Aalborg Universitet. Published in: I E E E Transactions on Smart Grid. DOI (link to publication from Publisher): /TSG.2015.

Transcription

1 Aalborg Universitet Review o Power Sharing Control Strategies or Islanding Operation o AC Microgrids Han, Hua; Hou, Xiaochao; Yang, Jian; Wu, Jia; Su, Mei; Guerrero, Josep M. Published in: I Transactions on Smart Grid DOI (link to publication rom Publisher): 0.09/TSG Publication date: 206 Document ersion arly version, also known as pre-print Link to publication rom Aalborg University Citation or published version (APA): Han, H., Hou, X., Yang, J., Wu, J., Su, M., & Guerrero, J. M. (206). Review o Power Sharing Control Strategies or Islanding Operation o AC Microgrids. DOI: 0.09/TSG General rights Copyright and moral rights or the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition o accessing publications that users recognise and abide by the legal requirements associated with these rights.? Users may download and print one copy o any publication rom the public portal or the purpose o private study or research.? You may not urther distribute the material or use it or any proit-making activity or commercial gain? You may reely distribute the URL identiying the publication in the public portal? Take down policy I you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to the work immediately and investigate your claim. Downloaded rom vbn.aau.dk on: juli 4, 208

2 This document downloaded rom is the preprint version o the inal paper: H. Han, X. Hou, J. Wu, M. Su, J. Yang, J. M. Guerrero, "Review o Power Sharing Control Strategies or Islanding Operation o AC Microgrids," I Trans. Smart Grid Review o Power Sharing Control Strategies or Islanding Operation o AC Microgrids Hua Han, Xiaochao Hou, Jia Wu, Mei Su, Jian Yang, Member, I, Josep M. Guerrero, Fellow, I Abstract Microgrid is a new concept or uture energy distribution system that enables renewable energy integration. It generally consists o multiple distributed generators (DGs) that are usually interaced to the grid through power inverters. For the islanding operation o AC microgrids, two important tasks are to share the load demand among multiple parallel connected inverters proportionately and maintain the voltage and requency stabilities. This paper reviews and categorizes various approaches o power sharing control principles. Simultaneously, the control schemes are graphically illustrated. Moreover, various control approaches are compared in terms o their respective advantages and disadvantages. Finally, the paper presents the uture trends. Index Terms AC microgrid; power electronic inverters; power sharing control strategies; islanding operation W I. INTRODUCTION ITH the expansion o the electrical power grid, conventional power system has become increasingly vulnerable to cope with the reliability requirements and the diverse demand o power users. Moreover, Distributed generation (DG) has advantages o pollution reduction, high energy utilization rate, lexible installation location, and low power transmission losses. DG units also present a higher degree o controllability and operability compared to the conventional generators [], which will allow microgrids to play a major and critical role in maintaining the stability o electrical networks [2]-[4]. So, microgrids will gradually be a strong and eective support or the main power grid and potentially one o the uture trends o power system [5]. The DG units o a microgrid can be classiied into grid-orming (voltage-controlled) and grid-ollowing (current controlled) DG units [6]. In grid-connected mode, the units are oten controlled as grid-ollowing. The most adopted control strategies or grid-ollowing inverters are discussed in [4], [7], [8]-[9]. In islanding mode, the electronic converter interaces between the loads and the micro-sources act as voltage sources, which are responsible or the power sharing according to their ratings and availability o power rom their corresponding energy sources or prime movers [0]-[5]. Manuscript received September 4, 204; revised October 6, 204 and March 30, 205; accepted or publication May 04, 205. This work was supported by the National Natural Science Foundation o China under Grants and H. Han, X. Hou, J. Wu, M. Su, and J. Yang (corresponding author) are with the School o Inormation Science and ngineering, Central South University, Changsha 40083, China ( jian.yang@csu.edu.cn). J. M. Guerrero is with the Department o nergy Technology, Aalborg University, Aalborg 9220, Denmark ( joz@et.aau.dk). This paper ocuses on control strategies o grid-orming DG units in islanding mode. Researches on control o grid orming units were perormed initially in uninterruptible power supply (UPS) systems with parallel operation [6]-[2]. Power sharing control strategies o DG units based on communication include concentrated control [22]-[27],master/ slave control [28]-[3], and distributed control [24], [32], [33]. On the other hand, the control strategies without communication are generally based on the droop concept, which include our main categories: conventional and variants o the droop control [6], [34]-[60], virtual ramework structure based method [6], [9], [53], [6]-[68], construct and compensate based methods [69]-[76] and the hybrid droop/signal injection method [36], [77]. The details and characters o various control methods will be illustrated later. Integrated control strategies reer to hierarchical structures which usually consist o primary, secondary and tertiary control [22], [6], [62], [78]. The primary control stabilizes the voltage and requency and oers plug-play capability or DGs. The secondary control, as a centralized controller, compensates or the voltage and requency deviations to enhance the power quality. Tertiary control considers the optimal power lowing o the whole microgrids or interaction with main grid [2]. In addition, Hierarchical control has other special unctions: distributed intelligent management system [79]; voltage unbalance compensation or optimal power quality [80]; sel-healing networks [8]; smart home with a cost-eective energy ecosystem [82]; generation scheduling [83]. So, the hierarchical structure o microgrids can be regarded as an intelligent, integrated and multi-agent system. Some reviews o microgrid control have been published recently [84]-[87]. Reerence [84] classiies all the control strategies (e.g., decentralized control, centralized control, model predictive control, multi-agent systems,) into three levels: primary, secondary and tertiary based on their speed o response and inrastructure requirements. The most excellent is that it proposes the uture challenges and trends in microgrid control. In [85], the next generation power system might adopt the distributed control techniques because o dividing the control task among dierent units. The characters o the distributed control are to use extensive integrated communication and advanced components. New amily o microgrid control and management strategies realizes the plug and play concept and the dynamics o the requency in [86]. Reerence [87] discusses the control methods and objectives rom the point o the voltage and requency stability, and presents the actors aecting power load sharing. This paper ocus on the inverter output control and power sharing control

3 which mostly belong to the primary control, especially or the droop-based control. Mostly, the decentralized control strategies are classed into our main categories more exactly. The rest o the paper is organized as ollows: Section II discusses three control methods based on communication. Section III presents the droop control methods, including the dierent variations. Section I introduces several virtual structures techniques. Section shows some construction and compensation methods. Hybrid droop/signal injection based methods are reviewed in Section I. Then, characters o various methods and the uture trends are summarized in Section II. Finally, this paper concludes in Section III. II. COMMUNICATION BASD CONTROL TCHNIUS Communication based control techniques can achieve excellent voltage regulation and proper power sharing. Moreover, in contrast to droop controllers, which will be discussed later, the output voltage amplitude and requency are generally close to their ratings without using a secondary control [22]. However, these control strategies, which require communication lines between the modules, result in increased cost o the system. Long distance communication lines will be easier to get interered, thus reducing system reliability and expandability. In the ollowing Section, several typical communication based control strategies are reviewed. A. Concentrated Control The concentrated/central control method is presented in [23]-[27], [88], and the control scheme is illustrated in Fig.. The control method requires common synchronization signals and current sharing modules. The phase locked loop (PLL) circuit o each module can ensure the consistency between the requency and phase o the output voltage and the synchronization signal. Also, the current sharing modules can detect the total load, which deine the reerence value o the current or each module. This reerence current i re is a raction o the load current i load. For N equal modules, i re = i load N. In the meantime, every inverter unit measures itsel output current in order to calculate the current error. In case o parallel units controlled by synchronization signals, they have negligible dierences o requency and phase among each other, thus the current sharing error o each unit can be caused by voltage amplitude inaccuracies. Thereore, this method directly adds current error to each inverter unit as a compensation component o the voltage reerence in order to eliminate the dierences among their output currents. In [26] and [27], the central limit control (CLC) scheme is discussed. In CLC mode, all the modules should have the same coniguration and each module tracks the average current to achieve equal current distribution. Reerence [88] proposes a multistage centralized control scheme with high penetration o plug-in electric vehicles. The coordination allows the Ps to play a pivotal role in the successul and optimized operation o the islanded microgrids. The one advantage o the concentrated method is that current sharing is maintained during both steady-state and transients. However, this control scheme must include a centralized controller, which makes diicult to expand the system and reduces system redundancies. Moreover, current reerence has to be distributed to all converters by using high bandwidth communication links, in order to achieve synchronization among the units. These techniques present high dependency on communications and reduce the reliability, which may be compromised with single-point aults. R R R controller Central control units DG unit controller controller PLL syn pulse PLL PLL Power electric inverter Power electric inverter Power electric inverter Fig.. Control schematic o the concentrated control. 0 i 2 i n i i g i g i g i i 2 i n i N i N i N Common Bus Load AC Bus B. Master/Slave Control Based on the master/slave control method, the unction o parallel control units is built into each inverter. Through the mode-selecting switch or automatic sotware setting, the initially starting module in parallel acts as master inverter, which is in charge o parallel control, while the others serve as slave-inverters [28]-[3],[89]. The structure o master/slave control is illustrated in Fig. 2. As shown in this igure, the master module regulates the output voltage and speciies the current reerence o the rest o slave modules. Then, slave units track the current reerence provided by the master in order to achieve equal current distribution. Inverters don t need any phase locked loop or synchronization since these units are communicated with the master units. However, the system isn t redundant since it presents a single point o ailure. I master unit ails, the whole system will ail. In order to overcome this drawback, several researchers have improved the master/slave control method. In [30], the rotating priority window, providing random selection o the master, is proposed to increase the reliability. An auto master-slave control strategy is proposed in [3], which is a variant o the master/slave control. The control circuitry contains an active power share bus and a reactive power share communication bus interconnecting all the paralleled units. The inverter with the highest output power becomes the master inverter, which drives the power bus. Also, its power is the reerence or the other inverters. The master-slave control in [89] regards the utility interace as master control at the common coupling point with the utility and the energy gateways, allows plug-and-play integration o DRs and ensures eicient and reliable operation o the microgrid in every operating condition. 2

4 v re Master Units Slave Units i m i Controller Current Controller v e PWM Converter PWM Converter L C L C Common Bus Load control scheme or requency restoration and economic dispatch. The most advantage is that the DGs can share loads according to their increment costs. A robust distributed controller is designed or sharing active and reactive power in [93]. It use partial eedback linearization and ensure the robustness by considering structured uncertainties. The concept o the graph theory is also adopted. i m i n Current Controller PWM Converter L C i ave v re i e v 0 Decompose controller i eq regulator PWM v 0 i 0 Fig. 2. Control structure o the master/slave control. In summary, master/slave control can achieve excellent power sharing perormance with advantages o ease implementation. I the master inverter ails, the improved control strategy would switch to another normal inverter which is then used as the new master. Thereore, parallel operation wouldn t be aected. However, an obvious issue with all master/slave control methods is that high output current overshoot may occur during transients since the master output current isn t controlled, so it doesn t ensure a good transient perormance. C. Distributed Control The distributed control is oten applied to parallel converters [24], [32]-[33], [90]-[93]. The instantaneous average current sharing is a typically distributed control or parallel converters. In this control technique, individual control circuit is used in each inverter, but no central controller is needed. Further, average current sharing requires a current sharing bus and reerence synchronization or the voltage. An additional current control loop is used to enorce each converter to track the same average reerence current, provided by the current sharing bus. When a deect happen in any module, it can smoothly detach rom the microgrid, and the rest o modules can still operate normally in parallel. Fig. 3 shows a control block diagram o the distributed control scheme. The average current sharing bus value is regarded as a current reerence o each paralleled converter. The current error i en is decomposed into active and reactive components, i end and i enq, then the output voltage requency and amplitude are regulated through current regulators respectively. The distinct eature o the distributed control is that the inormation required is not global but adjacent or any units. So it only needs lower band-width than the central control method. Because o dividing the control task among dierent units, it has many advantages compared to the droop control. Recently, it has become the lexible and reliable control strategies o uture trends. In [90], a distributed networked control system is used to restore the requency and amplitude deviations and ensure reactive power sharing. Without relying on a central control, the ailure o a single unit won t inluence the normal operation o the whole system. Reerence [9] provides a distributed two-layer control scheme or ac microgrids. The voltage/requency and active/reactive powers are decoupled, regulated by the irst and second layers respectively. Reerence [92] utilizes the ully distributed 3 i 0 i0 N k = Current sharing Bus i ave i en i 0n v re N k Decompose v 0n i enq controller i ed i end regulator regulator regulator Fig. 3. Control structure o the distributed control. PWM Load Common Bus In conclusion, the distributed control has no central control board and every module is symmetric. regulation and undamental power sharing are well controlled. However, interconnections between the inverters are still necessary. This degrades the lexibility and redundancy o the system. As the number o parallel modules and distance o the interconnected lines increase, more intererence is expected in the system. III.DROOP CHARACTRISTIC BASD TCHNIUS The control strategies that operate without inter-unit communications or power sharing control are based on droop concept [2]-[3], [], [6]-[2], [94]-[96]. Operation without communication links is oten essential to connect remote inverters. It can avoid complexity and high costs, and improve redundancy and reliability requirements o a supervisory system. Also, such a system is easier to expand because o the plug-and-play eature o the modules which allows replacing one unit without stopping the whole system. Thereore, communication lines are oten avoided especially or long distances and high investment cost. However, droop characteristic presents several drawbacks: Frequency and voltage deviations: In islanding mode, the voltage and requency o the microgrid are load-dependent. Steeper droop ensures better load sharing, yet results in larger requency and voltage deviations, and even may cause instabilities in the microgrid. This is the inherent trade-o between the requency and voltage regulation and load sharing accuracy or the droop method [34]. Harmonic loads: The original droop method ocuses on undamental power sharing but doesn t take harmonic sharing into account in the case o nonlinear loads. I it s not coped properly, it would lead to harmonic circulating currents and poor power quality. Moreover, the calculation and smoothing o active and reactive power take some v 0n i 0n

5 delays, thus it presents a slow dynamic response [35]-[36]. The dierent and unknown line impedances: The line impedances between the paralleled converters also aect the power sharing perormance. When the line impedances between the inverters and point o common coupling are dierent, it could result in a large circulating current and low precision o power sharing among inverters [37], [38]. Fluctuant and changeable output power o DGs: Another drawback o the original droop method is the poor perormance with renewable energy resources because the output active power o micro-sources is usually luctuant and changeable [39]. To overcome these drawbacks and minimize the circulating current under all situations, researchers have developed several improved droop control methods. These methods all into our main categories, namely ) Conventional and variants o droop control 2) irtual structure based methods 3) Construct and compensate based methods 4) Hybrid droop/signal-injection based methods. Besides the above our main categories, the most recent control methods based on droop characteristic has emerged rom [94]-[96]. In [94], drooping the virtual lux instead o the inverter output voltage can avoid the complicated inner multi-loop eedback control and requency-voltage deviations to some extent. Reerence [95] presents a multivariable control topology, which oers a systematic and straight orward design approach with the loop shaping technique. It also realizes real power sharing by simultaneous drooping o both requency and voltage amplitude, which enhances load sharing accuracy in resistive microgrids. The consensus-based droop control with sparse communication network obviously alleviates the eects o non-ideal line impedances with better dynamical perormances in [96]. This Section will discuss conventional and improved droop controllers, and show the control schemes in detail. A. Conventional Droop Control The droop control method or the parallel connected inverters can avoid the dependency on communications. It is sometimes named as wireless control with no interconnection between the inverters. However it can lead to a conusion with the wireless communications in the sense o radiorequency based communications. The basic idea o this control level (also named primary control) is to mimic the behavior o a synchronous generator, which is to reduce the requency as the active power increases. When the inverter output impedance is highly inductive, hence the active and reactive powers drawn to the bus can be expressed as i Pi = sinφ X () 2 i cosφ i = X Where X is the output reactance o an inverter, ϕ is the phase angle between the output voltage o the inverter and the voltage o the common bus, i and are the amplitude o the output voltage o the inverter and the grid voltage, respectively. It can be ound that the active power is predominately dependent on the power angle, while the reactive power mostly depends on the output voltage amplitude. This principle can be integrated in voltage source inverters (SIs) by using the well-known P/ droop method [6], [40], which can be expressed as i = rated mp ( Pi Prated ) (2) i = rated n ( i rated ) where i is the index representing each converter, rated and rated are the nominal requency and voltage o the micro-source, respectively, P i and i are the average active and reactive power, P rated and rated are the nominal active and reactive power, respectively; and m P and n are the active and reactive droop slopes, respectively. The choice o m P and n impacts the network stability, so they must be careully and appropriately designed [4], [42]. Usually, the droops are coordinated to make each DG system supply apparent power proportional to its capacity [6]. i min mp = Pi Pi, max (3) i, max i,min n = i, min i, max The control algorithm with conventional droop control is illustrated in Fig. 4. The power stage consists o SI with a LC ilter and a coupling line inductor. The controller consist o three control loops: (i) a power sharing controller is used to generate the magnitude and requency o the undamental output voltage o the inverter according to the droop characteristic; (ii) a voltage controller is used to synthesize the reerence ilter inductor current vector; (iii) and a current controller is adopted to generate the command voltage by a pulse width modulation (PWM) module. Micro source Local sample v 0 i 0-90 Data regulation v 0 v 0 i 0 PI Low-pass ilter Power Calculation Low-pass ilter SPWM Modulation P n m P v 0 Droop control LC ilter i 0 Inner control loops (voltage & current) i 0 The Main circuit line Reerence 2sin( t) AC Bus v 0 v re Local control without communication Fig. 4. Control structure o conventional P/ and / droop control. As discussed above, the conventional droop method can be implemented without communication between modules, and thereore is more reliable. However, it has some drawbacks as listed below: Multiple control objectives: Since there is only one control DSP 4

6 variable or each droop characteristic, it isn t possible to satisy multiple control objectives. For example, a design trade-o needs to be considered between the voltage / regulations and load P/ sharing [34]. Mixed resistive and inductive line impedance: The conventional droop method is developed assuming highly inductive equivalent impedance between the SC and the AC bus. However, this assumption is challenged in microgrid applications since low-voltage distribution lines are mainly resistive. Thereore, equation () isn t valid or AC microgrids [34], [38]. Furthermore, i the line impedance is mixed resistive and inductive, then the active and reactive power will be strongly coupled. This case is important in medium-voltage (M) microgrids, in which the power lines X/R ratio can be next to one. No a global voltage variable: As opposed to the requency, the voltage isn t a global variable in a microgrid. Thus, the reactive power control is diicult to share between the parallel inverters and may result in circulating reactive current [37], [38]. Same problem may occur in highly resistive lines, especially or circulation o active current controlled through the voltage. The nonlinear loads: In case o nonlinear loads, the conventional droop method is only based on undamental values and doesn t consider current or voltage harmonics. Since it only uses P and measurements which are usually average over one line cycle. The conventional droop method should be modiied in order to share the harmonic currents [35], [36]. These potential drawbacks have been widely discussed in the recent literatures [34]-[38]. Proposed solutions will be discussed in ollowing sections. B. PD/FB Droop Control While the conventional requency droop control method works well in a microgrid with mainly inductive line impedances, it may present problems when implemented it on a low-voltage microgrid, where the eeder impedance is mainly resistive. Note that the delivered active and reactive power o the inverter still increase with, but here, the reactive power increases with the power angle ϕ, and the active power remains increasing along with voltage variation ( com ), as can be seen in the ollowing well known small-angle approximation. 2 com com P Z (4) com φ Z Thus, voltage active power droop and requency reactive power boost (PD/FB) characteristics are alternatively considered [9], [46]-[48], as i = rated + m i (5) i = rated np Pi Droop/boost characteristics o PD/FB method are shown in Fig. 5. This kind o control oers an improved perormance or controlling low voltage AC microgrid with highly resistive transmission lines [46]. However, the PD/FB method strongly depends on system parameters which signiicantly restrict its application. Furthermore, it is also unable to properly share the load active current. 0 P P max min 0 ( a ) (b) Fig. 5. Droop/boost characteristics or low-voltage AC microgrid. max C. Complex Line Impedance Based Droop Method Many problems cannot be solved by using the conventional droop control method, such as line impedance dependency, inaccurate P or regulation and slow transient response [34]-[38]. In [53], considering the impact o complex impedance, it proposes the controller that can simpliy the coupled active and reactive power relationships, oer good dynamic perormance, and be more convenient when the line impedance resistance and inductance parts are similar (X R) in M microgrids. In this particular case, the droop unctions can be expressed as = 0 mp ( P ) (6) = 0 n ( P+ ) In [52], to acilitate simultaneous active and reactive powers control and regulate the PCC voltage,a P-- droop control method is proposed. For electric systems with complex impedance, both active and reactive powers aect the voltage magnitude. Thereore, the droop characteristics or the proposed P-- droop method is given by = re + ( nd P) + ( md ) (7) where re is the desired reerence value o the PCC voltage, in this case, p.u.; n d and m d are the active and reactive power coeicients or the proposed P-- droop method. Moreover, these droop coeicients are adjusted online through a lookup table based on the PCC voltage level. The control algorithm o the proposed P-- droop method is shown in Fig. 6. Furthermore, additional loops such as impedance voltage drops estimator [38], grid parameters estimator [54], and reactive current loop [55] have been added to the conventional droop control in order to deal with line impedance mismatches and ensure good power sharing perormance. In order to improve the dynamic perormances o parallel inverters in distributed generation systems, a wireless (droop) controller is proposed in [56]. t dp φ = m Pdτ m P P md dt (8) d = n nd dt 5

7 Loop d, re q, re Inner Current Loop I d, re I q, re I d, meas I q, meas sint cost abc φ PID PID PLL re re PWM+P inverter n d m d meas Lookup table n P meas n d d m Outer current loop Droop controller d m d re RMS Iabc θ abc D transormation & power calculation I I q, meas d, meas Fig. 6. The control algorithm with the proposed P-- droop method. In a relatively small AC microgrid, large load changes can be expected. Then an adaptive derivative term is used to add damping and to avoid large start-up transients and circulating currents [57], [58], as dpi = + KP ( Pi Pi, re ) + Kpd dt (9) di i = re + K ( i i, re ) + Kqd dt where the choice o K pd and K qd can be obtained through the pole placement method. Additionally, there are some other solutions to address dynamic response problems: droop based on coupling ilter parameters [59], droop based on H derived rom linear matrix inequality (LMI) control theory [60]. D. Angle Droop Control In urther investigation o the droop concept, some researchers have proposed power-angle droop control, in which the phase angle o the distributed source voltage, relative to a system-wide common timing reerence is set [43]-[45]. As a result, the power requirement can be distributed among DGs, similarly as conventional droop does, by dropping the voltage angle and magnitude. δi = δrated mp ( Pi Pi, rated ) (0) i = rated n ( i i, rated ) Where rated and δδ rated are the rated voltage magnitude and angle o the DG respectively, when supplying their rated power levels o P i,rated and i,rated. Coeicients m P and n indicate active and reactive power droop gains. These values are chosen to meet voltage regulation requirement in the microgrid. The coeicient values or dierent DGs are chosen in order to share the load in proportion to their ratings. The angle droop is able to provide proper load sharing among the DGs without a signiicant steady-state requency drop in the system. And it has advantageous as the requency maximum restricts the choice o droop gain in the conventional requency droop control. Moreover, no communication is needed between DGs. However, i the local control boards aren t synchronized each other, the imperection o the crystal clock o the digital processors makes requencies o each inverter slightly dierent, which will lead to running out o phase limits ater certain time, leading to system instability. Some authors suggest the CAN bus or even global positioning system (GPS) to synchronize DGs. However, the loss o the global synchronizing signal at some DG units should be urther investigated.. Based Droop Control This control method is another type o P/ control. The control strategy presents a constant power band control o islanding AC microgrid, which operates without inter-unit communication in a ully distributed manner and takes the speciic characteristics o the microgrid into account. These characteristics include the lack o rotating inertia, resistive line, and high share o DGs, which are less controllable than central generators and require optimal power exploitation [49]-[5]. The voltage-based droop (BD) control strategy [49] consists o a P/ droop controller which is divided into two droop controllers ( g / dc and P/ g droops) and constant-power bands, as illustrated in Fig. 7. g/dc-droop controller P/g-droop controller 6 dc g g, nom dc, nom dc /-droop controller nom nom g Fig. 7. Droop control with constant power band. P dc P dc, nom b g, nom g, nom controller First, the g / dc droop control principle is based on the speciic characteristics o islanding AC microgrid. I an unbalance occurs between the generated power and the absorbed power, the dc link voltage dc o the power source changes. Thereore, dc is the indicator or ac power change. g = g, nom + m ( dc dc, nom ) () where g,nom and dc,nom are the nominal voltage o AC and DC side o power converter. Note that a slightly change o g leads to a change o the power delivered to the electrical network. To limit the signiicant deviation o AC side voltage, P dc / g droop with constant power band is used [49], as Pdc, nom KP ( g ( + b) g, nom ), i g > ( + b) g, nom Pdc, nom, Pdc = (2) i ( b ) g, nom < g < ( + b ) g, nom Pdc, nom KP ( g ( b) g, nom ), i < ( b) g, nom where P dc,nom is the rated active power o the AC-side o power g P dc SI

8 converter, and K p is the power droop gain. Note that the width b o this band is dependent on the nature o the source. g / dc droop control can be used along with P dc / g control in AC microgrids in order to take the advantages o both control methods. With the g / dc droop control, the microgrid voltage can be changed by detecting changes o dc, and balance is achieved without the need to change P dc. In the meantime, requent power changes can be avoided. No communication or the primary control is required, and the tolerated voltage deviation rom its nominal value is eectively used or the control. The overall scheme o the droop control with constant power band is shown in Fig. 8. Rest o microgrid i g i g Sensors g g calculation LC-ilter dc C L g/dc-droop controller dc, nom nom Kv Kq g dc g, nom g, nom g DC-side sin(2 π t) Rv i g P/g-droop controller 2b Constant power band I dc, nom PWM module g K P P dc P dc, nom controller Resistive virtual /-droop controller output impedance Fig. 8. The overall scheme o droop control with constant power band. In summary, voltage based control strategy makes ull utilization o the allowable range o the output voltage. In this range, the renewable energy sources are actively dispatched as they operate at maximum power tracking point (MPPT). This is particularly advantageous or DGs since their energy can be used more eiciently. Additionally, by combining the P/ g droop control, P dc can be changed in case the constant power band is surpassed, which increases the power lexibility in AC microgrid and avoids the voltage-limit violation. However, this control requires the micro-source to have certain ability to dispatch energy easily. Thereore, DGs require the multi-stage controller to dispatch the energy, which may aect the system eiciency to some extent. I. IRTUAL-STRUCTUR-BASD MTHODS A. irtual Output Impedance Loop In order to avoid the active and reactive power coupling, a typical and popular approach is based on virtual output impedance method [6], [9], [53], [6]-[62]. This control method is implemented by including ast control loops in the droop control method, as shown in Fig. 9. As a result, the expected voltage can be modiied [9], as re = Z () D s i0 (3) where Z D (s) is the virtual output impedance, and is the output voltage reerence under no load condition. In general, the output inductance can be produced by emulating an inductive behavior. This can be achieved by drooping the output voltage proportionally to the derivative o the output current with respect to the time, so that Z D (s) is purely inductive, i.e. Z D (s) = sl D. However, dierentiation can ampliy high requency noise, which may destabilize the DG voltage control scheme, especially during transients. This issue can be overcome by using a low-pass ilter instead o a pure derivative term o the output current [9]. s re = LD i0 (4) s + 7 re irtual impedance loop Sin generator sin( wt) re loop Z () s 0 L v Current loop PD(s) PD(s) Sot-start P c PWM+ inverter P P & Calculation + LPF Fig. 9. The control scheme o droop control with virtual output impedance. I the virtual impedance Z D (s) is properly adjusted, it can prevent occurrence o current spikes when the DG is initially connected to the AC microgrid. The sot starting can be acilitated by considering the time-variant virtual output impedance, as tt Z () t = Z ( Z Z ) e (5) D i where Z i and Z are the initial and inal values o the virtual output impedance, respectively, and T is the time constant o the sot starting process. Also, i the output inductance can be produced by emulating a resistive behavior, the system stability can be improved [64]. Recently, the virtual output impedance method has been modiied or harmonic current sharing [63], which is introduced in ollowing subsection. B. nhanced irtual Impedance Loop The islanding AC microgrid may have serious power quality problems due to the increasing presence o nonlinear loads. To realize a better reactive and harmonic power sharing, the research [63] proposes an enhanced control method using virtual impedance at the undamental and selected harmonic requencies. Similar in virtual undamental output impedance, this enhanced control method introduces the harmonic virtual impedance. The overall scheme o droop control with enhanced virtual impedance is shown in Fig. 0. I re Ki K Iinv _ α K Pi Pi Iinv _ α Iinv _ α _ 2K s + s i c cs+ 2K s + s i c cs+ C α I Line _ α C _ α w sin( wt) droop _ α v 0 vh C _ α I Line _ α Feeder irtual undamental impedance irtual harmonic impedance PCC Droop control I Line _ α I Line _ β PCCh _ α PCCh _ β droop _ β P ave ave Fig. 0. The scheme o enhanced virtual impedance method. v P & Calculation Local PLL i dq / αβ droop _ α I Line _ α I Line _ β θ C droop _ α droop _ β PCCh _ d PCCh _ q This enhanced virtual impedance control method can realize

9 better reactive and harmonic power sharing, alleviate the computational load at DG unit local controller without using any undamental and harmonic components extractions, and mitigate the PCC harmonic voltages by reducing the magnitude o DG unit equivalent harmonic impedance. However, it requires the knowledge o the physical line impedance parameters, and low-bandwidth communications. Additionally, virtual impedance design rules are presented in [64], and a robust virtual impedance implementation is proposed, which can alleviate voltage distortion problems caused by harmonic loads. C. irtual Frame Transormation Method Another method based on a virtual structure is the virtual rame transormation [6], [65]. In general, both line reactance X and resistance R need to be considered. The active and reactive powers drawn to the bus can be expressed as Pi [( i )cos θ+ iφsin θ] Z (6) i [( i )sinθ iφcos θ] Z Where ϕ is the phase angle between the output voltage o the inverter and the common bus, and are the amplitude o the output voltage o the inverter and the grid voltage, Z and θ are the magnitude and phase o the impedance respectively. The use o an orthogonal linear rotational transormation matrix T rom active and reactive power P and to the modiied active and reactive power P ' and ' is proposed as P' P sinθ cosθ P TP ' = = cosθ sinθ (7) Despite the line impedance is mixed, P/ decoupling is achieved as i the network were purely inductive. In general, the accurate value R/X isn t known, but an estimation o R/X may be suicient to perorm the method [54]. Similarly to [6] and [54], a virtual requency/voltage rame transormation ( ' ' ) is proposed in [66]-[68]. ' sin cos T ϕ ϕ ' = = cosϕ sinϕ (8) where and are calculated through the conventional droop equations in (2). The transormed voltage and requency, ' - ', are then used as reerence values or the DG voltage control loop. The PD/FB method and the conventional droop control are special cases whereϕ = 0 andϕ = 90. The ' ' virtual rame transormation is shown in Fig.. The proposed real and reactive power control is based on the virtual requency and voltage ' ' rame, which can eectively decouple real and reactive power lows and improve the system transient and stability perormance. However, one issue with the virtual rame power control is that i the rame transormation angle isn t the same or all DG units, the microgrid requency and voltage will be converted to dierent values in dierent virtual rames. Consequently, i two DGs are injecting dierent powers or line impedances aren t matched, the transormation angle will be dierent and both reerence rames will be out o synchronism. 8 max min ' ϕ ' ' min ' min ' ' c ' d min Fig.. The details o the ' ' virtual rame transormation. b ' a max. CONSTRUCTION-AND-COMPNSATION-BASD MTHODS A. Adaptive Droop Control Recently, some researchers have proposed control methods based on construction and compensation ideas. In [69], it proposes a novel adaptive voltage droop scheme or the parallel operation o DGs in an islanding AC microgrid. In this method, two terms are constructed to the conventional reactive power (-) control. One term is used to compensate or the voltage drop across the transmission lines. The other term is added to hold the system stability and improve reactive power sharing under heavy loading conditions. In order to illustrate this control technique, a two-dg system with generic output impedances is shown in Fig. 2. The voltage o a single DG can be derived as rp i i x i i i = i Dii (9) i i The two latter terms represent the voltage drop on the internal line impedance. These terms can be added to the conventional reactive power control, which compensates or voltage drops on the power lines as rp i i x i i i = i Dii + ( + ) (20) i i Additionally, to improve the system stability and suit or any load conditions, the method presented in [69] adopts the voltage droop coeicient as a nonlinear unction o active and reactive power. rp i i x i i i = i Di( P i, i) i + ( + ) i i (2) 2 2 Di ( P i, i ) = Di + mii + mpipi where D i, m i and m Pi are droop coeicients. The three terms can mitigate the negative impacts o the active power control and the microgrid parameters on the reactive power control, improving the system stability and the reactive power sharing under heavy loading conditions. Nevertheless this method requires good knowledge o the power line parameters [69].

10 Small errors may result in a positive eedback, and thus may cause system instability. P, α L 0 2 α P, Z Line Line2 Z2 = r2 + jx = r+ jx 2 DS R DS2 l+ jx l Rl2 + jx l2, δ, 2, δ2, 2 Load Fig. 2. A typical two-dgs system. B. Synchronized Reactive Power Compensation Method To improve the reactive power sharing accuracy, an enhanced control strategy is proposed in [70]-[7], which estimates the reactive power control error by injecting a small real power disturbance that is activated by low-bandwidth synchronization signals rom the central controller. Also, a slow integration term is added to the conventional reactive power droop control in order to eliminate reactive power sharing error. With the proposed scheme, reactive power sharing errors are signiicantly reduced. Ater the compensation, the proposed droop controller will be automatically switched back to the conventional droop controller. The improved droop control can be described as = 0 ( DP P+ D ) K (22) C = 0 D + ( ) ( P PA ) s where K C is the integral gain, which is selected to be the same or all the DG units. Fig. 3 illustrates the diagram o the proposed synchronized reactive power compensation method. This control strategy is realized by two stages [7]. The conventional droop method is used in the irst stage, and the averaged real power in the steady-state should be measured or use in the second stage. In the last stage, the reactive power sharing error is compensated by introducing a real-reactive power coupling and using an integral voltage magnitude term. In summary, the synchronized reactive power compensated method injects a real-reactive power transient coupling term to identiy the errors o reactive power sharing, and improves the reactive power sharing accuracy [7]. However, the method needs synchronization signals rom a central controller. It can be seen as a classical event-triggered system whose stability isn t easy to be guaranteed. P 0 0 LPF LPF Compensation Flag Moving P Average Filter P Sot compensation gain P A S/H G G DP P+ D D K C s sin( wt) Fig. 3. Droop control with synchronized reactive power compensation. re To voltage tracking loop C. Droop Control Based Synchronized Operation The method mainly includes two important operations: error reduction operation and voltage recovery operation [00]. The sharing accuracy is improved by the sharing error reduction operation, which is activated by the low-bandwidth synchronization signals. However, the error reduction operation will result in a decrease in output voltage amplitude. Thereore, the voltage recovery operation is proposed to compensate the decrease. The needed communication in this method is very simple, and the plug-and-play is reserved. The improved droop control can be described as 9 k k n n i() t = n i i() t K i i + G n= n= (23) where k denotes the times o synchronization event until time t. According to (23), the control is a hybrid system with continuous and discrete traits. Thereore, the droop equation at the k-th synchronization interval could be expressed as k k k k n n i i i i i n= n= (24) = n K + G where G n is the voltage recovery operation signal at the n-th synchronization interval, G n has two possible values: or 0. I G n =, it means the voltage recovery operation is perormed. i n represents the output reactive power o DG-i unit at the n-th synchronization interval. K i is a compensation coeicient or the DG-i unit, Δ is a constant value or voltage recovery [00]. Besides, the control timing diagram is shown in Fig. 4. The sharing error operation and the voltage recovery operation are perormed in update interval. Sampling operation occurs in sampling interval. There is a time interval τ, which is long enough to guarantee the system in steady state. The method is robust to the time delay because all the necessary operations only need to be completed in an interval, not a critical point. k--th synchronization Update interval τ k k i i Sampling interval k-th synchronization tw Update interval τ k+-th synchronization k i Sampling interval Fig. 4. Control timing diagram o one DG with the two consecutive synchronization events. D. - Dot Droop Control Method This method constructs the relationships o reactive power and the change rate o the DG output voltage ( ) in order to improve the reactive power sharing [72]-[74]. The proposed droop control can avoid this coupling dependence. The change rate o voltage will drive continuously until the desired lows, and its perormance can be less dependent on the line impedances. The droop controller is expressed as x = 0x nx ( 0x x) (25) x = 0 x + dτ t x where n x is the droop coeicient, 0x is the nominal x which is set to zero, and 0x is the reactive power set point at the nominal x, which is related to the reactive power capacity o DG. Also, 0x is the nominal phase voltage magnitude and k i t w t x

11 is the voltage command. In the steady state, the x must be reset back to zero to prevent varying output voltage magnitudes. So, x restoration mechanism is designed [73] as d 0x = K res Rx ( 0x x ) (26) dt The control diagram o proposed droop control and the DG control block diagram are shown in Fig. 5. P& Calculation 0x i x x x P x, inst x, inst x res LPF K Rx Restoration P x x s P- Droop - dot Droop 0x x w sin( wt) Droop controller x nx x x 0x x s x x & Current controller x, re PWM+ power converter 0 ix x Droop controller Fig. 5. droop controller and the control block diagram o single DG. In the proposed control strategy, the control result is related to the initial condition o the voltage change rate. Despite the system is stability, the steady-state solutions may not exist. Moreover, the power sharing perormances aren t necessarily superior to those o conventional methods. The use o the integral term in (25), tries to restore the voltage with a local control loop, whose response will depend on the initial conditions o such an integrator, thus leading to system instability. Thereore, this controller isn t easible in real microgrid applications.. Common ariable Based Control Method The common variable is critical or the active and reactive power sharing. Because o the mismatch between the DG output interace inductors in microgrid, it is really diicult to achieve reactive power sharing. Similar with the active power control, some researchers have proposed the adjustable reactive power sharing method, where an integral controller is used to regulate the common bus voltage com [48], [75], [76]. i = Kq ( re com ) dt (27) where K q is the integral gain and re = D i (28) In the steady state, com and re o each DG are equal. Moreover, the steady-state reactive power can be calculated as com = (29) D From (29), it is known that the reactive power or each DG is equal. Then, microgrid operation parameters will no longer aect the reactive power control. Similarly, the strategy proposed in [76] is suited or inverters with resistive output impedance. The improved active power control is modiied i = Ke ( com ) Kq P i dt (30) In summary, the control method based on a common variable can achieve accurate proportional load sharing among parallel DGs, and is robust to the system parameter variations. However, these methods have a potential issue o requiring the load voltage inormation which is diicult to measure when it exists long distances between the DG and the common bus. Moreover, the common voltage may not exist when the coniguration o AC microgrid is complex or in a real distributed system with dispersed loads. I. HYBRID DROOP/SIGNAL-INJCTION BASD MTHOD Conventional droop control cannot ensure a constant voltage and requency, neither an exact power sharing. But an advantage o the control can avoid communication among the DGs. Communication based control is a simple and stable strategy providing a good current sharing, yet a low reliability and redundancy. Thereore, to take advantage o their respective advantages, a hybrid scheme combining two control methods is presented in [47], [97]-[99]. The sharing o real and reactive powers between the DGs is easily implemented by two independent control variables: power angle and voltage amplitude. However, adding external communication is still not desired. Such communications increase the complexity and reduce the reliability, since power balance and system stability rely on these signals. Several current sharing techniques based on requency encoding o the current sharing inormation have been presented in [36] and [77]. The power lines are used or the communication or the power sharing. Most importantly, this technique doesn t require extra control interconnections and automatically compensates or inverter parameter variations and line impedance imbalances. In [36], each DG injects a small AC voltage signal to the microgrid. Frequency signal w q is determined by the reactive power o the DG. q = q0 + D (3) where w q0 is the nominal requency o injected AC signals and D is the boost coeicient. The small real power transmitted through the signal injection is then calculated. And the value o the output voltage,, is adjusted [36], as = DP pq (32) In this way, a / droop is achieved, through the requency component w q. In the presence o nonlinear loads, the harmonic distortion D caused by non-linear loads is shared in similar way. A control signal with a requency that is drooped with D is injected. The power in this injected control signal is used to adjust the bandwidth o the voltage loop [36]. d = d0 md D = S P (33) BW = BW0 Dbw pd where BW 0 is the nominal bandwidth o the voltage loop and D bw is the droop coeicient. The block diagram o the signal injection method is shown in Fig.6. 0

12 I P Micro Source r le tro c o n reerence calculator SC PWM PI controller = DP pq BW = BW0 Dbw pd Droop characteristics q d re BW P D i 0 Power calculator Fig. 6. The block diagram o the requency signal injection method. AC Bus Signal injection method properly controls the reactive power sharing and isn t sensitive to variations in the line impedances [36], [77]. It is also suited or linear and nonlinear loads. However, it doesn t guarantee the voltage regulation. Other issues o this method are the complexity and the need or measuring and generating high-requency components. Also, signal injection method can deteriorate the power quality, which increases the losses on the transmission lines because o the harmonic current generated by the method. Moreover, this injected signal can result in the inter harmonic and resonance. Since this method adjusts the voltage droop bandwidth, it may attenuate the system stability. As an alternative, harmonic virtual impedance is proposed in [63]. II. DISCUSSION OF ARIOUS MTHODS AND FUTUR TRNDS From the previous discussion, it can be seen that each o these proposed control techniques has its own characteristics, advantages and disadvantages. The communication based methods can provide tight current sharing, high power quality, ast transient response, and reduce circulating currents between the inverters. However, it requires communication links and high bandwidth control loops. Further, it isn t easy to be expanded due to the need or load current measurement and to know the number o inverters in the system. The required interconnections make the system less reliable and not truly redundant and distributed. Droop control methods are based on local measurements o the network state variables which make DG truly distributed and absolute redundancy, as they don t depend on cables or reliable operation. It has many desirable eatures such as expandability, modularity, lexibility and redundancy [6], [62], [78]. However, the droop control concept has some limitation including requency and amplitude deviations, slow transient response and the possibility o circulating current among inverters due to line impedance mismatches between inverters and the common bus. Recently, researchers have improved the two control strategies, or combined these two control method to overcome the corresponding drawbacks. The potential advantages and disadvantages o the communication based methods and the 0 droop methods are summarized in Tables I and II. From these two Tables, it is diicult or only one control scheme to overcome all drawbacks or all applications. However, urther investigation o these control techniques will help improve the design and implementation o uture distributed AC microgrid architectures. The uture trends in control strategies or microgrid are essentially related to energy services and protection, which include the demand response, optimal power low, market participation, storage management, and so on. These technologies could be interesting when connecting microgrids to the main grid or when deploying a cluster o multiple microgrids. Thus, multi-agent systems and hierarchical control [62] could negotiate the interchange o energy between microgrids or microgrid clusters. Thereore, the multi-agent control and hierarchical control are becoming a clear trend o research in microgrids technologies, while communication systems are becoming more important to make these applications easible. In addition, the research about the impacts o stability and reliability with a large number o microgrids connection are still behind. So, it s also diicult to convert the current conventional distribution network structure in short time. The details o uture trends include the ollowing: Network-based Hybrid Distributed Control: Smart distributed grid has been proved that it can improve the eiciency and reliability o the power system. Network-based hybrid distributed control o microgrids is essential to optimize the perormance o microgrids under high penetration level o DG resources, which is treated by algebraic graph theory. A converge analyses are carried out in [0], and it proposes a control scheme which can not only realize requency recovery, accurate power sharing, high reliability and robustness, but also optimize the energy low in the system. Fault-tolerant control: The ault-tolerant control is a key technology area which should not only manage supply and demand o electricity more eectively, but also apply appropriate corrective actions to eliminate, mitigate and prevent various emergency situations such as aults, outages, disturbances to power quality or changes in the user needs [02]. Moreover, the ault tolerant control also can be implemented or sel-healing and anti-islanding which enhances the capability o ault-ride through and ensures the reliability and security o the systems. Cost-Prioritized Droop Schemes: All the optimization schemes are to reduce the cost, so it seems to more easibly and eectively propose a droop scheme based on cost-prioritized. Reerence [03] proposes several droop schemes in consideration o operating costs. These are nonlinear variable droop schemes which regard the related cost unction as the droop coeicients. ariable Inertia: The traditional bulk power system consists o many synchronous generators with a relatively large inertia. But microgrids don t have the kinetic energy and spinning reserve, which consist o many inverter-based distributed resources with a low inertia. Then, the low inertia may lead to severe voltage or requency deviations

13 in some big disturbances and sudden changes. So the system should show a large inertia when the requency will deviate, and a low inertia when to recover the requency. The objective o the variable inertia is always to keep the normal requency. Stability issues: The stability o microgrid has been studied or long years. However, the stability o the microgrid has never been studied perectly when it supplies some complex loads such as the dynamic loads, the constant loads, inductor motor, the pulsed loads and the electric vehicles. So, it s necessary to propose the special models and control methods to solve the voltage, requency and power-angle stabilities or these composite loads. III. CONCLUSION This paper has presented an overview o the dierent power sharing control strategies o DGs in islanding AC microgrids. Detailed description o the control schemes has been given. The communication based methods o concentrated control, master/slave control, and distributed control perorm a good current sharing, yet a low reliability and redundancy. However, the droop characteristic based control method has been presented to avoid communication lines/cables, thus, which can help increase the system reliability, modularity and lexibility. Also, improvement o virtual structure based method, constructed and compensated based method, common variable based method, and signal injection method, have been proposed to overcome the inherent drawbacks o the traditional droop methods or decoupling the active and reactive control laws, robustness with respect to the system parameters, addressing nonlinear loads, and proper voltage regulation. Moreover, various control approaches are compared in terms o their respective advantages and disadvantages. Finally, the uture trends or primary control techniques o AC microgrids are briely discussed. The studies show that in the process o development o microgrid, challenges and opportunities coexist. Communication based control Concentrated control [23]-[27] Master/slave control [28]-[3] Distributed control [24], [32],[33] TABL I. Potential Advantages And Disadvantages O The Communication Based Control MethodS Potential advantages Good power sharing in steady state and transients Constant voltage and requency regulation Recover the output voltage easily Good power sharing in steady state Symmetrical or every module Constant voltage and undamental power sharing. Potential disadvantages High bandwidth communication required Low reliability and expandability High current overshoot during transients Require high bandwidth communication Low redundancy Require communication bus Degrade the modularity o the system 2

14 TABL II. Potential Advantages And Drawbacks O Droop Characteristic Based Control Methods Droop characteristic based control Potential advantages Potential drawbacks Conventional and variants on droop control irtual structure based method Constructed and Compensated based method Conventional requent droop control [6],[40] PD/FB droop control [9], [46]-[48] Complex line impedance [52]-[58] Angle droop control [43]-[45] Droop control with constant power band [49]-[5] irtual output impedance control [9],[53],[6]-[62] nhanced virtual impedance control [63] irtual rame transormation method [6],[65]-[68] Adaptive voltage droop control [69] Synchronized reactive power compensation [70]-[7] Droop control based Synchronized operations [82] dot control method [72]-[74] Common variable based control method [48],[75], [76] Signal injection method [36], [77] asy implementation without communication High expandability, modularity and lexibility For highly resistive transmission lines asy implementation without communication Decoupled active and reactive controls Improved voltage regulation Constant requency regulation Considering the speciic characteristic o micro-source Operating in MPPT within a certain range and energy used more eiciently Avoiding voltage-limit violation Not aected by the physical parameters Improved perormance o power sharing and system stability Can handle linear and nonlinear loads power sharing Mitigates the PCC harmonic voltage Decoupled active and reactive power controls Improved voltage regulation Improved system stability and power sharing under heavy load condition Improved power sharing perormances Not inluenced by the physical parameters Improved power sharing perormances Not aected by the physical parameters Robust to communication delay Same as conventional droop Accurate reactive power sharing Not aected by the physical parameters Can handle linear and nonlinear loads Not aected by the system parameters Aected by the physical parameters Poor voltage-requency regulation Slow dynamic response Poor harmonic sharing Aected by the physical parameters Poor voltage and requency regulation Line impedances should be known in advance Require GPS signals Poor perormance o power sharing Micro-source requires dispatched abilities Require multi-stages controllers and aect system eiciency regulation isn t guaranteed Requires relatively high bandwidth or controller Requires the low-bandwidth communication The physical parameters should be known in advance Hard to exactly ensure the same transormation angle or all DGs The physical parameters should be known in advance The physical parameters should be known in advance Requires the low bandwidth synchronized communication Requires the simple low bandwidth synchronized communication Depend on the initial conditions Steady-state solution may not exist asy to destabilize Hard to measure the common voltage due to long distance Cause harmonic distortion o voltage IX. RFRNCS [] K. Moslehi and R. Kumar, "A reliability perspective o the smart grid," I Trans. Smart Grid, vol., no., pp.57-64, Jun.200. [2] R. H. Lasseter, "Microgrids," in Proc. I Power ngineering Society Winter Meeting, 2002, pp [3] R. H. Lasseter and P. Paigi, "Microgrid: A conceptual solution," in Proc. I power electro special con., 2004, pp [4] J. Rocabert, A. Luna, F. Blaabjerg, and P. Rodriguez, "Control o power converters in AC microgrids," I Trans. Power lectron., vol.27, no., pp , Nov.202. [5] A. Molderink,. Bakker, M. G. C. Bosman, J. L. Hurink, and G. J. M. Smit, "Management and control o domestic smart grid technology," I Trans. Smart Grid, vol., no.2, pp.09-9, Sep.200. [6] K. Debrabandere, B. Bolsens, J. an den Keybus, A. Woyte, J. Driesen, and R. Belmans, "A voltage and requency droop control method or parallel inverters," I Trans. Power lectron., vol.22, no.4, pp.07-5, Jul [7] F. Blaabjerg, R. Teodorescu, M. Liserre, and A.. Timbus, "Overview o control and grid synchronization or distributed power generation systems," I Trans. Ind. lectron, vol.53, no.5, pp , Oct [8] J. J. Justo, F. Mwasilu, and J. Lee, "AC microgrids versus DC microgrids with distributed energy resources: A review," Renewable and Sustainable nergy Reviews, vol.24, pp , Aug.203. [9] M. A. ltawil, and Z. Zhao, "Grid-connected photovoltaic power systems: Technical and potential problems-a review," Renewable and Sustainable nergy Reviews, vol.4, no. pp.2-29, Jan.200. [0] Y. Li, D. M. ilathgamuwa, and P.C. Loh,"Design, analysis, and real-time testing o a controller or multi-bus microgrid system," I 3

15 Trans. on Power lectron., vol.9, no.5, pp ,sep [] P. Piagi and R. H. Lasseter, "Autonomous control o microgrids," in Proc. Power ngineering Society General Meeting, 2006, pp.8-5. [2] P. Arboleya, D. Diaz, and J. M. Guerrero,"An improved control scheme based in droop characteristic or microgrid converters," in Proc. lectric power systems research, 200, pp [3] J. A. P. Lopes, C. L. Moreira, and A. G. Madureira, "Deining control strategies or microgrids islanded operation," I Trans. Power Syst., vol.2, no.2, pp , May [4] F. Katiraei, M. R. Iravani, and P. W. Lehn, "Micro-grid autonomous operation during and subsequent to islanding process," I Trans. Power Del., vol.20, no., pp , Jan [5] N. Lidula and A. Rajapakse, "Microgrids research: a review o experimental microgrids and test systems," Renewable and Sustainable nergy Reviews, vol.5, no., pp , Jan.20. [6] M. C, Chandorkar, D. M. Divan, and R. Adapa, "Control o parallel connected inverters in standalone ac supply systems," I Trans. Ind. Appl., vol.29, no. pp.36-43, Jan.993. [7] M. C. Chandrokar, D. M. Divan, and B. Blaabjerg, "Control o distributed UPS systems," in Proc. 25th Annual I Power lectron. Specialists Con., 994, pp [8] Y. B. Byun, T. G. Koo, and K. Y. Joe, "Parallel operation o three-phase UPS inverters by wireless load sharing control," in Proc. Telecommunications nergy Con., 2000, pp [9] J. M. Guerrero, L. Garciadeicuna, and J. Matas, "Output impedance design o parallel-connected UPS inverters with wireless load-sharing control," I Trans. Ind. lectron., vol.52, no.4, pp.26-35, Aug [20] J. M. Guerrero, L. Hang, and J. Uceda, "Control o distributed uninterruptible power supply systems," I Trans. Ind. lectron., vol.55, no.8, pp , Aug [2] J. M. Guerrero, J. C. asquez, and J. Matas, "Control strategy or lexible microgrid based on parallel line-interactive UPS systems," I Trans. Ind. lectron., vol.56, no.3, pp , Mar [22] T. L. andoorn, J. D. M. De Kooning, and B. Meersman, "Review o primary control strategies or islanded microgrids with power-electronic interaces," Renewable and Sustainable nergy Reviews, vol.9, pp , Mar.203. [23] D. Shanxu, M. Yu, and X. Jian, "Parallel operation control technique o voltage source inverters in UPS," in Proc. o the I 999 International Con. 999, pp [24] M. Prodanovic, T. C. Green, and H. Mansir, "A survey o control methods or three-phase inverters in parallel connection," in Proc. 8th International Con. Power lectron. and ariable Speed Drives, 2000, pp [25] K. Siri, C.. Lee, and T. F. Wu, "Current distribution control or parallel connected converters part ii," I Trans. Aerospace and lectronic Syst., vol.28, pp.84-85, Jul.992. [26] T. F. Wu, K. Siri, and J. Banda, "The central-limit control and impact o cable resistance in current distribution or parallel-connected DC-DC converters," in Proc. 25th Annual I Power lectron. Specialists Con., 994, pp [27] J. Banda and K. Siri, "Improved central-limit control or parallel operation o DC-DC power converters," in Proc. 26th Annual I Power lectron. Specialists Con.,995, pp [28] K. Siri, C.. Lee, and T.. Wu,"Current distribution control or parallel connected converters I," I Trans. Aerospace and lectronic Syst., vol.28, no.3, pp , Jul.992. [29] J. F. Chen and C. L. Chu, "Combination voltage-controlled and current-controlled PWM inverters or UPS parallel operation," I Trans. Power lectron., vol.0, no.5, pp , Sep.995. [30] F. Petruzziello, P. D. Ziogas, and G. Joos, "A novel approach to paralleling o power converter units with true redundancy," in Proc. 2st Annual I Power lectron. Specialists Con. 990, pp [3] Y. Pei, G. Jiang, and X. Yang, "Auto-master-slave control technique o parallel inverters in distributed AC power systems and UPS," in Proc. 35th Annual I Power lectron. Specialists, 2004, pp [32] J. Tan, H. Lin, and J. Zhang, "A novel load sharing control technique or paralleled inverters," in Proc. 34th Annual I Power lectron. Specialist Con., 2003, pp [33] X. Sun, Y. S. Lee, and D. Xu, "Modeling, analysis, and implementation o parallel multi-inverter systems with instantaneous average current sharing scheme," I Trans. Power lectron., vol.8, no.3, pp , May [34] J. He and Y. W. Li, "An nhanced Microgrid Load Demand Sharing Strategy," I Trans. Power lectron., vol.27, no.9, pp , Sep.202. [35] A. Tuladhar, H. Jin, and T. Unger, "Parallel operation o single phase inverter modules with no control interconnections," in Proc. o Twelth Annual I Applied Power lectron. and xposition, 997, pp [36] A. Tuladhar, H. Jin, and T. Unger, "Control o parallel inverters in distributed AC power systems with consideration o line impedance eect," I Trans. Ind. Appl., vol.36, no., pp.3-38, Jan [37] J. C. asquez, J. M. Guerrero, and A. Luna, "Adaptive droop control applied to voltage-source inverters operating in grid-connected and islanded modes," I Trans. Ind. lectron., vol.56, no.0, pp , Oct [38] Y. W. Li and C. N. Kao, "An accurate power control strategy or power-lectronic-interaced distributed generation units operating in a low-voltage multibus microgrid," I Trans. Power lectron., vol.24, no.2, pp , Dec [39] P. H. Divshali, A. Alimardani, and S. H. Hosseinian, "Decentralized cooperative control strategy o microsources or stabilizing autonomous SC-based microgrids," I Trans. Power Syst., vol.24, no.4, pp , Nov.202. [40] F. Katiraei and M. R. Iravani, "Power management strategies or a microgrid with multiple distributed generation units," I Trans. Power Syst., vol.2, no.4, pp.82-83, Nov [4]. A. A. Coelho, P. C. Cortizo, and P. F. D. Garcia,"Small-signal stability or parallel-connected inverters in stand-alone AC supply systems," I Trans. Ind. Appl., Mar.2002;38(2): [42] N. Pogaku, M. Prodanovic, and T. C. Green, "Modeling, analysis and testing o autonomous operation o an inverter-based microgrid," I Trans. Power lectron., vol.22, no.2, pp , Mar [43] R. Majumder, A. Ghosh, and G. Ledwich, "Angle droop versus requency droop in a voltage source converter based autonomous microgrid," in Proc. Power & nergy Society General Meeting, 2009, pp.-8. [44] R. Majumder, B. Chaudhuri, and A. Ghosh, "Improvement o stability and load sharing in an autonomous microgrid using supplementary droop control loop," I Trans. Power Syst., vol.25, no.2, pp , May.200. [45] R. Majumder, G. Ledwich, and A. Ghosh, "Droop control o converter-interaced microsources in rural distributed generation," I Trans. Power Del., vol.25, no.4, pp , Oct.200. [46] J. M. Guerrero, J. Matas, and L. G. de icuña,"decentralized control or parallel operation o distributed generation inverters using resistive output impedance," I Trans. Ind. lectron., vol.54, no.2, pp , Apr [47] X. Yu, A. M. Khambadkone, and H. Wang, "Control o parallel-connected power converters or low-voltage microgrid Part I: A hybrid control architecture," I Trans. Power lectron., vol.25, no.2, pp , Dec.200. [48] C. K. Sao and P. W. Lehn, "Control and power management o converter ed microgrids," I Trans. Power Syst., vol.23, no.3, pp , Aug [49] T. L. andoorn, B. Meersman, and L. Degroote, "A control strategy or islanded microgrids with dc-link voltage control," I Trans. Power Del., vol.26, no.2, pp , Apr.20. [50] T. L. andoorn, B. Renders, and L. Degroote, "Active load control in 4

16 islanded microgrids based on the grid voltage," I Trans. Smart Grid, vol.2, no., pp.39-5, Mar.20. [5] T. L. andoorn, B. Meersman, and J. De Kooning, "Automatic power sharing modiication o P/ droop controllers in low-voltage resistive microgrids," I Trans. Power Del., vol.27, no.4, pp , Oct.202. [52] A. Moawwad,. Khadkikar, and J. L. Kirtley, "A New P-- Droop Control Method or an Interline Photovoltaic (I-P) Power System," I Trans. Power Del., vol.28, no.2, pp , Apr.203. [53] W. Yao, M. Chen, and J. Matas, "Design and analysis o the droop control method or parallel inverters considering the impact o the complex impedance on the power sharing," I Trans. Ind. lectron., vol.58, no.2, pp , Feb.20. [54] J. C. asquez, J. M. Guerrero, and A. Luna, "Adaptive droop control applied to voltage-source inverters operating in grid-connected and islanded modes," I Trans. Ind. lectron., vol.56, no.0, pp , Oct [55] H. Hanaoka, M. Nagai, and M. Yanagisawa, "Development o a novel parallel redundant UPS," in Proc. the 25th International I Telecommunications nergy Con. 2003, pp [56] J. M. Guerrero, L. G. de icuña, and J. Matas, "A wireless controller to enhance dynamic perormance o parallel inverters in distributed generation systems," I Trans. Power lectron., vol.9, no.5, pp , Sep [57] G. Yajuan, W. Weiyang, and G. Xiaoqiang, "An improved droop controller or grid-connected voltage source inverter in microgrid," in Proc. the 2nd I International Symposium on Power lectron. or Distrib. Gen. Syst., 200, pp [58] Y. Mohamed and. F. l-saadany, "Adaptive decentralized droop controller to preserve power sharing stability o paralleled inverters in distributed generation microgrids," I Trans. Power lectron., vol.23, no.6, pp , Nov [59] A. M. Salamah, S. J. Finney, and B. W. Williams, "Autonomous controller or improved dynamic perormance o AC grid, parallel-connected, single-phase inverters," IT Gen., Trans. Distrib., vol.2, no.2, pp , Mar [60] T. Goya,. Omine, and Y. Kinjyo, "Frequency control in isolated island by using parallel operated battery systems applying H control theory based on droop characteristics, " IT Renewable Power Gen., vol.5, no.2, pp.60-66, Mar.20. [6] J. M. Guerrero, J. C. asquez, and J. Matas, "Hierarchical control o droop-controlled AC and DC microgrids A general approach toward standardization, "I Trans. Ind. lectron., vol.58, no., pp.58-72, Jan.20. [62] J. M. Guerrero, P. Loh, and M. Chandorkar, "Advanced Control Architectures or Intelligent MicroGrids Part I: Decentralized and Hierarchical Control," I Trans. Ind. lectron., vol.60, no.4, pp , Apr.203. [63] J. He, Y. W. Li, and J. M. Guerrero, "An Islanding Microgrid Power Sharing Approach Using nhanced irtual Impedance Control Scheme, " I Trans. Power lectro., vol.28,no.,pp , Nov.203. [64] J. He and Y. W. Li, "Analysis, design, and implementation o virtual impedance or power lectron," interaced distributed generation. I Trans. Ind. Appl., vol.47, no.6, pp , Nov.20. [65] C. T. Lee, C. C. Chuang, and C. C. Chu, "Control strategies or distributed energy resources interace converters in the low voltage microgrid," in Proc. I nergy Conversion Congress and xposition, 2009, pp [66] Y. Li and Y. W. Li, "Decoupled power control or an inverter based low voltage microgrid in autonomous operation," in Proc. I 6th International Power lectron. and Motion Control,2009, pp [67] Y. Li and Y. W. Li, "irtual requency-voltage rame control o inverter based low voltage microgrid," in Proc. I lectrical Power & nergy Con. 2009, pp.-6. [68] Y. Li and Y. W. Li, "Power management o inverter interaced autonomous microgrid based on virtual requency-voltage rame," I Trans. Smart Grid, vol.2, no., pp.30-40, Mar.20. [69]. Rokrok and M.. H. Golshan, "Adaptive voltage droop scheme or voltage source converters in an islanded multi-bus microgrid," IT generation, trans. distrib., vol.4, no.5, pp , May.200. [70] J. He and Y. W. Li, "An accurate reactive power sharing control strategy or DG units in a microgrid," in Proc. I 8th International Con. Power lectron. and CC Asia,20, pp [7] J. He and Y. W. Li, "An nhanced Microgrid Load Demand Sharing Strategy," I Trans. Power lectron., vol.27, no.9, pp ,sep.202. [72] C. T. Lee, C. C. Chu, and P. T. Cheng, "A new droop control method or the autonomous operation o distributed energy resource interace converters," in Proc. I nergy Conversion Congress and xposition, 200, pp [73] C. T. Lee, C. C. Chu, and P. T. Cheng, "A New Droop Control Method or the Autonomous Operation o Distributed nergy Resource Interace Converters," I Trans. Power lectron., vol.28, no.4, pp , Apr.203. [74] C. T. Lee, R. P. Jiang, and P. T. Cheng, "A grid synchronization method or droop controlled distributed energy resources converters," in Proc. I nergy Conversion Congress and xposition, 20, pp [75] C. K. Sao and P. W. Lehn, "Autonomous load sharing o voltage source converters, " I Trans. Power Del., vol.20, no.2, pp , Apr [76]. C. Zhong, "Robust droop controller or accurate proportional load sharing among inverters operated in parallel," I Trans. Ind. lectron., vol.60, no.4, pp , Apr.203. [77] D. J. Perreault, R. L. Selders Jr, and J. G. Kassakian, "Frequency-based current-sharing techniques or paralleled power converters," I Trans. Power lectron., vol.3, no.4, pp ,jul.998. [78] A. Bidram and A. Davoudi, "Hierarchical structure o microgrids control system," I Trans. Smart Grid, vol.3, no.4, pp ,dec.202. [79]. Mojica-Nava, C. A. Macana, and N. uijano, "Dynamic population games or optimal dispatch on hierarchical microgrid control," I Trans. Syst., Man, and Cybernetics: Systems, ol.44, No.3, pp , Mar.204. [80] L. Meng, F. Tang, M. Savaghebi, J. C. asquez, and J. M. Guerrero, "Tertiary Control o Unbalance Compensation or Optimal Power uality in Islanded Microgrids," I Trans. nergy Conv., ol.29, No.4, pp , Dec.204. [8] F. Shahnia, R. P. S Chandrasena, S. Rajakaruna, and A. Ghosh, "Primary control level o parallel distributed energy resources converters in system o multiple interconnected autonomous microgrids within sel-healing networks," IT Gen., Trans. and Distrib., ol.8, No.2, pp , Feb.204. [82] J. Bingnan, and F. Yunsi, "Smart Home in Smart Microgrid: A Cost-ective nergy cosystem with Intelligent Hierarchical Agents," I Trans. Smart Grid, ol.6, No., pp.3-3, Jan.205. [83] X. Wu, W. Xiuli, and. Chong, "A Hierarchical Framework or Generation Scheduling o Microgrids," I Trans. Power Del., ol.29,no.6, pp , Dec.204. [84] D.. Olivares, A. Mehrizi-Sani, A. H. temadi, C. A. Cañizares, R. Iravani, M. Kazerani, A. H. Hajimiragha, O. Gomis-Bellmunt, M. Saeediard, R. Palma-Behnke, G. A. Jiménez-stévez, and N. D. Hatziargyriou, "Trends in microgrid control," I Trans. Smart Grid, ol.5,no.4,pp , Jul.204. [85] M. Yazdanian and A. Mehrizi Sani, "Distributed Control Techniques in Microgrids," I Trans. Smart Grid, ol.5, No.6, pp , Nov.204. [86] M. A. shabani and Y. A. I. Mohamed, "New Family o Microgrid Control and Management Strategies in Smart Distribution Grids Analysis, Comparison and Testing," I Trans. Power Syst., ol.29, No.5, pp , Sep.204. [87] B. M. id, N. A. Rahim, J. Selvaraj, and A. H. l Khateb, "Control methods and objectives or electronically coupled distributed energy resources in microgrids: a review," I Syst. J., pp.-3, 204. [88] M. M. A. Abdelaziz, M. F. Shaaban, H.. Farag, and. F. l-saadany, "A multistage centralized control scheme or islanded microgrids with P," I Trans. Sustainable nergy, ol.5, No.3, pp , Jul

17 [89] T. Caldognetto, and T. Paolo, "Microgrids Operation Based on Master-Slave Cooperative Control," I Journal o merging and Selected Topics in Power lectron., ol.2, No.4, pp , Dec.204. [90]. Shaiee, J. M. Guerrero, and. C. Juan, "Distributed secondary control or islanded microgrids A novel approach," I Trans. Power lectron., ol.29, No.2, pp.08 03, Feb.204. [9] A. Bidram, A. Davoudi, and F. Lewis, "A Multi-objective Distributed Control Framework or Islanded AC Microgrids," I Trans. Ind. Inormatics, ol.0, No.3, pp , Aug [92] H. Xin, L. Zhang, Z. Wang, D. Gan, and K. P. Wong, "Control o Island AC Microgrids Using a Fully Distributed Approach," I Trans. Smart Grid, ol.6, No.2, pp , Mar.205. [93] M. A. Mahmud, M. J. Hossain, H. R. Pota, and A. M. T. Oo, "Robust Nonlinear Distributed Controller Design or Active and Reactive Power Sharing in Islanded Microgrids," I Trans. nergy Conv., ol.29, No.4,pp , Dec.204. [94] J. Hu, J. Zhu, D. G. Dorrell, and J. M. Guerrero, "irtual Flux Droop Method A New Control Strategy o Inverters in Microgrids," I Trans. Power lectron., ol.29, No.9, pp , Sep.204. [95] M. Ashabani, Y. Mohamed, M. Mirsalim, and M. Aghashabani, "Multivariable Droop Control o Synchronous Current Converters in Weak Grids/Microgrids With Decoupled textit {dq}-axes Currents," I Trans. Smart Grid, to be published. [96] L. Y. Lu, and C. C. Chu, "Consensus-Based Droop Control Synthesis or Multiple DICs in Isolated Micro-Grids," I Trans. Power Systems, to be published. [97] X. Yu, A. M. Khambadkone, and H. H. Wang, "A hybrid control architecture or low voltage microgrid," in Proc. I nergy Conversion Congress and xposition, 200, pp [98] Y. Zhang and H. Ma, "Theoretical and experimental investigation o networked control or parallel operation o inverters," I Trans. Ind. lectron., vol.59, no.4, pp ,apr.202. [99] M. N. Marwali, J. W. Jung, and A. Keyhani, "Control o distributed generation systems-part II: Load sharing control," I Trans. Power lectron., vol.9, no.6, pp ,nov [00] H. Hua, L. Yao, S. Yao, S. Mei, and J. M. Guerrero, "An Improved Droop Control Strategy or Reactive Power Sharing in Islanded Microgrid," I Trans. Power lectron., ol.30, No.6, pp , Jun.204. [0] A. Kahrobaeian and Y. Abdel-Rady Ibrahim Mohamed, "Networked Based Hybrid Distributed Power Sharing and Control or Islanded Micro-Grid Systems," I Trans. Power lectron.,ol.30, No.2, pp , Feb.205 [02] A. argas-martínez, L. I. Avila Minchala, Y. Zhang, L.. Garza-Castañón, and H. Badihi, "Hybrid Adaptive Fault-Tolerant Control Algorithms or and Frequency Regulation o an Islanded Microgrid," Int. Trans. lectron. nerg. Syst., Jan.204. [03] I. U. Nutkani, P. C. Loh, and F. Blaabjerg, "Cost-Prioritized Droop Schemes or Autonomous AC Microgrids," I Trans. Power lectron., ol.30, No.2, pp.09-9, Feb.205. Hua Han received the M.S. and Ph.D. degrees rom the School o Inormation Science and ngineering, Central South University, Changsha, China, in 998 and 2008, respectively. She was a visiting scholar o University o Central. Florida, Orlando, FL, USA, rom April 20 to April 202. She is currently an associate proessor with the School o Inormation Science and ngineering, Central South University, China. Her research interests include microgrid, and power electronic equipment. Xiaochao Hou received the B.S. degree in Automation rom the Central South University, Changsha, China, in 204, where he is currently working toward M.S. degree in electrical engineering. His research interests include Renewable nergy Systems, Distributed generation and micro-grid. Jia Wu received the B.S. and M.S. degrees rom the School o Inormation Science and ngineering, Central South University, Changsha, China, in 2006 and 2009, respectively, where he is currently working toward Ph.D. degree in electrical engineering. His research interests include photovoltaic power generation system and wind energy conversion system. Mei Su received the B.S., M.S. and Ph.D. degrees rom the School o Inormation Science and ngineering, Central South University, Changsha, China, in 989, 992 and 2005, respectively. Since 2006, She has been a Proessor with the School o Inormation Science and ngineering, Central South University. Her research interests include matrix converter, adjustable speed drives, and wind energy conversion system. Jian Yang (M 09) received the Ph.D. degree in electrical engineering rom the University o Central Florida, Orlando, in He was a Senior lectrical ngineer been with Delta Tau Data Systems, Inc., Los Angeles, CA, rom 2007 to 200. Since 20, he has been with Central South University, Changsha, China, where he is currently an Associate Proessor with the School o Inormation Science and ngineering. His main research interests include control application, motion planning, and power electronics. Josep M. Guerrero (S 0-M 04-SM 08-F'5) received the B.S. degree in telecommunications engineering, the M.S. degree in electronics engineering, and the Ph.D. degree in power electronics rom the Technical University o Catalonia, Barcelona, in 997, 2000 and 2003, respectively. Since 20, he has been a Full Proessor with the Department o nergy Technology, Aalborg University, Denmark, where he is responsible or the Microgrid Research Program. Since 202 he has been a guest Proessor at the Chinese Academy o Science and the Nanjing University o Aeronautics and Astronautics; and since 204, he has been chair Proessor in Shandong University.His research interests is oriented to dierent microgrid aspects, including power electronics, distributed energy-storage systems, hierarchical and cooperative control, energy management systems, and optimization o microgrids and islanded minigrids. Pro. Guerrero is an Associate ditor or the I TRANSACTIONS ON POWR LCTRONICS, the I TRANSACTIONS ON INDUSTRIAL LCTRONICS, and the I Industrial lectronics Magazine, and an ditor or the I TRANSACTIONS on SMART GRID. He has been Guest ditor o the I TRANSACTIONS ON POWR LCTRONICS Special Issues: Power lectronics or Wind nergy Conversion and Power lectronics or Microgrids; the I TRANSACTIONS ON INDUSTRIAL LCTRONICS Special Sections: Uninterruptible Power Supplies systems, Renewable nergy Systems, Distributed Generation and Microgrids, and Industrial Applications and Implementation Issues o the Kalman Filter; and the I TRANSACTIONS on SMART GRID Special Issue on Smart DC Distribution Systems. He was the chair o the Renewable nergy Systems Technical Committee o the I Industrial lectronics Society. In 204 he was awarded as ISI Highly Cited Researcher, and in 205 he was elevated as I Fellow or his contributions on distributed power systems and microgrids. 6