed method s robust to the communcaton delay because all the necessary operatons only need to be completed n an nterval, not a crtcal pont. Fg.24 shows the expermental results when the synchronzaton sgnal of unt fals, whch s equvalent to the tme delay s nfnty. It s obvous that, before t=t 1, the crculatng current s ept to be a small value because the mproved droop control s n effect. After t=t 1, the sharng error reducton operaton and voltage recovery operaton are dsabled due to the lost of the synchronzaton sgnal of unt. As a result, the pea value of the crculatng current ncreases to about 2.8A from a small value. In concluson, the results n Fg.23 and Fg.24 ndcate that the proposed method only needs a low-bandwdth requrement, and t s robust to a small tme delay of communcaton. However, once communcaton fals completely, the reactve power sharng accuracy performance may be worse. E (25v/dv) t 1 Fg. 21 Crculatng current and PCC voltage waveforms of DGs wth Communcaton delay I 02 (2A/dv) only voltage recovery operaton performed. I 01 (2A/dv) (2A/dv) t 1 (2A/dv) U L Fg.23 Output current and crculatng current waveforms when 0.2 s tme delay occurs n synchronzaton sgnal of unt. (25v/dv) Fg.22 Crculatng current and PCC voltage waveforms of DGs wth the mproved droop. To test the senstvty of the proposed method to 9

11 10 t 1 Communcaton falure t 1 I 02 I 01 (2A/dv) (2A/dv) (2A/dv) Fg.24 Output current and crculatng current waveforms when the synchronzaton sgnal s lost n unt. V CONCLUSIONS In ths paper, a new reactve power control for mprovng the reactve sharng was proposed for power electroncs nterfaced DG unts n AC mcro-grds. The proposed control strategy s realzed through the followng two operatons: sharng error reducton operaton and voltage recovery operaton. The frst operaton changes the voltage bas of the conventonal droop characterstc curve perodcally, whch s actvated by the low-bandwdth synchronzaton sgnals. The second operaton s performed to restore the output voltage to ts rated value. The mproved power sharng can be acheved wth very smple communcatons among DG unts. Furthermore, the plug-and-play feature of each DG unt wll not be affected. Both smulaton and expermental results are provded to verfy the effectveness of the proposed control strategy. REFERENCES [1] W. Deng, W. Pe, and Z. Q, Impact and mprovement of dstrbuted generaton on voltage qualty n mcro-grd, n Thrd Internatonal Conference on Electrc Utlty Deregulaton and Restructurng and Power Technologes, Apr. 2008, pp. [2] Lasseter R H. Mcrogrds [C]//Power Engneerng Socety Wnter Meetng, IEEE. IEEE, 2002, 1: [3] Lopes J A P, Morera C L, Madurera A G. Defnng control strateges for mcrogrds slanded operaton [J]. Power Systems, IEEE Transactons on, 2006, 21(2): [4] Y. W. L, D. M. Vlathgamuwa, and P. C. Loh, Desgn, analyss, and real-tme testng of a controller for multbus mcrogrd system, IEEE Trans. Power Electron., vol. 19, no. 5, pp , Sep [5] Y. A.-R. I.Mohamed and E. F. El-Saadany, Adaptve decentralzed droop controller to preserve power sharng stablty of paralleled nverters ndstrbuted generaton mcrogrds, IEEE Trans. Power Electron., vol. 23,no. 6, pp , Nov [6] Guerrero, Josep M., et al. "Control strategy for flexble mcrogrd based on parallel lne-nteractve UPS systems." Industral Electroncs, IEEE Transactons on 56.3 (2009): [7] Guerrero, Josep M., et al. "Herarchcal control of droop-controlled AC and DC mcrogrds a general approach toward standardzaton." Industral Electroncs, IEEE Transactons on 58.1 (2011): [8] Chandorar M C, Dvan D M, Adapa R. Control of parallel connected nverters n standalone ac supply systems [J]. Industry Applcatons, IEEE Transactons on, 1993, 29(1): [9] Guerrero J M, Matas J, de Vcuña L G, et al. Wreless-control strategy for parallel operaton of dstrbuted-generaton nverters [J]. Industral Electroncs, IEEE Transactons on, 2006, 53(5): [10] De Brabandere K, Bolsens B, Van den Keybus J, et al. A voltage and frequency droop control method for parallel nverters [J]. Power Electroncs, IEEE Transactons on, 2007, 22(4): [11] Guerrero J M, GarcadeVcuna L, Matas J, et al. Output mpedance desgn of parallel-connected UPS nverters wth wreless load-sharng control [J]. Industral Electroncs, IEEE Transactons on, 2005, 52(4): [12] L Y W, Kao C N. An accurate power control strategy for power-electroncs-nterfaced dstrbuted generaton unts operatng n a low-voltage multbus mcrogrd [J]. Power Electroncs, IEEE Transactons on, 2009, 24(12): [13] Guerrero J M, Matas J, de Vcuña L G, et al. Decentralzed control for parallel operaton of dstrbuted generaton nverters usng resstve output mpedance[j]. Industral Electroncs, IEEE Transactons on, 2007, 54(2): [14] Guerrero J M, de Vcuña L G, Matas J, et al. A wreless controller to enhance dynamc performance of parallel nverters n dstrbuted generaton systems[j]. Power Electroncs, IEEE Transactons on, 2004, 19(5): [15] Tuladhar A,Jn H,Unger T,et al.control of parallel nverters n dstrbuted AC power systems wth consderaton of lne mpedance effect [J].IEEE Trans. Industry Electroncs,2000, 36(1): [16] He J, L Y W. An enhanced mcrogrd load demand sharng strategy [J]. Power Electroncs, IEEE Transactons on, 2012, 27(9): [17] Sao C K, Lehn P W. Autonomous load sharng of voltage source converters[j]. Power Delvery, IEEE Transactons on, 2005, 20(2): [18] Moawwad A, Khadar V, Krtley J L. A New P-Q-V Droop Control Method for an Interlne Photovoltac (I-PV) Power

12 11 System[J]. IEEE transactons on power delvery, 2013, 28(2): [19] Coelho E A A,Cortzo P C,Garca P F D.Small-sgnal stablty for parallel connected nverters n stand-alone AC supply systems.ieee Transactons on Industry Applcatons,2002, 38(2): [20] Pogau N,Prodanovc M,Green T C.Modelng, Analyss and Testng of Autonomous Operaton of an Inverter-Based Mcrogrd.IEEE Transactons on Power Electroncs,2007, 22(2): [21] Mahmood, H.; Mchaelson, D.; Jang, J., Accurate Reactve Power Sharng n an Islanded Mcrogrd usng Adaptve Vrtual Impedances, Power Electroncs, IEEE Transactons on, vol.pp, no.99, pp.1,1. [22] Qng-Chang Zhong, Robust Droop Controller for Accurate Proportonal Load Sharng Among Inverters Operated n Parallel, Industral Electroncs, IEEE Transactons on, vol.60, no.4, pp.1281,1290, Aprl [23] C. N. Rowe, T. J. Summers, R. E. Betz, D. J. Cornforth, and T. G. Moore, Arctan power frequency droop for mproved mcrogrd stablty, IEEE Trans. Power Electron., vol. 28, no. 8, pp , Aug [24] A. Luna, F. Blaabjerg, and P. Rodríguez, Control of power converters n AC mcrogrds, IEEE Trans. Power Electron., vol. 27, no. 11, pp , Nov