Abstract— This paper presents the operation and controllerdesign of a microgrid consisting of a direct drive wind gener-ator and a battery storage system. A model predictive controlstrategy for the ac-dc-ac converter of wind system is derived andimplemented to capture the maximum wind energy as well asprovide desired reactive power. A novel supervisory controlleris presented and employed to coordinate the operation of windfarm and battery system in the microgrid for grid-connectedand islanded operations. The proposed coordinated controllercan mitigate both active and reactive power disturbances thatare caused by the intermittency of wind speed and load change.Moreover, the control strategy ensures the maximum powerextraction capability of wind turbine while regulating the pointof common coupling bus voltage within acceptable range in bothgrid-connected and islanded operations. The designed concept isverified through various simulation studies in EMTDC/PSCAD,and the results are presented and discussed.
Index Terms— Coordinated controller, direct drive windturbine, islanded operation, maximum power extractioncapability, model predictive control (MPC).
I. INTRODUCTION
RENEWABLE energy such as wind is attracting greatattention as a clean and inexhaustible resource and it
has been turning to the alternatives of fossil. Therefore,maximizing the usage of wind energy has received nationaland worldwide attention. The wind turbine technology can bedivided into two categories: constant speed constant frequency(CSCF) and variable speed constant frequency (VSCF) [1].Although as the main member of CSCF wind turbine, squirrelcage inductor generator (SCIG) has merits of economy andsimplicity, it cannot capture maximum wind energy. VSCFwind generator, on the other hand, is more attractive because ofthe improvement in wind energy production and the flexibilityof output power regulation. VSCF wind turbine is connectedto the electric power network through ac-dc-ac inverter. Withproperly designed controller, VSCF can generate 20%–30%more energy than the CSCF, and can reduce active poweroscillation and provide dynamic reactive power support [2].
Manuscript received June 17, 2013; revised September 11, 2013; acceptedSeptember 15, 2013. Date of publication September 18, 2013; date of currentversion October 29, 2013. This work was supported by the NETL-RUA GridTechnology Collaborative under Grant 10013520.269.1005682R. Recom-mended for publication by Associate Editor Wenzhong Gao.
The authors are with the Lane Department of Computer Scienceand Electrical Engineering, West Virginia University, Morgantown, WV26505 USA (e-mail: [email protected]; [email protected]; [email protected]).
Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JESTPE.2013.2282601
Kim and Kim [3] present a linear proportional integral (PI)controller, which can capture the maximum energy fromvarying wind speed as well as deliver desired reactive power.However, the controller’s performance is compromised whenthe wind speed changes rapidly. Nonlinear control strategy likefuzzy logic and neural network are discussed in [4]–[9]. Thesecontrollers have fast response and are capable of predictingwind generation; however, they are difficult to implement inindustry standard simulation tools like PSCAD/EMTDC. Amodel predictive control (MPC) strategy for wind ac-dc-acinverter is presented in this paper to look-ahead current windprofile that optimizes system operation in transient situation.
However, in remote distribution network, it is desirable thatdistributed generators can supply local load. Under wind farmnormal operation, the power generation is fluctuated becauseof the intermittency of wind speed. The control design of windpower electronic interface itself cannot solve the discontinu-ity of active power supply. Mismatch between demand andsupply may occur and induce voltage stability issues, and theinfluence may vary depending on penetration level, fault ridethrough ability, and reactive power support. Hence, auxiliarydevices are required to address the active power intermittency.Storage devices such as batteries are considered to be asolution to the active power intermittency related to windgeneration. Battery storage provides operation freedom towind generation that allows time-shifting between generationand demand. Energy storage in a microgrid is very similarto any other inverter-based voltage source with the abilitiesof plug-and-play, fast response, and bidirectional power-flowcapabilities [10]. Since battery is used to compensate thevariability in wind energy, the coordination between energystorage and wind farm is critical for managing variationsin load demand. Prior works detailing modeling and designof hybrid systems with different renewable resources werepresented in [11]–[14]. A microgrid’s supervisory control canregulate the active power dispatch among different renewableresources as presented in [11]–[13]. A coordinated control isproposed in [14] to maintain the system frequency by adjustingthe output power of different distributed generators. How-ever, reactive power coordination between battery and windgeneration has received little research attention. Insufficientreactive power compensation can cause voltage stability issues,which are worsened by the remote location of wind farm.Other than the well-known capability of providing instantactive power supply, battery can also provide dynamic reactivepower support when equipped with inverter interface, and itis capable and should be used to maintain voltage stability
HAN et al.: COORDINATED PREDICTIVE CONTROL OF A WIND/BATTERY MICROGRID SYSTEM 297
Fig. 1. One-line diagram of the microgrid.
together with wind farm. This paper designs a MPC-basedcoordinated control scheme, which enables the microgridto provide sustainable power as well as dynamic reactivepower support to the load in both grid-connected and islandedoperations, thus reducing active power oscillation as well astracking maximum wind energy.
II. SYSTEM DIAGRAM
The schematic representation of the microgrid system isshown in Fig. 1. The system is composed of a VSCF wind gen-erator, battery storage, and two different kinds of load: a fixedimpedance load and a dynamic load with controllable power.Wind generator has a gearless direct drive wind turbine, amultipole permanent magnet synchronous generator (PMSG),and an ac-dc-ac converter. The battery system consists ofa lithium-ion storage unit and a dc-ac inverter. An infinitevoltage source in series with a high inductor is used torepresent a weak electric grid. The controller design will beintroduced in the following sections.
III. MPC STRATEGY OF WIND AC-DC-AC INVERTER
The modeling of wind turbine introduced in [2] is adoptedin this paper. However, instead of using diode rectifier, an Insu-lated Gate Bipolar Transistor (IGBT) based ac-dc-ac inverter isused as the interface between wind turbine and grid as shownin Fig. 2.
A. Basic Operation of ac-dc-ac Inverter
A six IGBT bridge voltage source inverter (VSI) is locatedon both sides of the dc bus as shown in Fig 2. In a two-levelVSI in this paper, there are a total of C1
2 available switchingstates in each phase [15]. The switching function of phasesj = a, b, c of VSI-k (k = 1, 2) are controlled so that at anyinstant only one bridge is switched on for each phase, whichmeans either the upper or lower arm is on. Taking advantageof the limited switching states, the desired switching states canbe obtained to track the maximum wind energy and providedesired reactive power.
B. Mathematical Model of a VSI
Assuming the three-phase voltage is balanced, the dynamicbehavior of VSI-k phase j can be expressed as in (1) and (2)
Vjk + Ldijk
dt+ Rijk + Vpjk = Edc
2(1)
Vjk + Ldijk
dt+ Rijk − Vnjk = Edc
2(2)
where V and i are the voltage and current of VSI, respec-tively, Vp and Vn are the upper and lower bridge voltages,respectively, jk denote phase-j of VSI-k, and Edc is the dc busvoltage. Adding (1) with (2), the phase voltage of VSC canbe expressed as
dijk
dt= Vnjk − Vpjk − 2Rijk − 2Vjk
2L. (3)
Based on the switching function, Vpjk and Vnjk can berepresented as
⎧⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
⎡
⎣VpjkVnjkSnjk
⎤
⎦ =⎡
⎣0
Edc0
⎤
⎦ , Spjk = 1
⎡
⎣VpjkVnjkSnjk
⎤
⎦ =⎡
⎣Edc01
⎤
⎦ , Spjk = 0
(4)
where Sp and Sn are the switching states of upper and lowerbridge arms, respectively. In PSCAD, 1 is close and 0 is open.Note that, different softwares use different representation. Thedc bus voltage Edc can be calculated as
idc1 − idc2 = Cd Edc
dt. (5)
C. MPC Strategy
The objectives of the generator side VSI are to regulatethe ac current to capture maximum wind energy as well asmaintain zero reactive power exchange. MPC strategy is sodesigned to obtain the next time step switching states basedon the forecasted wind speed and the dynamic behavior ofVSI. Assuming the next time step wind speed is known atnext time instant, the wind generation and torque with respectto wind speed can be described as [16]–[18]
Pwind_opt = 1
2ρπ R2Cp_optV
3w (6)
Tw = Pw_opt
ωw= 1
2ρπ R2Cp_opt
V 3w
ωw(7)
where ρ is the air density, R is the blade radius,Vw is thewind speed, ωw is the mechanical speed of wind turbine, Tw
is the mechanical torque of wind turbine, and Cp is the powercoefficient and it is a function of ωw , Vw and blade pitchangle β. Cp_opt is the optimum power coefficient and capturesthe maximum wind energy Pwind_opt. The approximation ofCp_opt used in this paper can be found in [19] and [20].
The power exchange between wind turbine and generatorside VSI in next time step can be expressed in synchronous
298 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 1, NO. 4, DECEMBER 2013
Fig. 2. Circuit diagram of wind converter.
Fig. 3. Control scheme of wind converter.
dq0 reference frame as
[P (t + Ts)Q (t + Ts)
]
=[ 1
2ρπ R2CpV 3w (t + Ts)
0
]
=[
32 Vq(t + Ts)Iq (t + Ts)32 Vq(t + Ts)Id (t + Ts)
]
(8)
where d and q are the d-axis and q-axis sequences, respec-tively. Ts is the time step. By selecting a sufficiently smallTs , Vq(t + Ts) can be approximated by the measured valueof Vq(t) [14]. It is seen that the active and reactive powerP(t + Ts) and Q(t + Ts) can be obtained by regulatingIq (t + Ts) and Id (t + Ts). The reference abc current of nexttime step can be calculated by the PI controller as shown inFig. 3. Similarly, the objectives of the grid side VSI are todeliver desired reactive power to the grid as well as maintainthe dc bus voltage. The reference current of grid side VSI canbe obtained in similar manner as shown in Fig. 3. The phasetracking system presented in [21] is applied in this paper toget the synchronous rotating speed.
D. MPC Formulation
An accurate prediction of next time step current is essentialfor tracking. Based on (3), the dynamic discrete-time modelof generator side VSI can be deduced with a backward Eulerapproximation as shown in (9)
i j k (t+Ts) = L
L + R×Tsijk (t)+Ts
× Vnjk (t+Ts) − Vpjk (t+Ts)
2L + 2R×Ts−Ts×Vjk (t + Ts)
L + R×Ts.
(9)
To reduce the error between predicted and reference current,a cost function is presented to evaluate the predicted currentunder all combinations of switching states as shown in (10)
Jjk = ∣∣ijkref (t + Ts) − ijk (t + Ts)
∣∣ . (10)
The MPC strategy is applied both on the VSI of the windgenerator and VSI of battery system, since the battery isconnected to the system through a VSI similar to the grid-side inverter of direct drive wind turbine. The mathematicalmodel and the MPC formulation of VSI can be used for both
HAN et al.: COORDINATED PREDICTIVE CONTROL OF A WIND/BATTERY MICROGRID SYSTEM 299
Fig. 4. Control scheme of battery VSI.
wind generator and battery system. Similar strategy is appliedto battery inverter as shown in Fig. 4, and the control objectiveof battery inverter is set to provide desired power to the grid.The acquisition of the active and reactive power reference forbattery system will be described in the following sections.
IV. MICROGRID OPERATION AND COORDINATED
CONTROL STRATEGY
The coordinated controller for the microgrid is as shown inFig. 5. The monitored power system quantities include point ofcommon coupling (PCC) bus voltage, real and reactive poweroutput of wind and battery as well as the breaker statuses ofwind, battery and load. This section presents the reactive andactive power coordination under different operations.
A. Grid-Connected Operation
1) Reactive Power Coordinated Control: In grid-connectedmode, the synchronous generators in main power grid maintainthe voltage and frequency of the microgrid constant even whenthere is no power exchange on the interconnected transmissionline. However, since the microgrid is normally connected atthe remote end of the power system where lack of reactivepower support is expected, wind and battery are responsible forproviding dynamic reactive power and maintaining the PCCbus voltage. A reactive power coordinated control scheme forgrid-connected operation is designed considering the afore-mentioned criterions as shown in Fig. 5 for the coordinationof wind converter and battery inverter. Six different operationmodes are implemented as follows.
1) Start mode: This mode occurs during the system startingprocedures. The wind farm starts in zero reactive powercontrol (0-RPC) mode and then switches to voltagecontrol mode in steady state. The battery is connected2 s after wind farm starts operating in 0-RPC modeto reduce transients. Wind converter generates reactivepower until PCC bus voltage is constant.
2) Wind dominant mode: This mode is activated after theinitialization mode where wind farm is the primary
reactive power compensator. In this mode the battery isat zero reactive power output regardless of charging ordischarging. In the wind dominant mode, small transientdisturbances are mitigated by the wind farm therebyimproving the life cycle of battery storage.
3) Wind and battery mode: This mode is activated whenthe PCC voltage violates the high limit and the windconverter reaches its maximum reactive power capabil-ity. The battery switches to voltage control mode toprovide dynamic reactive power support. When windand battery both are in voltage control mode, thecoordinated control prevents oscillations by utilizinga basic droop in the voltage versus reactive powercharacteristics of both wind and battery to ensure thevoltage change in PCC bus is absorbed by both windand battery system. The parameters of the controllers areoptimized through a genetic algorithm to achieve opti-mal performance in voltage regulation and active powercontrol.
4) Shedding mode: This mode is activated when the PCCvoltage violates the low limit and both wind farm andbattery are at their maximum reactive power output.Under this condition some of the load may be discon-nected for voltage recovery.
5) Stall mode: This mode is activated when the PCCvoltage violates the high limit and both wind andbattery inverters absorb maximum reactive power. Thewind turbine goes to stall regulation, and the maximumpower control is overridden. The blade pitch angleis increased to reduce the wind energy capture. Theactive power output of the wind turbine is reduced toincrease the maximum reactive power capability of windconverter.
6) Protection mode: This mode is activated when the PCCvoltage violates the protection limits. To protect thepower electronics device from over voltage or undervoltage conditions and other safety issues, both the windand battery system are disconnected from the powergrid.
Time deadband is introduced in this coordinated controller toavoid the excessive mode changes. Table I gives the controlstrategy map for each operation mode. With this algorithm,autonomous operation of wind and battery is achieved, pro-viding flexibility of operation.
The coordinated control expands the operation range andprovides the system four levels of protection:
1) voltage regulation by wind farm only;2) voltage regulation by both farm and battery;3) load shedding and stall regulation;4) wind and battery splitting.The system can also restore from any operation point to
initial conditions if certain requirements are met.2) Active Power Coordinated Control: The uncontrollable
nature of wind energy leads to intermittent active power supplyand hence unbalance between supply and demand. However,in remote distribution network, it is desirable that distributedgenerators supply local load. The power imbalance can bemitigated by the active power coordination shown in Fig. 6.
300 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 1, NO. 4, DECEMBER 2013
Supervisory Control
Grid Connected Operation
Islanded Operation
Reactive Power CoordinationMode SelectionMode I . Starting
Mode II. Wind dominateMode III. Wind & Battery
Mode IV. SheddingMode V. Stall controlMode VI. Protection
Active Power CoordinationMode Selection
Mode I: Over generationMode II: Under generation
Wind TurbineI-IV, VI: MPPTV: stall control
Wind InverterII-V: V control
I, VI: 0-RPC
BSS InverterIII -V: V ControlI,II , VI: 0-RPC
Wind BreakerI-V: CloseVI: Open
BSS BreakerI-V: CloseVI: Open
Load BreakerI-III , V: CloseIV, VI: Open
Wind turbineI-II: MPPT
Wind InverterI-II: MPPT
BSS InverterI: Charging
II: Discharging
Active Power CoordinationMode I: Wind Alone
Mode II: Wind & Battery
Reactive Power Coordination
PV InverterMagnitude
Control
BSS InverterV Control
Wind InverterI-II: Frequency
Control
BSS InverterII: Governor
Control
Wind TurbineI: Primary Mover Control
II: MPPT
Power system quantities
Fig. 5. Coordinated control scheme of the mircrogrid with wind and battery storage.
TABLE I
CONTROL STRATEGY MAP FOR OPERATION MODES UNDER
REACTIVE POWER COORDINATE CONTROL
Fig. 6. Active power coordination.
Equation (11) define the battery active power output
Pwind + Pbatt + Pgrid = Pt + Ploss (11)
where Pwind, Pbatt , and Pgrid are the active power commandof wind, battery and grid, Pt is the total load demand, andPloss is the system loss. The microgrid is so designed that thewind farm and battery can support the total local load, andhence under most conditions, no real and reactive power willbe needed from the grid, resulting in low power losses also.The active power from grid is needed when the wind speed islow and the total load Pt exceeds the maximum active powergeneration of Pwind + Pbatt. After inclusion of Ploss in the
Fig. 7. Islanded control strategy.
equation, no significant change is observed due to negligibletransmission losses in the microgrid.
B. Island Operation
In a standalone operation, the network is isolated from themain ac supply system. There are no synchronous alternatorsthat can provide voltage and frequency references. In thispaper, a seamless transition method is proposed as shown inFig. 7 to calculate the reference voltage angle for islandedcontrol operation. An ideal three-phase sinusoid wave is usedas the base reference angle, and the phase angle of PCCvoltage before islanding is used as the initial phase angleof the sinusoid wave to minimize the oscillation caused bylarge angle shift before and after islanding. Wind farm isthe primary source of supply in the island. Wind convertergenerates a 60 Hz three-phase balanced constant voltage signalas reference. A voltage drop or voltage rise at dc link can occurdue to fluctuations in wind power. Battery is used to regulatethe PCC bus voltage as well as to ensure the maximum powerextraction capability of wind farm during islanded operation.However, in certain cases, partial dynamic load will be usedto maintain power balance.
1) Reactive Power Coordination: The variable reactivepower output of wind farm is compensated by the dynamicreactive power of the battery. Thus the battery inverter is undervoltage control mode in islanded operation.
HAN et al.: COORDINATED PREDICTIVE CONTROL OF A WIND/BATTERY MICROGRID SYSTEM 301
2) Active Power Coordination: For wind system to beoperated in maximum power tracking mode, a battery is usedas governor of the wind turbine and is regulated as shownin Fig. 6. The battery absorbs energy or injects energy tomaintain the dc-link voltage of wind converter. As a result, thewind farm is capable of providing the load. In case of largewind speed drop, load shedding operation will disconnect apart of the load to restore the system frequency.
C. Battery Operation Control
The state of charge (SOC) of the battery is essential inthe coordinated control. It is necessary to avoid depletingor overcharging the battery. Moreover, time deadband tdbprevents frequent operation mode changes and here tdb thetime duration under charging/discharging modes should begreater than time deadband of 0.033 s
BatteryMode
={discharging, SOC>SOClow &Pbattset or Qbattset>0 & tb≥tdbcharging, SOC<SOCup & Pbattset or Qbattset<0 & tb≥tdb
(12)
where SOClow and SOCup are the lower and upper limitsof SOC, Pbattset and Qbattset are the reference active andreactive power output of battery. To increase the life span ofbattery storage without compromising the control objectives, athorough charging/discharging control scheme is applied. Theconcept of this control strategy is to introduce a hysteresisband and add an idle mode in addition to the charging anddischarging modes. In idle mode, battery is still connected tothe system but it is not charging or discharging unless theidle mode has been deactivated. The idle mode is activated toprotect battery life if one of the following criterions is met:
1) the reference power is within the deadband [−Pband,Pband];
2) the reference power exceeds the deadband upper limit,and the battery SOC violates SOC_low;
3) the reference power violates the deadband lower limit,and the battery SOC violates SOC_up.
However, the charging mode is activated based on (12), untilthe battery SOC reaches its upper limits SOC_up accordingto criterion 3. The battery may oscillate between charging andidle mode at this operation point. To avoid excessive chargingor discharging frequently, the battery is charged fully unlessdischarging mode is activated.
V. SIMULATION STUDY
A comprehensive simulation study was carried out inPSCAD/EMTDC to verify the performance of the MPC-basedcoordinated controller. A 2-MW direct drive wind turbineconsisting of a PMSG and an ac-dc-ac converter is connectedto the system through LR filter as shown in Fig. 1. The PMSGhas 84 poles with a base angular frequency of 117.93 rad/s.The simulation time step is chosen as 0.0001 s. A 0.5-MWbattery is also connected to the 0.48 kV PCC bus. The totalload of the system is 1.5 MW. Several scenarios are simulatedand the results are presented in this section.
Fig. 8. Comparison between MPC and optimized PI controller.
A. Comparison of MPC Controller With Industry Standard
The performance of MPC is compared with the PI controllerwith optimized parameters. The optimized PI parameters areobtained by a genetic algorithm. An actual 3 s wind speed pro-file is used as shown in Fig. 8. P and Q are the active and reac-tive power of VSI-1, and Q1 is the reactive power of VSI-2.The subscripts mpc, pi, and ref denote MPC, PI controller,and reference, respectively. Fig. 8 shows that both the PIcontroller as well as MPC controller maintains 0 reactivepower exchange between wind turbine and VSI-1 and betweengrid and VSI-2.
From Fig. 8, it is seen that the active power output ofwind generator can track the wind speed variations with theproposed MPC and the optimal PI controller. The reactivepower output and dc link voltage are fluctuations are within±0.001 MVAR for reactive power and ±0.02 p.u. for dc linkvoltage. The maximum overshoot of reactive power is less than0.003 MVAR, and the overshoot of 1.09 p.u. in dc link voltageat 0.35 s is caused by the excessive wind speed. The rotorside converter controller maintains the active power output at2 MW when the wind speed exceeds the rated value. It canbe seen from the figures that MPC has faster response, lessovershoots and less steady-state error in dc bus voltage, activeand reactive power. The tracking error in Fig. 7 is caused bythe current limiter in wind converter to prevent the converterfrom overloading. The rating of wind converter is 2 MWin this paper. The active power reference is obtained from(6) and (7), and the value can be higher than 2 MW when thewind speed exceeds the rated value. Under such condition,the current of wind converter is limited by the current limiterto prevent the overloading of wind converter. Thus, the rotor
302 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 1, NO. 4, DECEMBER 2013
Fig. 9. System performance under low voltage in grid-connected mode.
Fig. 10. System performance under high voltage in grid-connected mode.
side converter controller maintains the active power output at2 MW when the wind speed exceeds the rated value.
B. Reactive Power Coordination Under Grid-ConnectedOperation
The four-level protection capability of the proposed reactivepower coordination under grid-connected operation is evalu-ated under severe voltage variations owing to large inductiveor capacitive load changes. The wind speed is maintained asconstant. Fig. 9 shows the PCC bus voltage, real and reactivepower, the operation mode (0: 0-RPC, 1: voltage control), andthe switch status of dynamic load (0: close, 1: open). Thesubscripts wind, batt, t, wind + batt, and s denote wind farm,
Fig. 11. System performance under wind speed variation.
Fig. 12. Performance of the generator and grid side MPC.
battery, total load, wind and battery, and system, respectively.Initially, the wind system is operating in wind dominant modewith Vrms as 0.48 kV. A 2 MVAR load step change at1 s results in Vrms violating the lower voltage limit andan increase in reactive power output of wind turbine from0.2 MVAR to its maximum capacity of 0.88 MVAR. Thesystem transitions to wind and battery mode at 1.5 s, whichimproves the voltage profile but cannot remove violations eventhough battery reaches its maximum reactive power output.The system transitions to shedding mode where the dynamicload is disconnected to restore Vrms to 0.48 kV.
Similarly, when a load step at 1 s results in Vrms violat-ing higher voltage limit, the reactive power output of windfarm decrease to −1 MVAR as shown in Fig. 10. Sincethe bus voltage is not within limit in the wind dominantmode even under wind farm generating maximum reactivepower output, the system transitions to wind and batterymode at 1.5 s which improves voltage profile but cannotremove violations even though battery reaches its maxi-mum reactive power output. The system transitions to stalloperation mode at 2.2 s, where the blade angle gradually
HAN et al.: COORDINATED PREDICTIVE CONTROL OF A WIND/BATTERY MICROGRID SYSTEM 303
Fig. 13. System response to operation mode change.
Fig. 14. System performance during the transition from grid-connected toislanded mode.
increases which results in decrease in Pwind and increasein Qwind from −1 to −1.5 MVAR and restoring Vrms to0.48 kV.
C. Active Power Coordination Under Grid-ConnectedOperation
The active power output of wind farm changes as shown inFig. 11, which changes in wind speed as shown in Fig. 8. Windand battery maintain power balance in the microgrid withno power exchange with the main power system. However,under low wind speed at 2 s the battery is at its maximumactive power generation. The coordinated controller adjust thereactive power output of wind farm to maintain the PCC bus
Fig. 15. System response during the transition from grid-connected toislanded mode.
Fig. 16. System performance during the transition from grid-connected toislanded mode.
voltage at 0.48 kV, and the power mismatch is compensated bythe main grid. The proposed coordinated control also providesactive output power control that maintains system frequency.
Fig. 12 shows currents ia1 and ia2 of grid side and generatorside MPC, respectively. It is seen that ia1 and ia2 trackthe reference current ia1ref and ia2ref while interconnectingthe low-varying frequency system on generator side with the60 Hz electric power system.
D. Active and Reactive Power Coordination Under TransitionFrom Grid-Connected to Islanded Operation
The PCC bus voltage and wind converter output currentwhen the microgrid is disconnected from main grid at 1.5 s isas shown in Fig. 13. There is no phase shift in the PCC busvoltage Va2 during the islanding. However, the current of VSIoscillates for one cycle.
As seen from Fig. 14 the island power balance is disturbedwhen there is a large wind speed drop at 2.4 s. Some of
304 IEEE JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, VOL. 1, NO. 4, DECEMBER 2013
Fig. 17. System performance during black start in islanded operation.
the dynamic load is disconnected at 2.6 s and battery adjustsreactive power output to compensate for voltage oscillationdue to active power output variations of wind converter tomaintain Vrms at 0.48 kV. The system frequency is maintainedat 60 Hz.
E. Active and Reactive Power Coordination Under TransitionFrom Islanded to Grid-Connected Operation
The system performance from islanded to grid-connectedoperation is shown in Figs. 15 and 16. As seen from thefigures, there is no phase shift in the PCC bus voltage duringthe transition from islanded to grid-connected mode. Themicrogrid is able to provide the local load before and after thetransition, and the PCC bus voltage is maintained at 0.48 kV.
F. Reactive and Active Power Coordination Under BlackStart in Islanded Operation
Under black start, the average wind energy calculated bywind forecasting combine with the maximum battery capacitydetermines the load that can be supplied. Fig. 17 shows thesystem performance under black start with 1.2 MW loadconnected. Wind farm is connected at 0.3 s and battery isconnected 0.1 s after. It is seen that the wind farm reachessteady state at 0.6 s, and the PCC bus voltage Vrms ismaintained at 0.48 kV.
VI. CONCLUSION
This paper presents a MPC-based supervisory control sys-tem that coordinates the operation of wind farm and batterystorage system in a microgrid for grid-connected and islandedoperations. The MPC increases the accuracy of maximumwind energy capture as well as minimizes the power oscil-lations caused by varying wind speed. The proposed coor-dinated controller mitigates both active and reactive power
disturbances that are caused by the intermittencies in windspeed and load change. Thus the control strategy adds an extradegree of flexibility to the microgrid by providing four levelsof protection, and maintaining power balance.
REFERENCES
[1] Y. Zhang, “Study on operation characteristics of grid-connected doublyfed induction generators,” M.S. thesis, Dept. Electr. Eng., North ChinaElectric Power Univ., Baoding, China, 2008.
[2] M. R. Patel, Wind and Solar Power Systems. Boca Raton, FL, USA:CRC Press, 1999, pp. 82–83.
[3] S.-K. Kim and E.-S. Kim, “PSCAD/EMTDC-based modeling and analy-sis of a gearless variable speed wind turbine,” IEEE Trans. EnergyConvers., vol. 22, no. 2, pp. 421–430, Jun. 2007.
[4] R. M. Hilloowala and A. M. Sharaf, “A rule-based fuzzy logic controllerfor a PWM inverter in a standalone wind energy conversion scheme,”IEEE Trans. Ind. Appl., vol. 32, no. 1, pp. 57–65, Jan./Feb. 1996.
[5] M. H. Ali, M. Park, I.-K. Yu, T. Murata, and J. Tamura, “Improvement ofwind-generator stability by fuzzy-logic-controlled SMES,” IEEE Trans.Ind. Appl., vol. 45, no. 3, pp. 1045–1051, May/Jun. 2009.
[6] M. El Mokadem, V. Courtecuisse, C. Saudemont, B. Robyns, andJ. Deuse, “Fuzzy logic supervisor-based primary frequency controlexperiments of a variable-speed wind generator,” IEEE Trans. PowerSyst., vol. 24, no. 1, pp. 407–417, Feb. 2009.
[7] G. N. Kariniotakis, G. S. Stavrakakis, and E. F. Nogaret, “Windpower forecasting using advanced neural networks models,” IEEE Trans.Energy Convers., vol. 11, no. 4, pp. 762–767, Dec. 1996.
[8] H. Li, D. Zhang, and S. Y. Foo, “A stochastic digital implemen-tation of a neural network controller for small wind turbine sys-tems,” IEEE Trans. Power Electron., vol. 21, no. 5, pp. 1502–1507,Sep. 2006.
[9] S. Li, D. C. Wunsch, E. A. O’Hair, and M. G. Giesselmann, “Usingneural networks to estimate wind turbine power generation,” IEEE Trans.Energy Convers., vol. 16, no. 3, pp. 276–282, Sep. 2001.
[10] R. Lasseter and M. Erichson, “Integration of battery-based energy stor-age element in the CERTS microgrid,” in Proc. IEEE Energy Convers.Congr. Exposit., Sep. 2010, pp. 2570–2577.
[11] S.-K. Kim, J.-H. Jeon, C.-H. Cho, J.-B. Ahn, and S.-H. Kwon, “Dynamicmodeling and control of a grid-connected hybrid generation system withversatile power transfer,” IEEE Trans. Ind. Electron., vol. 55, no. 4,pp. 1677–1688, Apr. 2008.
[12] J.-H. Jeon, J.-Y. Kim, H.-M. Kim, S.-K. Kim, C. Cho, J.-M. Kim,J.-B. Ahn, and K.-Y. Nam, “Development of hardware in-the-loopsimulation system for testing operation and control functions of micro-grid,” IEEE Trans. Power Electron., vol. 25, no. 12, pp. 2919–2929,Dec. 2010.
[13] J.-Y. Kim, J.-H. Jeon, S.-K. Kim, C. Cho, J. H. Park, H.-M. Kim, andK.-Y. Nam, “Cooperative control strategy of energy storage system andmicrosources for stabilizing the microgrid during islanded operation,”IEEE Trans. Power Electron., vol. 25, no. 12, pp. 3037–3048, Dec. 2010.
[14] M. Datta, T. Senjyu, A. Yona, T. Funabashi, and C.-H. Kim,“A frequency-control approach by photovoltaic generator in a PV-dieselhybrid power system,” IEEE Trans. Energy Convers., vol. 26, no. 2,pp. 559–571, Jun. 2011.
[15] J. Qin and M. Saeedifard, “Predictive control of a modular multilevelconverter for a back-to-back HVDC system,” IEEE Trans. Power Del.,vol. 27, no. 3, pp. 1538–1547, Jul. 2012.
[16] PSCAD/EMTDC Power System Simulation Software User’s Manual,Version 3, Manitoba HVDC Research Center, Winnipeg, MB, Canada,1998.
[17] O. Anaya-Lara and E. Acha, “Modeling and analysis of custom powersystems by PSCAD/EMTDC,” IEEE Trans. Power Del., vol. 17, no. 1,pp. 266–272, Jan. 2002.
[18] J. G. Slootweg, S. W. H. de Haan, H. Polinder, and W. L. Kling,“General model for representing variable speed wind turbines in powersystem dynamics simulations,” IEEE Trans. Power Syst., vol. 18, no. 1,pp. 144–151, Feb. 2003.
[19] S. Morimoto, H. Nakayama, M. Sanada, and Y. Takeda, “Sensorlessoutput maximization control for variable-speed wind generation systemusing IPMSG,” IEEE Trans. Ind. Appl., vol. 41, no. 1, pp. 60–67,Jan./Feb. 2005.
HAN et al.: COORDINATED PREDICTIVE CONTROL OF A WIND/BATTERY MICROGRID SYSTEM 305
[20] W. Qiao, L. Qu, and R. G. Harley, “Control of IPM synchronousgenerator for maximum wind power generation considering magneticsaturation,” IEEE Trans. Ind. Appl., vol. 45, no. 3, pp. 1095–1105,May/Jun. 2009.
[21] S.-K. Chung, “A phase tracking system for three phase utility interfaceinverters,” IEEE Trans. Power Electron., vol. 15, no. 3, pp. 431–438,May 2000.
[22] P. Cortes, S. Kouro, B. L. Rocca, R. Vargas, J. Rodriguez, J. I. Leon,S. Vazquez, and L. G. Franquelo, “Guidelines for weighting factorsdesign in model predictive control of power converters and drives,” inProc. IEEE Int. Conf. Ind. Technol., Feb. 2009, pp. 1–7.
[23] M. C. Chandorkar, D. M. Divan, and R. Adapa, “Control of parallelconnected inverters in standalone AC supply systems,” IEEE Trans. Ind.Appl., vol. 29, no. 1, pp. 136–143, Jan./Feb. 1993.
Junbiao Han (S’11) received the B.S. degree inelectrical engineering from Zhengzhou University,Zhengzhou, China, in 2004, and the M.S. degreein electrical engineering from North China ElectricPower University, Beijing, China, in 2008. He iscurrently working toward the Ph.D. degree with WestVirginia University, Morgantown, WV, USA.
He was a Research and Development Engineerwith RXPE, Liaoning, China, from 2007 to 2009,on industrial application of power electronic devicesuch as SVC and STATCOM. His current research
interests include the application of power electronic device in renewableenergy such as solar energy and wind energy.
Sarika Khushalani Solanki (M’09) received B.E.degree in electrical engineering from Nagpur Uni-versity, Nagpur, India, the M.E. degree in electri-cal engineering from Mumbai University, Mumbai,India, in 1998 and 2000, respectively, and the Ph.D.degree in electrical and computer engineering fromMississippi State University, Starkville, MS, USA,in 2006.
She has been an Assistant Professor with the LaneDepartment of Computer Science and ElectricalEngineering, West Virginia University, Morgantown,
WV, USA, since August 2009. She was with Open Systems International, Inc.,Minneapolis, MN, USA, as a Senior Engineer. Her current research interestsinclude smart grid, power distribution system, computer applications in powersystem analysis, and power system control.
Jignesh Solanki (M’02) received the B.E. degreefrom the Visvesvaraya National Institute of Tech-nology, Nagpur, India, and the M.E. degree fromMumbai University, Mumbai, India, in 1998 and2000, respectively, and the Ph.D. degree in electricaland computer engineering from Mississippi StateUniversity, Starkville, MS, USA, in 2006.
He was involved in research activities at IndianInstitute of Technology Bombay, Mumbai. He hasbeen a Research Assistant Professor with the LaneDepartment of Computer Science and Electrical
Engineering, West Virginia University, Morgantown, WV, USA, since August2009. He was with Open Systems International, Inc., Minneapolis, MN,USA, as a Senior Engineer. His current research interests include smart grid,multiagent applications in power system, and power system control.
Dr. Solanki received the IEEE Multiagent Systems Working Group Awardin 2008.
