Smart Power Management

Energy Conversion and Management 84 (2014) 562–582

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier .com/locate /enconman

Smart power management algorithm in microgrid consistingof photovoltaic, diesel, and battery storage plants considering variationsin sunlight, temperature, and load

http://dx.doi.org/10.1016/j.enconman.2014.04.0720196-8904/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Electrical Engineering, Faculty ofEngineering, University of Malaya, 50603 Kuala Lumpur, Malaysia. Tel.: +60 379675238; fax: +60 3 79675316.

E-mail addresses: [email protected] (S. Koohi-Kamali), [email protected](H. Mokhlis).

Sam Koohi-Kamali a,b, N.A. Rahim a,c, H. Mokhlis a,b,⇑a UM Power Energy Dedicated Advanced Centre (UMPEDAC), Level 4, Wisma R&D UM, Jalan Pantai Baharu, 59990 Kuala Lumpur, Malaysiab Department of Electrical Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysiac Renewable Energy Research Group, King Abdulaziz University, Jeddah 21589, Saudi Arabia

a r t i c l e i n f o

Article history:Received 12 February 2014Accepted 24 April 2014Available online 20 May 2014

Keywords:PhotovoltaicSmart gridBattery energy storagePower smoothingRenewable energyEnergy management system

a b s t r a c t

Integration of utility scaled solar electricity generator into power networks can negatively affect theperformance of next generation smartgrid. Rapidly changing output power of this kind is unpredictableand thus one solution is to mitigate it by short-term to mid-term electrical storage systems like battery.The main objective of this paper is to propose a power management system (PMS) which is capable ofsuppressing these adverse impacts on the main supply. A smart microgrid (MG) including diesel, batterystorage, and solar plants has been suggested for this purpose. MG is able to supply its local load based onoperator decision and decline the power oscillations caused by solar system together with variable loads.A guideline algorithm is also proposed which helps to precisely design the battery plant. A novel appli-cation of time domain signal processing approach to filter oscillating output power of the solar plant ispresented as well. In this case, a power smoothing index (PSI) is formulated, which considers both loadand generation, and used to dispatch the battery plant. A droop reference estimator to schedule genera-tion is also introduced where diesel plant can share the local load with grid. A current control algorithm isdesigned as well which adjusts for PSI to ensure battery current magnitude is allowable. MG along withits communication platform and PMS are simulated using PSCAD software. PMS is tested under differentscenarios using real load profiles and environmental data in Malaysia to verify the operational abilities ofproposed MG. The results indicate that PMS can effectively control the MG satisfying both operator anddemand sides.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Public needs in modern societies beside optimal consumptionand/or generation of electricity necessitate the integration of intelli-gent power management systems (PMSs) into power networks. Thismatter has brought a new concept which is so-called ‘‘Smartgrid’’.Smartgrid incorporates advanced measurement technologies, con-trol algorithms, and communication platforms into present powergrid. These features are helpful to optimize the utilization of renew-able energy (RE) prime movers which contribute in the generation ofelectricity in large scales [1,2]. A combination of distributed storage(DS), RE distributed generation (DG) systems and loads which can

operate in parallel with the grid or in autonomous modes is so-called‘‘Microgrid’’. Microgrid (MG) can be considered as a cluster of loadand generation in smartgrid that brings many advantages for thesystem. The benefits can be pointed out i.e. increasing RE sourcesdepth of penetration, decreasing environmental emissions, utilizingwaste heat, providing ancillary services, making the balancebetween generation and consumption, and bringing continuousbackup power supply for redundant and sensitive processes [3].Renewable resources such as wind and solar photovoltaic (PV) arenaturally intermittent and hence energy storage systems (ESSs) likebattery can be exploited together with them to compensate for thisdrawback [4]. Solar PV plant in high penetration levels can modifythe load profile and create technical challenges for the system insteady-state and transient operating modes. The fluctuating outputpower is one example brought to this end [5,6]. Ramp ups/downs insolar plant output power are completely unpredictable. These fluc-tuations can be governed by several factors i.e. passing clouds, PV

http://crossmark.crossref.org/dialog/?doi=10.1016/j.enconman.2014.04.072&domain=pdf

http://dx.doi.org/10.1016/j.enconman.2014.04.072

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.enconman.2014.04.072

http://www.sciencedirect.com/science/journal/01968904

http://www.elsevier.com/locate/enconman

S. Koohi-Kamali et al. / Energy Conversion and Management 84 (2014) 562–582 563

power plant placement, depth of penetration, and power networktopology. In the case that these oscillations are out of control, mainAC network equipments such as motors, generators, and voltage reg-ulating devices can be affected adversely [4]. This oscillatory naturemay create voltage and angle instability in the main grid to whichDG is connected especially where DG depth of penetration is high.There have been suggested many methods in the literature toresolve this problem [7–10]. For instance, in [7], generation curtail-ment, dump load usage, and using ESS have been recommended asremedies for this purpose. However, the role of load variation basedon a real pattern has not been studied in this research. The use of ESShas become very popular recently [8,9] in collaboration with fuel cell[10] in order to suppress addressed oscillations. In this case, batteryenergy storage (BES) has been mostly proposed by the researchersfor short-term to mid-term applications [6,8,9].

Batteries are expensive equipments and thus adoptingelaborated control strategies in order to efficiently exploit them ismandatory. A conventional inertial filter has been utilized in [11]to smooth output power of a wind farm and make a reference valuefor current controlled inverter of BES. In [6,12], state of charge (SOC)feedback controllers for BES have been suggested which limit bat-tery charging/discharging currents within the acceptable range.However, finding an accurate time constant for SOC by these twomethods is not a straightforward task and hence highly dependson assumptions made by the planner which may not be the sameat all the times. In [6], genetic algorithm (GA) has been also usedto optimize the control parameters for solar plant which mayincrease the computational burden instead. In all above mentionedcases, there has not been considered the role of rotary based DGsystems such as diesel generator power plant in conjunction withthe intermittent RE sources where the system supplies loads basedon actual profiles. The lack of a unified power management system(PMS) which can govern a MG including RE sources together withthe conventional DGs is completely obvious as well. In [13,14],coordinated energy management algorithms have been proposedwhich can control a MG consisting of only electronically interfaced(EI) DGs in grid-connected and stand-alone modes. However, therole of rotary based DG has not been investigated and smoothingpower fluctuations of solar plant together with a real load profilehave not been addressed in these works.

In this paper, MG is investigated which consists of both rotary(diesel power plant) and EI based DGs. A power smoothing indexhas been formulated to mitigate the fluctuations resulted by inter-mittent solar PV system together with the variable load. In thiscase, moving average filtering (MAF) which is a time domain signalprocessing approach is utilized to smooth these oscillations. Thisindex can be applied for any kind of intermittent RE sources sinceit is easy to implement and has not relied on complicated compu-tational methods. A load model is also proposed which can besuitable for real-time applications where the actual load profilehas to be simulated for dynamic studies. A current control algo-rithm is designed as well to ensure the battery charge/dischargecurrent is within the specified limitation. This work also suggestsa guideline algorithm for the purpose of the battery house sizingtaking into account the ramp rate limits of the main network. Inaddition, a power management algorithm (PMA) is suggestedwhich helps the system owner to exploit the battery plant in themost efficient way. The concept of agent is included in this algo-rithm to define different level of hierarchy where a communicationchannel acts as the platform to exchange information between theoperator, DGs, and loads. To dispatch diesel and BES plants a newapplication of droop control mechanism is introduced whichmakes it possible for the operator to schedule the generation unitsfor both active and reactive powers. A droop mechanism for dieselplant excitation controller is also designed so that it can share thelocal reactive power with BES proportional to their ratings.

In what follows, Section 2 presents the proposed MG configura-tion. Section 3 describes the MG components dynamic models andcontrollers. Section 4 explains the data input preparationapproaches for generation units and loads. Section 5 analyses theproposed PMA and highlights the role of MAF in smoothing outthe aforementioned fluctuations. In Section 6, simulation resultshave been brought to the readers and technical matters have beeninvestigated in depth. Section 7 concludes this work.

2. System configuration and operation

The proposed microgrid (MG) incorporates both rotary andelectronically interfaced distributed generation (EIDG) systems(see Fig. 1). MG is subject to operate in grid-connected (G.C) mode.The primary source of power is a diesel engine which provides themechanical torque required for a 1.28 MV A synchronous genera-tor. Another DG unit is a 1080 kV A (1026 kW h, 1125 A h) batteryenergy storage system (BESS) which consists of a Lead-acid batterybank connected to the grid through a three-phase bi-directionalvoltage sourced converter (VSC). BESS is capable of operating aseither source or sink of power. As a demand, a six steps AC loadin two categories (i.e. industrial and domestic loads) are intercon-nected to the load bus and each group consists of three similarfeeders. The demand is supposed to vary during 24 h. A 1 MWp

photovoltaic distributed generation (PVDG) plant is also consid-ered to inject the available power from the sun into the MG inunity P.f during the whole day.

BESS is dispatched to smooth the power fluctuations in systemcaused by solar plant together with loads and hence it reduces theramp rate stresses on the main AC network. Diesel plant is dis-patched to shift up or down the grid active power profile and thusshares the load active power with AC network.

Depending on the power management strategy, BESS operateseither in inverting or rectifying modes. Diesel plant is assumed todecrease or increase its active power generation according toPMS commands. BESS and diesel plant can be dispatched in orderto share the load reactive power proportional to their ratings.

As shown in Fig. 1, there exist four agents in MG, namely, unitagents, generation agent, load agent, and main agent [15]. Genera-tion agent is assigned to receive and/or send the data from/to DGunit agents through the communication channel (bus) indicatedby dotted black line. Each DG unit agent collects the local informa-tion such as DG breaker status, output voltage and current, andavailability of prime mover. There is a forecasting module embed-ded in PV unit agent and along with this module the estimated solarpattern is sent to the main agent to be filtered yielding averagedsolar irradiation profile. In BESS, unit agent calculates SOC of batteryand sends it to generation agent. All these information is gathered ingeneration agent and then sent to the main agent. Local agents alsogenerate and compute the feedback signals required for the internalcontrollers (red dotted lines) such as current or power loops, gover-nor, and excitation controllers. Load agent also registers the statusof load breakers and the power which is flown in each feedertogether with the forecasted load profile. These data are sent tothe main agent (which is in the highest level of hierarchy) to informthe operator about the system states (dotted blue lines). Then theoperator calculates the dispatching references and issues requiredcommands for the DGs and loads breakers.

3. System components and controllers

3.1. Voltage sourced converter (VSC)

Two-level three-leg converter topology has been utilized in thiswork. This topology consists of six insulated gate bipolar transistor

Fig. 1. Smart microgrid configuration.

564 S. Koohi-Kamali et al. / Energy Conversion and Management 84 (2014) 562–582

(IGBT) semiconductor switches. There exist two IGBTs in each legand hence they supposed to be switched on/off in a complemen-tary manner to prevent the short circuit at the VSC terminal. Thegate firing signals are generated through pulse width modulation(PWM) technique which benefits from high frequency modulationsaw-tooth waveform crossing the reference signal oscillating atfundamental frequency (50 Hz). To prevent the propagation of highswitching frequency harmonics in the main network, VSC is con-nected to point of common coupling (PCC) through a LC filterand delta-star step-up isolation transformer. LC filter includes inequivalent inverter line inductance in parallel with a capacitorbank (design value). Bandwidth of LC filter should be lower thanthe VSC controller bandwidth to make sure that the noiselessstates are fed back into the controller. Thus, the control loopswould be in charge of attenuating the low frequency disturbanceswhich pass through the filter.

3.1.1. Phase locked loop (PLL)To measure the frequency of a measured AC signal, PLL mecha-

nism can be used. The quadrature component of the input signal(Vq) into PLL controller is compared to zero and hence the errorpasses through a PI regulator whose output would be consideredas the input for voltage-controlled oscillator (VCO) unit (seeFig. 2). This method of PLL implementation force Vq to become zeroand thus the direct component of signal (Vd) is aligned with the ref-erence phasor rotating at the same frequency. Another method hasbeen presented in [16] in which the difference between the PLLoutput angle and its input reference angle can turn into zero if itis limited to between zero and �180�.

3.1.2. Current controlThe former control technique which is mainly utilized in the

presence of a master grid is so-called ‘‘current control’’ scheme.

In current control mode, the flow of power can be in two directionsi.e. from the VSC to the master network or vice versa. The masterunit or network is assumed to be robust enough to provide con-stant frequency and voltage at PCC and thus VSC can simply followthe reference phasor using the phase PLL controller [17]. To modelcurrent controller, states of the system should be measured andtransformed to dq0 rotating synchronous reference. Park’s trans-formation is applied for this purpose as shown in Fig. 2. The angleof this conversion is generated through the PLL controller whichsimultaneously oscillates along with the instantaneous voltagephasor measured at PCC.

As shown in Fig. 3, current controller is composed by a fastinner current loop which tracks the reference current value (Iref).Direct and quadrature components of reference current can be cal-culated either through Eqs. (1) and (2) or can be generated using anouter slower power loop in P–Q control mode (see Fig. 4). The outerpower loop follows the given power references which are eithernegative (in rectifying mode) or positive (in inverting mode).

Id;ref ¼2

3VdPref ð1Þ

Iq;ref ¼ �2

3VdQ ref þ Cf 0Vd ð2Þ

where Vd is the direct component of VSC output voltage (in kV), Cf isthe capacitor value in LC filter (in F), Pref and Qref are the desired set-points for active and reactive powers (in MW and MVAR),respectively.

3.2. Solar PV generator dynamic model

PV generator should be modeled accurately because the dynam-ics of VSC and controllers highly depend on the PV model. For the

Isc

Id

Rsh

Rs

IIsh

-

+

V

Fig. 5. Equivalent single-diode model of solar PV cell.

++

-2 /3

++

2 /3

SIN Va

SIN Vb

SIN Vc

+++ Vq

Vq,ref = 0.0

PI

VCO

++

-2 /3

++

2 /3

COS Va

COS Vb

COS Vc

+++ Vd

Park’s Transformation

Va

Vb

Vc

freq

2/3

2/3

-

+ ∫PLLθ

Fig. 2. PLL controller used in this work.

Fig. 3. Inner current loop regulator.

--

+ Id,ref

--+ Iq,ref

PI

PI

Fig. 4. Outer power loop regulator in P–Q control mode.


circuitry based modeling techniques of PV cell, single-diode, dou-ble-diode, and three-diode models have been suggested in the lit-erature. The single-diode model can be further improved by addinga second diode in parallel with the first diode. The second dioderepresents the recombination effect of carriers in depletion regionof semiconductor where current value of PV cell is low [18]. In

three-diode model, a third diode has been included in the modelto represent the current flows through the peripheries [19]. How-ever, single-diode model has shown a reasonable trade-offbetween accuracy and simplicity. This model is generally utilizedin power system studies since determination of parametersbecomes slightly complicated for double-diode and three-diodemodels. In single-diode model, a current source is anti-parallelwith a diode. As depicted in Fig. 5, shunt and series resistancesare also considered in this equivalent circuit.

When the PV cell is illuminated by the sun, it generates DCphoto-current (Isc). Photo-current varies linearly against changesin solar irradiance. The current through the anti-parallel diode(Id) is responsible for non-linearity of I–V characteristic. TheKirchhoff’s current law, based on equivalent circuit shown inFig. 5, yields [20]:

I ¼ Isc � Id � Ish ð3Þ

and hence,

I ¼ Isc � I0 expV þ IRs

nkTc=q

� �� 1

� �� V þ IRs

Rsh

� �ð4Þ

Isc can be calculated as:

Isc ¼ IscRGGR

1þ aT Tc � TCRð Þ½ � ð5Þ

whereGR: reference solar radiationTCR: reference cell temperatureIscR: short circuit current at GR and TCR

aT: temperature coefficient of photo currentand I0 is dark current and can be calculated as:

: ( ), ( )Inputs V t I t

( ) ( )( ) ( )

dI I t I t dtdV V t V t dt

= − −= − −

0dV =

dI dV I V= − = 0dI

> 0dI

Increase Vref

Increase Vref

Decrease Vref

Decrease Vref

dI dV I V> −Yes

Yes

Yes

Yes

Yes

No

No No

No No

( ) ( )( ) ( )

I t dt I tV t dt V t

− =− =

Return

Fig. 6. IC algorithm in order to the estimate Vref for DC bus controller.


I0 ¼ IoRT3

c

T3cR

!exp

1TcR� 1

Tc

� �� qeg

nkð6Þ

whereIoR: dark current at TCR

q: electron chargek: Boltzman constanteg: band gap energyn: diode ideality factor (between 1 and 2, typical value forsilicon solar cell is 1.3).

The basic unit of PV generator is solar cell which is able to gener-ate electrical power about 1–2 W. Series and/or parallel electricallycoupled PV cells make PV modules and further PV arrays. A PV arrayencompasses series and parallel connected modules and hence thesingle cell equivalent circuit can be scaled up in order to rearrangefor any series/parallel configuration. Total number of 160 stringsthat each consists of 24 modules in series (Voc,Plant = 1221.6 V) havebeen connected in parallel (Isc,Plant = 891.2 A) to build up 1 MW PVpower plant. Parameters of PV module used in this paper are citedin Table 1.

3.2.1. Maximum power point tracking (MPPT) algorithmThe amount of power which can be captured from the solar cell

depends on the operating point on the I–V characteristic. To drawas much power as possible from the PV generator, different MPPTtechniques have been introduced in the literature [21–24]. MPPTalgorithm is designed to generate voltage reference value at DClink and hence keep the operating point about the knee point ofI–V curve. The main duty of MPPT algorithm is to match impedanceat the PV generator terminal where the load impedance varies andforce the operating point to move away the knee point. A com-monly used algorithm is Perturb and Observe (P&O) technique.However, this method has its own limitations. For example, theexact maximum power point (MPP) can never be found and hencethe power oscillates about MPP [20,25]. The method adopted inthis work is Incremental Conductance (IC) algorithm. Fig. 6 showsthe flowchart of this algorithm which has been implemented inPSCAD [24]. The IC algorithm is designed to evaluate Eq. (5) atthe MPP as:

dPdV¼ dðVIÞ

dV¼ I þ V

dIdV¼ 0 ð7Þ

where I and V are the output current and voltage at the terminal ofPV generator, respectively.

3.2.2. PVDG control mechanismAccording to [26,27], PVDG is not allowed to participate in volt-

age regulation and thus it can only operate in unity power factor.By forcing the reactive power to zero in steady-state, the DC busequation which shows relationship between input power (PPV)and output power (PVSC) can be written as:

ddt

V2dc ¼

2C

PPV � PVSCð Þ ð8Þ

Table 1Electrical specifications of solar module (HIT-N210A01).

Parameter Value

Rated Power (Pmax) 210 WMaximum Power Voltage (Vpm) 41.3 VMaximum Power Current (Ipm) 5.09 AOpen Circuit Voltage (Voc) 5.09 VShort Circuit Current (Isc) 5.57 ATemperature Coefficient (Voc) �0.142 V/CTemperature Coefficient (Isc) 1.95 mA/CNOCT (Normal Operating Cell Temperature) 46 �C

PVDG system is assumed to operate in current control mode at allthe times. To regulate the voltage at DC bus, there are two tech-niques i.e. single-stage and double-stage methods. In this work, sin-gle-stage controller has been utilized through which only theinverter is controlled. There exists no interfacing unit between thePV array and inverter except DC link. In single-stage mechanism,whole power generated in DC side by PV prime mover is instanta-neously transferred to AC side and hence the voltage at DC link getsregulated. MPPT sets the voltage reference. As illustrated in Fig. 7,the error between the square of set-point and square of measuredDC bus voltage passes through a PI compensator. PI regulator out-put is added to instantaneous power generated by PV array makingup Pref for the inner power loop. Assuming that PVSC to be the inputpower for the inverter and if the switching losses is ignored, the V2

dc

and PPV would be the controller inputs and hence Pref would be thecontroller output. PVSC would become equal to Pref in steady-state sothat the variation of DC link voltage would be forced to become zeroand the feed-forward path including the PPV ensures that the wholepower is transferred to the AC side.

However, in double-stage approach a DC–DC converter isexploited which regulates for voltage at DC bus at desired valueby tracking the MPPT reference. DC–DC converter should be

PI--+

PPV

+++ Pref2

dcV

2refV

Fig. 7. Outer power loop in PVDG controller.


controlled separately and so one control level has to be added tothe controller with additional filtering and switching hardware.

3.3. Battery energy storage system (BESS)

3.3.1. Battery design approachIn this paper, a design guideline has been proposed to accu-

rately size the battery storage plant taking into account the techni-cal and empirical aspects. In conventional rotary based generators,the high inertia of rotating mass is capable of supplying for anytransient power mismatch between the load and generation. Bycontrast, in low or non-inertial systems (e.g. a PVDG system) anychanges in the load or generation results in an abrupt power oscil-lation. To compensate for this drawback, BESS can be exploited inpower networks [4]. BESS also can bring ancillary services for ACnetwork such as dispatching ability, ride-through capability, andnetwork stability.

In this work, BESS is directly connected to the DC-link of VSC. Toprevent battery electrolyte decomposition, state of charge (SOC) isconsidered as the operating control variable in the power manage-ment system (PMS). To control the charge/discharge current PMSlimits the dispatching reference (Pref) through the proposed algo-rithm shown in Fig. 8. This current control strategy is embeddedin BESS unit agent and confines dispatching reference (smoothingpower index) of BESS to make sure that the charge/discharge cur-rents are controllable accordingly. Since the MG is in grid-con-nected mode any mismatch between the smoothing index andBESS power limits would be supplied or absorbed by the grid. Inthis algorithm, the PVDG and load forecasting modules releasetheir estimated waveforms and then smoothing index is calculatedby the main agent accordingly. It sends the dispatching signal toBESS through the generation agent. Once the unit agent receivesthe dispatching commands from the generation agent, the currentcontrol algorithm is activated and determines whether the charge/discharge current is within the allowed boundary or not. In anothermethod, the storage unit is connected to DC link via DC–DC con-verter. DC–DC converter is a bidirectional buck-boost converterthat has two duties. Firstly, it boosts up battery terminal voltageto that level is required at DC bus. Secondly, it contributes in charg-ing/discharging process of battery in constant current (CC) or con-stant voltage (CV) modes. To control DC–DC converter additionalelaborated control mechanisms together with filtering and switch-ing elements have to be considered.

+

PV FORECASTING

MODULE

MOVING AVERAGE

FILTER

-

Pset

23 dV

,maxref DischI I< ,maxref CharI I<

LOAD FORECASTING

MODULE +

PΔ

refP P= Δ ,maxref CharP P=

Yes

No

refP P= Δ,maxref DicharP P=

Yes

No

-

PPV

PLavg

PL

Fig. 8. Proposed battery charge/discharge current controller.

For the utility applications, lead-acid battery is a proper solu-tion in terms of technical aspects and its cost per kilo watt [28].Lead-acid battery (LAB) is able to meet the ramp rate requirementsof the grid. LAB can be used for deep-cycle applications where thepower is supplied for a long duration. LAB is also suitable to becharged or discharged in a short period of time (e.g. for powersmoothing purpose).

To determine the charge or discharge rate of a battery cell, C rateis defined. C rate specifies the amount of constant current multipliedby the duration (which is 5, 10, or 20 h) when the battery can con-tinuously supply for this current. It is nominally determined in man-ufacturers’ datasheets as C5, C10, or C20. The battery cell terminalvoltage (Vcell) strongly depends on the C rate which the battery issized for. Designing of a BESS for a typical microgrid is begun withcharacterization of battery cell which is the smallest unit in a bat-tery bank. Open-circuit terminal voltage of the cell in 100% stateof charge is considered as the float or nominal voltage (Voc). Finalvoltage at the completion zone of discharge is symbolized by Vfinal

for each cell. Keeping the Vcell above the Vfinal retains the operatingpoint in the linear zone. Therefore, Vfinal should be chosen as highas possible and normally 80% to 90% of Voc might be a reasonablevalue. Vfinal is specified in the manufacturers’ datasheets for highrate of discharge (e.g. 5 min in here) applications. The nearest valueof voltage available in the datasheet to the calculated Vfinal can beselected as the final voltage and used to size BESS (see Fig. 9).

Let VBESS to be the BESS terminal voltage after discharge. BESS issupposed to contain Ns battery cell in series to meet the requiredvoltage level at DC bus. The minimum DC bus voltage of VSC, inwhich the VSC can operate normally, is considered as VBESS andcan be formulated as:

VBESS ¼2ffiffiffi2p

VLL

mffiffiffi3p ð9Þ

where m is the PWM switching modulation index and VLL is RMSvalue of line-to-line VSC voltage at the grid side. Thus, the numberof cells in series in each string is given by:

Ns ¼VBESS

Vfinalð10Þ

If the calculated Ns is not an integer, the subsequent higher valuewould be selected as Ns. BESS should be able to absorb or supplythe maximum power in a short duration and at the same rate. Todesign it for maximum operating ramp rate, it is enough to findthe maximum difference between reference waveform estimatedby moving average processing unit and estimated solar plant outputpower added to difference between estimated load profile and aver-age load (smoothing index). The maximum ramp rate can be han-dled by BESS is found through (see Section 5):

PmaxBESS ¼ PL � PLavg

� �þ Pset � PPVð Þ

RðtÞ ð11Þ

where PL is the instantaneous load (in MW), PLavg is the average load(in MW), Pset is reference waveform generated by MAF (in MW), andPPV is the PVDG output power (in MW). R(t) is the unity exponentialor ramp function and reaches to its final value in T (s) and thus themaximum energy which can be absorbed or supplied by BESS isdetermined by:

EmaxBESS ¼

Z T

0Pmax

BESS tð Þdt ¼ PL � PLavg� �

þ Pset � PPVð Þ

� K � T ð12Þ

where

K ¼Z T

0R tð Þdt ð13Þ

K is equal to 0.5 and 0.2 where R(t) is either ramp or exponentialfunction, respectively.

Fig. 9. Proposed battery house sizing and designing algorithm.


Taking into account active power capability of BESS themaximum energy can be written as:

EmaxBESS ¼

K � SVSC � cos u � Tg

ð14Þ

where SVSC is VSC rated apparent power in VA and g is VSCefficiency which is between 0.8 and 0.9.

To find the charge and discharge rate of BESS, the C rate (A h) ofbattery plant should be found which is equal to:

CAh ¼Emax

BESS

VBESS:3600ð15Þ

and thus the charge/discharge rate of BESS is denoted as a C ratecoefficient given by:

D ¼ 3600T

ð16Þ

so the discharge current in ampere(s) is written as:

I ¼ D � CAh ð17Þ

The time taken for BESS to ramp up/down depends on the applica-tion which BESS is designed for. In G.C mode, BESS is responsible forsmoothing the PVDG power fluctuations together with the load.Smoothing index is considered to be the dispatching reference forbattery plant VSC and hence BESS should be sized based on this var-iable boundary of variation. The ramp rate of PVDG is very fast(because it is EIDG) and a sudden change in solar radiation levelresults in abrupt PVDG output variation. Load is also variable. In thiscase, generation and load forecasting can be proper remedies inorder to denote what would be the boundary of these fluctuationsin advance. In this work, the load and generation profiles areassumed to be forecasted one day in advance. For example, we con-sider a noise of ±10% about the PVDG average output power. If thePVDG is designed for 1 MWp, the peak time generation would varyfrom 0.9 to 1.1 MWp. So the BESS system should be capable of sup-plying or absorbing 100 kW within several seconds.

3.3.2. Battery dynamic modelShepherd model of lead-acid battery is used in this paper

[29,30]. To improve this model, the initial state of charge is substi-tuted in the equations to accurately consider how the battery levelof charge affects other parameters. The electrochemical behavior ofbattery is described in terms of current and voltage. This empiricalmodel is often used incorporating with Peukert equation in orderto obtain battery voltage and SOC as follow:

VT ¼ Voc � R � i� Ki1

SOCi � DODð18Þ

where VT is the battery terminal voltage (V), Voc is the battery open-circuit voltage, R is the internal resistance (ohm) calculated as:

R ¼ R0 þ KR1

SOCi � DODð19Þ

where R0 is residual resistance (ohm) calculated through:

R0 ¼ R0 þ KR ð20Þ

KR is electrolyte resistance at full charge and R0 is initial batteryresistance at full charge. i is the instantaneous battery current (A).Ki is polarization coefficient, and DOD is depth of discharge givenby:

DOD ¼ 1Q max

Zidt ð21Þ

where Qmax is the maximum nominal capacity of the battery. SOCi isinitial state of charge which can vary between 0.3 and 0.9 while thesimulation starts running. Instantaneous battery state of charge isformulated as:

SOC ¼ Q max � Q used

Q max¼ SOCi � DOD ð22Þ

where

Qused ¼Z

idt ð23Þ

As shown in Fig. 10, the equivalent circuit of battery has beenimplemented in PSCAD/EMTDC by a variable DC voltage sourcein series with a variable resistor. The amplitude of voltage sourceand resistor are determined through the Eqs. (18)–(23).

Technical specifications of lead-acid battery module used inBESS are cited in Table 2. The lowest final voltage value for 5 mindischarge is 9.6 V. To provide the voltage level of 912 V at the VSCDC link, 95 battery modules are estimated to be connected in seriesin each string. The maximum ramp rate considered for PVDG

+

-

R

+

-

VT− 1oc iV K

SOC

Fig. 10. Equivalent dynamic model of battery module.


together with the load are supposed to be 2 [�300, 350] kW/min(refer to Section 5). Each battery string can supply up to 75 A hwhich means that the current can be drawn within 5 min wouldbe 265.3 A and hence to supply 350 kW (384 A) power, two batterystrings are adequate in discharge mode. On the other hand, themaximum allowed charge current is 22.5 A for each string andthe BESS should be able to absorb 300 kW (329 A) in the matterof minutes. To ensure the charge current in within the boundary,15 battery strings have been allocated for this purpose. If the chargecurrent is above 329 A, the proposed current control algorithm lim-its the smoothing reference to this value.

3.3.3. BESS control mechanismBESS controller is designed to operate in grid-connected mode.

P–Q control mechanism is adopted which is a variety of currentcontrol of VSC. The references are generated by main agent in orderto dispatch the dispatchable DGs (BESS and diesel plant). BESS isresponsible for smoothing power oscillations due to the changesin solar radiation and load. A new dispatching index is formulatedin this paper which would be investigated in Section 4.

To employ the storage system efficiently, BESS must quicklyrespond within the acceptable duration. In this case, undesirableoperation of load tap changers (LTCs) and capacitor banks, whichcan be due to the power fluctuations of PVDG and the load, is min-imized and hence the AC network would be less stressed.

To make sure that the power is delivered in the microgrid withacceptable ramp rate, from the system operator view, BESS mustquickly counteract to sudden changes in load and PVDG outputpower. Allowed ramp rate is normally mentioned in kilowatt perminute (kW/min) and is the common feature of solar power

Table 2Technical specifications of 12 V battery module (RM12-75DC).

Nominal Voltage Voc 12 VNominal Capacity C20 (Qmax) 75 AH (20 h)Internal Resistance R0 64.8 mO (Fully charged bPolarization Coef Ki 0.003Electrolyte Resistance KR 0.7 mOUnits in series Ns 1.0Units in parallel Np 1.0Max. Charge Current Ich,max 22.5 A

Final Voltage Time (Mins)

5 10

9.6 V A 265.3 188.2W 2501.6 1880

10.02 V A 235.9 176.5W 2416.7 1822

10.2 V A 216.6 167.4W 2271 1756

10.5 V A 193.1 152W 2066.5 1620

10.8 V A 174.9 136.4W 1820 1491

purchase agreement between the utility companies and powerproducers [31].

Active and reactive power reference values for VSC of batteryplant are determined by PMS. These values are passed throughthe droop mechanism taking into account as P0 and Q0, respec-tively. Since the AC network is robust, the frequency and voltageare fixed in G.C mode and thus PMS dispatching values are usedto directly dispatch BESS. Other types of droop control have beenreported in the literature where all DGs are electronically dispatch-able units [32,33]. In this case, each DG generates its own fre-quency clock by itself through measuring the frequency andvoltage at DG system point of connection (POC).

In power grids with high X/R ratio, the flow of active power pre-dominantly depends on power angle and thus the frequency [34].The flow of reactive power can be regulated by changing the voltagemagnitude at POC. Therefore, P and Q are controlled independentlythrough frequency and voltage droop regulations, respectively, as:

Pref ¼ MP xPOC �x0ð Þ þ P0 ð24Þ

where x0 is nominal grid angular frequency in p.u, xPOC is themomentary angular frequency of voltage at POC in p.u, and MP isactive power droop coefficient as:

MP ¼Pmax � Pmin

xmin �xmaxð25Þ

and P0 (in p.u) is determined directly by system operator (mainagent) or calculated as:

P0 ¼Pmax þ Pmin

2ð26Þ

In steady-state (Verr = 0), and thus:

Qref ¼ MQ VPOC � V0ð Þ þ Q 0 ð27Þ

where V0 is nominal grid voltage in p.u, VPOC is the instantaneousVSC voltage at POC in p.u, and MQ is reactive power droop coeffi-cient given by:

MQ ¼Q max � Q min

Vmin � Vmaxð28Þ

and Q0 is set-point for reactive power in p.u which is determined byPMS or calculated as:

Q0 ¼Q max þ Q min

2ð29Þ

attery)

15 30 45 60

148.5 92.6 67.9 53.3.3 1518.7 925.2 698.6 579

141.5 90 66.7 53.3 1473.9 896.3 687.4 572.3

137.3 88.2 66.1 52.4.4 1426.3 881.5 675.9 564.5

128.1 84 64 51.3.4 1347.3 863.9 665.6 558

116.6 77.9 61.6 49.7.5 1242.3 836 647 547.8

Fig. 11. BESS active power droop control module.

Fig. 13. Torque map diagram of IC engine.

Fig. 12. BESS reactive power droop control module.

Table 4Synchronous generator (LSA 50.1-4P) model parameters.

Parameter Value Unit Description

Vb 0.40 kV Rated RMS line-to-line voltageIb 1.85 kA Rated RMS line current


As shown in Figs. 11 and 12, output of droop mechanisms setsthe active and reactive power references for VSC outer powerloops. The outer power loops would set current references forthe inner current loops.

xb 314.16 rad/s

Base electrical angular frequency

H 1.0 s Inertia constant (including flywheel and engineinertias)

Ta 0.041 s Armature time constantXd 3.53 p.u d-axis synchronous reactanceX0d 0.246 p.u d-axis transient reactance

X00d 0.135 p.u d-axis sub-transient reactanceXq 2.12 p.u q-axis synchronous reactanceX00q 0.169 p.u q-axis sub-transient reactance

X2 0.152 p.u Negative sequence reactanceT 0d 0.222 s d-axis short-circuit transient time constant

T 00d 0.02 s d-axis sub-transient time constant

T 0do 2.72 s d-axis open-circuit time constant

T 00do 0.043 s d-axis sub-transient open-circuit time constant

T 00qo 0.25 s q-axis sub-transient open-circuit time constant

3.4. Diesel generator plant

Some typical models of diesel generator plant have been pre-sented in [35–37]. Diesel generator set has to perform three tasksin here. Firstly, it generates constant active power to shift down/upthe grid active power profile or it shares the instantaneous loadactive power with the utility. Secondly, in the case that the gridis not able to charge the battery, diesel plant must go through thistask. Thirdly, the local load reactive power is assumed to be sharedbetween BESS and diesel plant proportional to their ratings.

The diesel generator plant includes in an internal combustion(IC) engine which drives a synchronous generator. Diesel primemover should respond quickly to any changes in demand and thusreject the disturbances. In diesel power station, a 1.2 MW diesel ICengine, which is mechanically coupled with a 1.28 MV A (cos /= 0.8) synchronous generator, operates as the prime mover. Themodel of IC engine which is available in PSCAD library has beenutilized for the simulation purpose. The IC engine model takesthe mechanical speed of generator and the fuel intake as the inputsand gives the mechanical torque as the output. Input parameters ofIC engine have been set according to Table 3. As shown in Fig. 13,the output torque in any specific cranking angle is defined for thesoftware as a torque map lookup table. Diesel engine has beenrated 17% larger than the synchronous generator capacity to keepthe microgrid stable during the overload condition. Synchronousgenerator and IC engine are supposed to spin at the same speedof 1500 rpm and hence there is no need to exploit gear box.

The synchronous generator has 4 poles and produces 50 Hzsinusoidal voltage waveform. There is a direct and quadrature axesmodel for this component in PSCAD library which has been used tosimulate the diesel plant [38]. The generator parameters (seeTable 4) have been imported into the PSCAD software according

Table 3IC engine model parameters.

Parameter Value Unit

Engine rating 1024 kWMachine rating 1280 kV AEngine rotating speed 1500 rpmNumber of cylinders 6 –Number of engine cycles Four strokes –Misfired cylinder No –

to the manufacturer’s datasheet. Some parameters of generatorhave not been provided by the manufacturer and thus can beapproximated as [39,40]:

Xd � Xq ð30Þ

since the rotor of generator is non-salient,

T 00do ¼Xd � T 0d � T

00d

X00d � T0do

ð31Þ

X00q ¼ 2X2 � X00d ð32Þ

where X2 is the negative sequence reactance, and

T 00q � T 00d ð33Þ

T 00qo ¼Xq

X 00q� T 00q ð34Þ

The manufacturer datasheet provides the moment of inertia J in(kg m2), however, the PSCAD needs inertia constant H in second(s)as a data entry given by:

H ¼12 � J �x2

m

SGð35Þ

where SG is the generator rating base in MVA and,

xm ¼2P�xs ð36Þ

where P is the number of poles and xs is the synchronous angularfrequency in rad/s.

Table 5IC engine governor controller parameters.


T1 0.05 sc Time constant of actuatorT2 0.02 s Engine dead-timeK 1.0 p.u Actuator gain


3.4.1. Governor controllerA governor control mechanism is proposed in this work which

enables the diesel generator to be dispatched by main agentthrough the droop mechanism set-point. Main agent calculatesfor the coefficient of power sharing with AC network and considersthe diesel plant capability. If the power sharing is not the issue,diesel plant is dispatched in fixed values.

The governor of IC engine is to regulate for the fuel intake andthus the engine rotating torque. As shown in Fig. 14, there aretwo time constants in the proposed governor model. The firstone (T1) is the time required for the actuator to move and replacein a new position and is modeled by a first-order lag compensator.Another time constant is the dead-time (T2) taken for all the cylin-ders to receive the fuel since they are not in a similar position at amoment. A hard limiter is also added to ensure that the fuel intakeis not negative and there is an upper limit for the fuel intake thatcorresponds to the maximum generation capacity. The values setfor the parameters of IC engine governor model are cited in Table 5.Active power droop coefficient (MPD) of IC engine governor is givenby:

MDP ¼Pmax;D � Pmin;D

xmin �xmaxð37Þ

where Pmax,D and Pmin,D are the maximum and minimum of activepower generation capability of diesel generator plant in p.u, respec-tively. In steady-state mode, when the frequency is restored to itsnominal value, the momentary angular frequency of diesel genera-tor (xD,POC), measured at POC in p.u, would be equal to synchronousangular frequency (xs) and thus xerr becomes zero. Active powerset-point of plant (P0D) in p.u can be set by main agent in G.C modeor obtained as:

P0D ¼Pmax;D þ Pmin;D

2ð38Þ

3.4.2. Excitation controllerExcitation controller comprises a voltage compensator together

with an exciter. The operational characteristics of excitation sys-tem have been described in [39,41] in detail. IEEE has developedstandard mathematical transfer functions for available commercialexcitation systems for software modeling purposes [42]. The exci-ter AC5A model has been utilized in this paper and its transferfunction is illustrated in Fig. 15. All the variables definitionstogether with their values have been cited in Table 6. AC5A modelhas been already implemented in PSCAD library. Parametersdefault values, which have been chosen in this work, confirmsthe proper response of exciter.

To set for Vref and prevent the circulation of reactive currentbetween the DGs, a reactive power versus Vref droop mechanismis proposed to be added to the excitation controller. This controlloop is also responsible for sharing the load reactive powerbetween the BESS and diesel plant. As shown in Fig. 16, the Vref

can be set for G.C mode. Operator commands the reactive powerset-point (to share the local reactive power demand).

Fig. 14. Diesel plant governor model and su

where QD,POC is the instantaneous reactive power output of syn-chronous generator in p.u, Q0D is the reactive power set-point inp.u given by:

Q0D ¼Qmax;D þ Qmin;D

2ð39Þ

where Qmax,D and Qmin,D are the maximum and minimum of reactivepower generation capability of diesel generator plant, respectively.MDQ is the reactive power droop coefficient and can be calculatedthrough:

MDQ ¼Q max;D � Q min;D

Vmin � Vmaxð40Þ

where Vmin and Vmax are the allowed boundary of variation for Vref inp.u (±5%).

3.5. Load model

To ensure the stability of a power network, the generation ofpower should be closely matched with its consumption. In thiscase, the load dynamic behavior is very important and has to beconsidered in power grid modeling and evaluating stages. Severalkinds of load have been introduced by [39] e.g. constant power(P), constant current (I), and constant impedance (Z) loads. Model-ing of load is not a straightforward task and hence so many factorsare involved such as time, metrological constrains, and economystatus.

There exist two load categories, in this work, i.e. industrial andresidential loads. Each consists of three similar feeders and hencethere are totally six load feeders. The load model is composed bytwo portions. The first portion includes in base load and the secondpart represents as the alternating load. This MG totally supplies260 kV A domestic base load (cos / = 0.963) and 357 kV A indus-trial base load (cos / = 0.98), respectively. Second portion ofdomestic load varies between 85 and 260 kV A (0.9 6 cos /6 0.96). Variable part of industrial load varies between 95 and175 kV A (0.8 6 cos / 6 0.9). Base load has been implemented bystatic load model and the variable load modeled using the dynamicdefinition suggested in Section 3.5.2.

3.5.1. Static loadBase load has been modeled by algebraic functions in which the

instantaneous voltage and frequency are independent variables as:

P ¼ P0 �VV0

� �a

� 1þ KPF � dfð Þ ð41Þ

ggested active power droop controller.

Fig. 15. AC5A excitation controller transfer function.

Table 6Synchronous generator excitation controller parameters.


TE 0.8 s Exciter time constantTA 0.02 s Regulator amplifier time constantTF1, TF2, TF3 1.0, 0, 0 s Regulator stabilizing circuit time constantKA 400 p.u Regulator gainKE 1.0 p.u Exciter constant related to self-excited fieldKF 0.03 p.u Regulator stabilizing circuit gainSE(EFD1) 0.86 p.u Saturation at EFD1

SE(EFD2) 0.5 p.u Saturation at EFD2

EFD1 5.6 p.u Excitation voltage for SE1

EFD2 4.2 p.u Excitation voltage for SE2

VRMIN �7.3 p.u Minimum regulator outputVRMAX 7.3 p.u Maximum regulator output

--+ 1/MDQQD,POC

Q0D

Vref+++

V0

Fig. 16. Synchronous generator proposed reactive power droop controller.


Q ¼ Q0 �VV0

� �b

� 1þ KQF � dfð Þ ð42Þ

where a and b are equal to the voltage indices for active and reac-tive powers, respectively. These two parameters have been set tozero. KPF and KQF are the frequency indices for active and reactivepowers, respectively, and have been considered to be zero. The zerovalue of parameters ensures that the base load is constant powerload and thus is independent from voltage and frequency variations.

3.5.2. Dynamic loadThis portion has been simulated through the variable induc-

tance L in series with variable resistance R and this branch is con-nected to the ground in parallel with the base load. R and L valueshave to be altered during the simulation to represent the variableportion of the load. Since the actual load profile is available, bydeducting the base load whatever remains would be the dynamicportion. Given the values of load active (P) and reactive (Q) powers,R and L values can be calculated as:

GðMhoÞ ¼ P

V2 ð43Þ

BðSiemensÞ ¼ Q

V2 ð44Þ

where G and B are the load conductance and susceptance, respec-tively. V is the nominal RMS line voltage and thus,

RðXÞ ¼ G

Y2 ð45Þ

LðHÞ ¼ B

Y2 �xs

ð46Þ

where xs is synchronous angular frequency and Y is the load admit-tance given by:

Y2ðohm�2Þ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiG2 þ B2

qð47Þ

4. Data inputs

Power system behavior is inherently unpredictable. Generation,transmission, and demand all have a degree of uncertainty. To keeppower system stable and cost effective, grid codes and regulationshave been mandated by the utilities and regulatory authoritiesover the last decades [26,27]. From the system planning and oper-ation perspectives, load scheduling, forecasting, and economic dis-patching are those kinds of remedies to cope with the variablenature of power system. In this case, planners and operators attaina prior knowledge, from their databases, about the problematicuncertainties. They can propose their most efficient and cheapestsolutions in order to resolve these shortcomings.

Load and supply forecasting are both beyond the scope of thisresearch. However, to examine the impacts of load and generationvariations on dynamics of system, it is necessary to collect the realload and PV radiation data. In this paper, the PVDG irradiation andthe daily load profiles are assumed to be forecasted by data centreor a forecasting module (embedded in unit and load agents) oneday in advance. This information, after some logical manipulation,would be used as inputs for PSCAD software.

4.1. Solar irradiation profile

A typical location in Malaysia in the city of Kuala Lumpur withlatitude of 3�,70N and longitude of 101�,390E has been chosen[43,44] as the sampling point. The solar radiation pattern has beenextracted through the HOMER software which is a Micropoweroptimization model developed and supported by Homer EnergyLLC. This software is able to estimate the average hourly irradiationprofile using the method suggested by [45]. HOMER creates a set of8760 solar radiation values for each hour of the year. On the otherhand, to study the dynamic behavior of microgrid in presence ofPVDG, solar radiation profile is needed with higher precision thanthe hourly data [46]. For example, to see the power flow changes,in the system, followed by PVDG power output variation (ramprate) within 5 min, the solar radiation profile should be sampledin every 5 min during a day. In this case, the BESS is sized so thatit can absorb or supply the power difference between Pset and Ppv


(together with that part of smoothing index related to the load)during the period PVDG ramps up or down, respectively. PV plantgenerates power in an unpredictable manner. This fact can be con-sidered as the result of cloudy sky, in worst condition, or rising andsetting of the sun during the day which causes 10–13% variation inPV plant output power. The abrupt changes in generated power arethe main concerns for the system planners and operators. A pass-ing cloud can lead to more than 60% change in PVDG output in amatter of seconds [47]. Fig. 17 shows the average solar irradiation(G) profile for a typical day on May approximated at the coordi-nates of sampling. Twenty four hourly irradiation values have beeninterpolated to obtain 288 values and redraw the solar pattern forevery 5 min during a day. To investigate the ramp rate effects onthe network, a normal distributed random noise with 0 meanand standard deviation 1 was applied on the averaged solar profile[6]. Hence, the noisy signal Gn which is finally estimated throughthe MATLAB module is utilized as the input of PV plant modelgiven by:

Gn ¼ Gþ randn size Gð Þð Þ � 0:5½ � � 0:15 ð48Þ

where G is a (m � n) array in which m = 288 and n = 1, respectively.

4.2. Operating temperature profile

To study the impacts of variation in temperature on the PVDGoutput power, the temperature profile of sampling location wascollected from Malaysia Metrological Department (MMD). Theinput of PV generator accepts the cell operating temperature givenby [48]:

Tcell ¼ TAmb þTNOCT � 20

0:8

� �� Gn ð49Þ

where TAmb is the ambient temperature shown in Fig. 18 and TNOCT isnominal cell operating temperature available in manufacturer’sdatasheet (46 �C). Tcell profile was estimated (in Celsius) duringthe day of sampling and exploited as solar PV generator model datainput in PSCAD.

4.3. Load data

The load characteristic is one of the main factors in power systemmodeling for dynamic stability and transient studies. Similar to thestochastic generation, load is also variable and changes during the

Fig. 17. Forecasted solar irradiation profile by HOMER in Malaysia on May and thenoise applied on.

day. In some hours in a day the load profile is in its maximum valuewhen the utility charges the customer for highest prices. The loadchanges also generate power mismatch between the generationand consumption. System power quality can be highly affectedwhere DGs generate intermittent power and the main grid is weak.In this work, there are two kinds of customer i.e. the residential andindustrial. The former is considered to be non-vital load. The indus-trial processes are the vital loads since any power disruption mayresults in huge money wastage and has hazardous consequencesas well. The load profile associated with each category was collectedfrom Tenaga National Berhad (TNB) office of sampling distributionin Malaysia. As shown in Fig. 19, the industrial load profile is almostconstant since the industrial processes are often redundant whilethe domestic load varies during the day.

5. Power management system (PMS)

A coordinated power management algorithm (PMA) has beenproposed in this paper. PMS, as the highest hierarchical level ofcontrol, runs PMA and dispatches the dispatchable DGs. Economicshas not been considered in suggested PMA and hence the dispatch-ing strategy has been defined based on the technical issues. Themost important PMA objective is to decline the stresses on theAC network due to PV plants ramp ups/downs and load fluctua-tions. In addition, BESS and diesel plant are supposed to sharethe local reactive power demand proportional to their ratings. Die-sel can also share the local active power demand with AC networkas in Eq. (50) or it can supply a constant active power to shift up/down the grid active power profile.

P00D ¼bPL if PD;min � P00D � PD;max

P0D o �w:

�ð50Þ

where P00D is dispatching reference of diesel plant and b is the activepower sharing coefficient determined by the system operator. Theinstantaneous power mismatch between the local generation andload would be compensated by grid.

PMS receives all the field information from lower stream agents(generation and load) and saves them in its database. Data pointedout in here can be battery SOC, PV power production status, break-ers states, DGs apparent power outputs, DGs power capabilities,and forecasting signals.

The integration of DG system should not put stress on the maingrid components. In this MG, since the demand and generation areboth variable, an efficient smoothing index has been proposed inorder to smooth the active/reactive power oscillations generatedby PVDG and load together. PVDG operates in unity power factorand hence it only contributes in the active power ramp ups/downs.In the case of load, both active and reactive power change accord-ing to load profiles. To smooth the active power oscillation duringthe time when the PVDG generation is down (night time), dieseland grid contribute to supply the demand. SOC of BESS is con-trolled instantaneously to make sure that BESS operates withinits power capabilities limits as:

SOCmin � SOC � SOCmax ð51Þ

Diesel can operate in its capacity at nominal frequency which canbe determined by the PMS unit (P0D) or dispatched up to its maxi-mum generation capacity by operator (bPL). It can be used to chargethe battery up to SOCref which is 60% in here. The reference SOC isdetermined by the operator which has the knowledge in advanceabout the load and PV active power profiles. If SOCref is properlyselected, by end of the day, the battery can be recharged up to SOCref

(if needed) by diesel or grid depending on the operator decision.During the daytime, as long as the SOC battery is within the opera-tional boundary, BESS is controlled in P–Q mode and thus active

Fig. 18. Temperature profiles (a) Malaysia ambient temperature. (b) Cell operating temperature.

Fig. 19. Distribution network load profile in Malaysia.


power reference is equal to smoothing index calculated by the mainagent as:

DP ¼ PL � PLavg� �

þ Pset � PPVð Þ ð52Þ

where DP is smoothing index in MW, PL is the instantaneous loadestimated by the load agent in MW, PLavg is the active power loadprofile mean value in MW given by:

PLavg ¼PL;max þ PL;min

2ð53Þ

where

PL;min ¼ Pinds;min þ Pdoms;min ð54Þ

PL;max ¼ Pinds;max þ Pdoms;max ð55Þ

where Pinds,min and Pdoms,max denote the minimum industrial activeload and maximum domestic active load, respectively. PPV is PVDGoutput power which is calculated by PV plant unit agent and is sentto main agent through generation agent. Pset is the moving averagevalue of PVDG output power in MW. Embedded moving averagemodule in main agent receives the noisy PPV profile one day aheadand thus Pset is generated through this module. Taking into accountthat PLavg is 872 kW and PL 2 [787, 1000] kW and thus first term in(52) would be member of [85, 128] kW. If the boundary of changesfor second term in (52) is between 215 and 222 kw (see Fig. 20), thesmoothing index would be member of [�300, 350] kW.

Moving average is often used to reduce the random noise and isthe best offer for time domain encoded signals. Principle of opera-tion is to get average from a number of input signal points andhence to reproduce each point in the output signal as [49]:

Fig. 20. Generation of PV average reference using moving average filter.


Y i½ � ¼ Fk

Xnp�1ð Þ=2

j¼� np�1ð Þ=2

X iþ j½ � ð56Þ

where X and Y are the input (PPV) and output (Pset) vectors, respec-tively. np is the number of points used in the moving average filter.Eq. (56) represents two sides averaging which is so-called ‘‘symmet-rical averaging’’. Symmetrical averaging requires np as an odd num-ber. Moving average is a convolution of input signal with arectangular unity area pulse which exploits filter kernel (Fk) given by:

Fk ¼1np

ð57Þ

As the number of points rises, the level of noise declines (seeFig. 20). In this paper, 21-point moving average filter has beenexploited in order to smooth the noisy signal and generate Pset.

To supply the microgrid reactive power, Q0D of synchronousgenerator excitation controller, and Qref of BESS are both setthrough the commands issued by the main agent. A constantsmoothen share of reactive power (QLavg) is supposed to be sup-plied by the power grid. In this case, the load reactive power fluc-tuation about QLavg is smoothed by the BESS and diesel plant as:

Q Lavg ¼Q L;max þ Q L;min

2ð58Þ

where

Q L;min ¼ Q inds;min þ Q doms;min ð59Þ

Q L;max ¼ Q inds;max þ Qdoms;max ð60Þ

where Qinds,min and Qdoms,max denote the minimum industrial reac-tive load and maximum domestic reactive load, respectively. Inanother case, if the operator decides to supply the whole reactivepower locally, QLavg is set to zero and hence BESS and diesel plantshare the local demand.

Reactive power smoothing difference (DQ) should be sharedbetween BESS and diesel generator proportional to their ratings.In this case, the set point reference values can be calculated as:

Q 0B Q 0D½ �T ¼ ½DQ �1

1þ1=a1

1þa

" #ð61Þ

where a is droop coefficient given by:

a ¼ SBESS

SDieselð62Þ

and,

DQ ¼ Q L � Q Lavg ð63Þ

where SBESS, SDiesel are the MVA ratings of BESS and diesel plant,respectively. Therefore, five operating zones are defined for BESSas follows:

a. DP > 0 and DQ < 0 thus BESS delivers active power (dis-charge) and absorbs reactive power;

b. DP < 0 and DQ < 0 thus BESS absorbs both active (charge)and reactive powers;

c. DP > 0 and DQ > 0 thus BESS delivers both active (discharge)and reactive powers;

d. DP < 0 and DQ > 0 thus BESS absorbs active power (charge)and delivers reactive power;

e. Standby mode in which no active power is supplied orreceived but reactive power still can be exchanged.

As shown in Fig. 21, PMS is able to smartly handle the MG basedon the solar radiation level together with BESS and diesel planttechnical constrains. BESS can be safely charged or discharge whenneeded as well as it shares reactive power with diesel plant at thesame time. PMS also makes it possible for the network operator todetermine the active and reactive power portions supplied by eachgeneration unit. In this case, the step-wise operation of PMS to runPMA is described as follows:

1. Main agent calculates for active and reactive powers smoothingindices. At all the times, the reactive power references are setfor BESS and diesel through Qref and Vref, respectively (seeSection A);

2. From evening to morning when the solar radiation is low, thelocal load is supplied by the grid and diesel plant. If batterySOC is bellow SOCref, depending on operator decision, it has tobe charged by diesel plant or grid up to the level of SOCref (seeSection B).

3. During the daytime (see Section C), BESS smoothing function isactivated if:

Input data received by main agent

PPV

PL

Pset

PLavg Calculate active power smoothig

indexEq. 52

∆P

Calculate reactive Power

smoothing indexEq. 63

Q0B

Q0D

QLavg

QL

DroopQref

Vref

NoYes

NoYes

Yes No

max( ) ( 0)SOC SOC P= ∧ Δ >

refP P= Δ0Diesel DP P=



min( ) ( 0)SOC SOC P= ∧ Δ <

0refP =

0Diesel DP P=

min maxSOC SOC SOC≤ ≤refSOC SOC<

1 2( ) ( )t t t t< ∨ >

NoYes

0refP =0Diesel DP P=NoYes

ref CharP P= −

0Diesel D CharP P P= +

BESSIs charged locally?

ref CharP P= −

0Diesel DP P=

0

P0D

Eq. 50

Pconstant

Eq. 58

Operator decisions

Yes No

Operator’scommands

Section A

Section BSection C

Fig. 21. Proposed smart power management algorithm.

Fig. 22. Comparison between the microgrid without BESS and with BESS. (a and d) Grid active power profile. (b and e) Diesel plant active power output. (c and f) BESS activepower output.


� SOC of BESS is between 30% and 90% when time is between t1

and t2;� BESS is fully-charged and smoothing index is positive;� BESS SOC is fully-discharged and smoothing index is

negative;

Otherwise battery keeps on operating in idle mode (Pref = 0).

6. Simulation studies

Several test cases have been conducted to evaluate the opera-tion of proposed PMA in MG (see Fig. 1) when the system worksin grid-connected mode. Microgrid with the agents and communi-cation channels has been simulated in PSCAD/EMTDC software.Solar radiation, cell operating temperature, and load vary during

Fig. 23. Comparison between the microgrid without BESS and with BESS. (a and d) Grid reactive power profile. (b and e) Diesel plant reactive power output. (c and f) BESSreactive power output.

Fig. 24. Comparison between the microgrid without BESS and with BESS. (a and d) Battery house SOC profile. (b and e) Battery house terminal voltage. (c and f) Battery houseoutput current.


the day. PMA should be able to manage the system and hence reliefthe stresses imposed on AC network. The network operator con-cern is about the active power ramp ups/downs caused by PVDGsince integration of DG systems must not negatively affect theproperties of main supply. For each test case, the proposed activepower smoothing index has been evaluated as well. The ability ofBESS, diesel plant, and grid to share the local active and reactivepower demand has also been investigated in conjunction withother test cases. The role of operator’s decision has been taken intoaccount as well.

6.1. Test case 1: Power smoothing index impact on decreasing stress onAC network, Diesel plant shares load active power with grid and loadreactive power with BESS while grid is to charge BESS

In this test case, diesel plant is supposed to supply 50% of loadactive power (PL) and hence operator decides to dispatch this unit(Eq. (50)) with b = 0.5 and rest of load is supplied by distributionnetwork. BESS plant is assumed to smooth the power fluctuationresulted by load and PVDG plant. The operator, which has beenreceived the forecasted waveforms, sets SOCref to be 60% at the

Fig. 25. (a) Grid active power profile. (b) Diesel plant active power output. (c) BESS active power output.

Fig. 26. (a) Battery house SOC. (b) Battery house terminal voltage. (c) Battery house current.


beginning of the day and at the end of day battery should berecharged (if any) up to this level by the grid. Operator also setsPLavg = 0.872 (MW) and thus the load fluctuations about this valuewould be compensated by BESS. Operator also assigns the dieselplant and BESS to share the load reactive power (QL) locallybetween themselves and sets QLavg = 0. PMA is activated betweent1 and t2 for the purpose of smoothing. In here, t1 and t2 are sup-posed to be at 6:30 a.m and 7:30 p.m, respectively. There are twoscenarios which have been implemented in this test case. MG oper-ates while BESS smoothing functionality is disabled in the formercase and is enabled in the latter case. In both cases, load reactivepower still can be shared properly between two DGs. As shownin Figs. 22–24, between 12:00 a.m and 6:30 a.m the diesel suppliesthe load together with the grid. From 6:30 a.m to 7:30 p.m, as

shown in Fig. 22, AC grid encounters with too many fluctuationsas the PVDG output starts increasing, however, these oscillationsare mitigated for the second case. In the first case, SOC of batterykeeps unchanged since BESS is in idle mode.

In second case, all the active power fluctuations have been sup-pressed by BESS and hence the grid active power profile becomessmoothened (see Fig. 22d). As depicted in Fig. 23, the load reactivepower has been shared between two DGs proportional to their rat-ings. The grid supplies zero reactive power in both cases because QL

is supplied locally. The reactive power outputs of DGs togetherwith the grid reactive power profile for both cases are the samewhich confirms that the control of active power is independentfrom the reactive power. The battery house terminal voltagechanges stably above the final voltage value in the second case

Fig. 27. (a) Grid reactive power profile. (b) Diesel plant reactive power output. (c) BESS reactive power output.

Fig. 28. (a) Grid active power profile. (b) Diesel plant active power output. (c) BESS active power output. (d) Active power smoothing index profile (DP).


(see Fig. 24) and remains constant in the first case. At the end ofday when the SOC becomes less than 60%, the grid starts chargingthe battery (see the grid active power profile which peaks between08:30 p.m and 09:30 p.m in Fig. 22d) BESS current control unitchecks for the reference value for charging power which is sentby operator. If this value is within the boundary, battery keepson charging until the SOC reaches to 60% (see BESS active powerprofile which is absorbed between 08:30 p.m and 09:30 p.m inFig. 22f) and the system gets ready for the next day. If the amountof charging current is above the limit assigned in current controlalgorithm (see Fig. 8), algorithm limits charging active power refer-ence value to maximum allowed value (300 kW).

6.2. Test case 2: Diesel plant can be remotely dispatched to shift up/down the grid active power profile, BESS smoothing function isenabled, diesel plant is to charge the battery, and load reactive poweris shared between diesel plant and BESS

In this test case, operator decides to dispatch the diesel plant byfixed active power reference. The aim is to shift down/up the gridactive power profile while BESS smoothing function is enabled atall the times during the day. As shown in Fig. 25–27, from 12 a.mto 2 a.m the diesel plant active power output is 200 kW (referenceset by operator is 200 kW). Between 2 a.m and 4 a.m active powerreference increases to 400 kW by the operator and then declines to

Fig. 29. (a) Battery house SOC. (b) Battery house terminal voltage. (c) Battery house current.

Fig. 30. (a) Grid reactive power profile. (b) Diesel plant reactive power output. (c) BESS reactive power output.


300 kW from 4 a.m until the end of day. BESS smoothes the powerfluctuations similar with the first case and the changes in dieselplant output power has no effect on its operation. At the end ofday, diesel plant receives the command from the system operatorin order to increase its active power generation and charge the bat-tery. SOC of BESS that has fallen to less than 60% is charged back toits reference SOC (by setting for P0D = 350 kW and Pref = �150 kW)and then system gets ready for the next day.

Reactive power is also shared between diesel plant and BESS. Asillustrated in Fig. 27, whenever the diesel output power changes,its output reactive power slightly oscillates. This matter happensbecause in rotary based DGs control of active and reactive powerto some extent is dependent to each other. To deliver more activepower by synchronous generator, diesel engine should rotate fasterand hence power angle would increase. Along with suddenincrease in power angle, the amount of reactive power deliveredby diesel plant shortly grows. Then, the excitation acts against thisphenomenon and reduce the reference voltage until the operatingpoint returns to its original value in a short while.

6.3. Test case 3: Diesel plant is remotely dispatched to generateconstant active power, BESS smoothing function is enabled, dieselplant is to charge the battery, load reactive power is shared betweendiesel plant and BESS, and QLavg is supplied by the grid

This test case is to show the efficacy of proposed reactive powersmoothing index (DQ). Between 12 a.m and 6:30 a.m BESS is idleand diesel plant generates constant active power. The grid com-pensates for the active power difference between the diesel plantoutput (PDiesel) and the load (see Fig. 28). At 4 a.m operator issuesa command to supply the average load reactive power by the gridand hence the load reactive power fluctuations about this value isshared between diesel plant and BESS. Once the operator com-mand is received by the generation agent, this agent informs unitagent associated with BESS and diesel plant. These DGs adjust theirreactive power generation accordingly (see Fig. 30). QLavg can becalculated through Eq. (58) and it is also possible to be consideredas constant offset value provided that the rest of reactive loadabove this offset complies with BESS and diesel plant power


capabilities as being shared between them. At the end of day, bat-tery is charged by diesel plant (by setting for P0D = 350 kW andPref = �150 kW) up to reference SOC and system gets prepared forthe next day.

As shown in Fig. 28, BESS follows the active power index closelyand the ramp ups/downs are completely compensated by this unit.The voltage at the battery house terminal is fluctuating due to thebattery output current oscillations. Since the final voltage valuehas been chosen properly, these variations have no undesirableeffect on the operation of BESS (see Fig. 29). The current controllermodule in BESS unit agent instantaneously monitors and controlsthe current of battery to make sure that its value is allowable.

7. Conclusion

This paper proposes a novel power management algorithm(PMA) which utilizes the battery and diesel plants efficiently insideof a microgrid which operates in grid-connected mode. This paperalso investigates the role of battery storage in the smartgrid in mit-igating solar plant output power oscillations where both rotary andelectronically interfaced DGs are present in the system. PMA inte-grates moving average filtering method into the scheduled dis-patching of battery plant in order to smooth the grid activepower profile and reduce the undesirable effects on its compo-nents. Moving average filter reduces the computational burdenand helps the operator to determine the degree of smoothness.Results show that if one day ahead forecasting information aboutthe load and solar radiation is available (considering a ramp rateof �50% to +10% for solar plant together with load), a 21-pointsymmetrical filter produces an acceptable smoothened referencewaveform. A practical algorithm is also suggested to size the bat-tery plant accurately based on data available through forecastingmodules together with network constraints. Size of the batteryplant obtained through this algorithm is quite matched to suppresssolar plant power fluctuations along with the variable load. PMA isimplemented in an agent oriented communication environmentwith four hierarchical levels. Communication platform helps thesystem operator to interfere in the operation of PMA and hencedispatch the DGs based on the real-time system requirements.The contribution of diesel plant in shifting up/down the grid activepower profile is evident and makes it possible for the system oper-ator to decide about the locally or remotely (by the grid) supplyingthe loads. The proposed droop mechanism as an outer loop forexcitation controller ensures the accurate reactive power sharingbetween the battery and diesel plants. A load model is designedwhich is suitable for power system real-time simulation andresults prove the efficacy of this model. The overall operation ofmicrogrid to manage the generation and consumption as well asto bring ancillary services for the main supply is quite satisfactory.

Acknowledgment

This work has been supported by High Impact ResearchSecretariat (HIR) at University of Malaya through the ‘‘CampusNetwork Smart Grid System for Energy Security’’ project (ProjectNo: H-16001-00-D000032 & Grant No: UM.C/HIR/MOHE/ENG/32).

References

[1] Acharjee P. Investigation of the power scenario in India for the implementationof smart grid. In: 2012 3rd IEEE PES international conference and exhibition oninnovative smart grid technologies (ISGT Europe); 2012. p. 1–8.

[2] Acharjee P. Strategy and implementation of smart grids in India. Energy StratRev 2013;1:193–204.

[3] Cornforth D. Role of microgrids in the smart grid. J Electron Sci Technol2011;9:9–16.

[4] Koohi-Kamali S, Tyagi VV, Rahim NA, Panwar NL, Mokhlis H. Emergence ofenergy storage technologies as the solution for reliable operation of smartpower systems: a review. Renew Sustain Energy Rev 2013;25:135–65.

[5] Shah R, Mithulananthan N, Bansal RC. Oscillatory stability analysis with highpenetrations of large-scale photovoltaic generation. Energy Convers Manage2013;65:420–9.

[6] Daud MZ, Mohamed A, Hannan MA. An improved control method of batteryenergy storage system for hourly dispatch of photovoltaic power sources.Energy Convers Manage 2013;73:256–70.

[7] Omran WA, Kazerani M, Salama MMA. Investigation of methods for reductionof power fluctuations generated from large grid-connected photovoltaicsystems. IEEE Trans Energy Convers 2011;26:318–27.

[8] Teleke S, Baran ME, Bhattacharya S, Huang AQ. Rule-based control of batteryenergy storage for dispatching intermittent renewable sources. IEEE TransSustain Energy 2010;1:117–24.

[9] Teleke S, Baran ME, Huang AQ, Bhattacharya S, Anderson L. Control strategiesfor battery energy storage for wind farm dispatching. IEEE Trans EnergyConvers 2009;24:725–32.

[10] Ahmed NA, Miyatake M, Al-Othman AK. Power fluctuations suppression ofstand-alone hybrid generation combining solar photovoltaic/wind turbine andfuel cell systems. Energy Convers Manage 2008;49:2711–9.

[11] Yun-Hyun K, Soo-Hong K, Chang-Jin L, Sang Hyun K, Byeong-Ki K. Controlstrategy of energy storage system for power stability in a wind farm. In: 2011IEEE 8th international conference on power electronics and ECCE Asia (ICPE &ECCE); 2011. p. 2970–3.

[12] Yoshimoto K, Nanahara T, Koshimizu G. New control method for regulatingstate-of-charge of a battery in hybrid wind power/battery energy storagesystem. In: Power systems conference and exposition, 2006 PSCE ‘06 2006IEEE PES; 2006. p. 1244–51.

[13] Tan KT, So PL, Chu YC, Chen MZQ. Coordinated control and energymanagement of distributed generation inverters in a microgrid. IEEE TransPower Deliv 2013;28:704–13.

[14] Tan KT, Peng XY, So PL, Chu YC, Chen MZQ. Centralized control for paralleloperation of distributed generation inverters in microgrids. IEEE Trans SmartGrid 2012;3:1977–87.

[15] Colson CM, Nehrir MH. Algorithms for distributed decision-making for multi-agent microgrid power management. In: Power and energy society generalmeeting, 2011 IEEE; 2011. p. 1–8.

[16] Zhang Y, Ooi BT. Stand-alone doubly-fed induction generators (DFIGs) withautonomous frequency control. IEEE Trans Power Deliv 2013;28:752–60.

[17] Yazdani A, Iravani R. Voltage-sourced converters in power systems. Wiley;2010.

[18] Taheri H. Accurate MATLAB simulink PV system simulator based on a two-diode model. J Power Electron 2011;11:179–87.

[19] Nishioka K, Sakitani N, Uraoka Y, Fuyuki T. Analysis of multicrystalline siliconsolar cells by modified 3-diode equivalent circuit model taking leakage currentthrough periphery into consideration. Sol Energy Mater Sol Cells2007;91:1222–7.

[20] Rajapakse AD, Muthumuni D. Simulation tools for photovoltaic system gridintegration studies. In: Electrical power & energy conference (EPEC), 2009IEEE; 2009. p. 1–5.

[21] Bhatnagar P, Nema RK. Maximum power point tracking control techniques:state-of-the-art in photovoltaic applications. Renew Sustain Energy Rev2013;23:224–41.

[22] Ishaque K, Salam Z. A review of maximum power point tracking techniques ofPV system for uniform insolation and partial shading condition. Renew SustainEnergy Rev 2013;19:475–88.

[23] Salas V, Olías E, Barrado A, Lázaro A. Review of the maximum power pointtracking algorithms for stand-alone photovoltaic systems. Sol Energy MaterSol Cells 2006;90:1555–78.

[24] Eltawil MA, Zhao Z. MPPT techniques for photovoltaic applications. RenewSustain Energy Rev 2013;25:793–813.

[25] Yazdani A, Di Fazio AR, Ghoddami H, Russo M, Kazerani M, Jatskevich J, et al.Modeling guidelines and a benchmark for power system simulation studies ofthree-phase single-stage photovoltaic systems. IEEE Trans Power Deliv2011;26:1247–64.

[26] IEEE recommended practice for utility interface of photovoltaic (PV) systems.IEEE Std 929-2000; 2000.

[27] IEEE application guide for IEEE Std 1547. IEEE standard for interconnectingdistributed resources with electric power systems. IEEE Std 15472-2008;2009. p. 1–207.

[28] Poullikkas A. A comparative overview of large-scale battery systems forelectricity storage. Renew Sustain Energy Rev 2013;27:778–88.

[29] Moore S, Eshani M. An empirically based electrosource horizon lead-acidbattery model. SAE Trans 1996;105:421–4.

[30] Rekioua Pr D. Optimization of photovoltaic power systems; 2012.[31] Hill CA, Such MC, Dongmei C, Gonzalez J, Grady WM. Battery energy storage

for enabling integration of distributed solar power generation. IEEE TransSmart Grid 2012;3:850–7.

[32] Hassan MA, Abido MA. Optimal design of microgrids in autonomous and grid-connected modes using particle swarm optimization. IEEE Trans PowerElectron 2011;26:755–69.

[33] Pogaku N, Prodanovic M, Green TC. Modeling, analysis and testing ofautonomous operation of an inverter-based microgrid. IEEE Trans PowerElectron 2007;22:613–25.

http://refhub.elsevier.com/S0196-8904(14)00384-7/h0010






































































[34] De Brabandere K, Bolsens B, Van Den Keybus J, Woyte A, Driesen J, Belmans R,et al. A voltage and frequency droop control method for parallel inverters. In:Power electronics specialists conference, 2004 PESC 04 2004 IEEE 35th annual,vol. 4; 2004. p. 2501–7.

[35] Bo K, Youyi W, Yoke-Lin T. An H1 controller design for diesel engine systems.In: 2000 Proceedings PowerCon 2000 international conference on powersystem technology, vol. 1; 2000. p. 61–6.

[36] Roy S, Malik OP, Hope GS. A least-squares based model-fitting identificationtechnique for diesel prime-movers with unknown dead-time. IEEE TransEnergy Convers 1991;6:251–6.

[37] Roy S, Malik OP, Hope GS. An adaptive control scheme for speed control ofdiesel driven power-plants. IEEE Trans Energy Convers 1991;6:605–11.

[38] Stavrakakis GS, Kariniotakis GN. A general simulation algorithm for theaccurate assessment of isolated diesel-wind turbines systems interaction. I. Ageneral multimachine power system model. IEEE Trans Energy Convers1995;10:577–83.

[39] Kundur P. Power system stability and control. New York: McGraw-Hill Inc.;1994.

[40] Sarma MS. Electric machines: steady-state theory and dynamic performance.2nd ed. Boston, MA: Cengage Learning; 1997.

[41] Anderson P, Fouad A. Power system control and stability; 2002.[42] IEEE recommended practice for excitation system models for power system

stability studies. IEEE Std 4215-2005 (Revision of IEEE Std 4215-1992); 2006[0_1-85].

[43] Kamali SK, Mekhilef S. Evaluation study on grid-connected PV system atUniversity of Malaya. In: 2009 International conference for technicalpostgraduates (TECHPOS); 2009. p. 1–7.

[44] KoohiKamali S, Yusof S, Selvaraj J, Esa MNB. Impacts of grid-connected PVsystem on the steady-state operation of a Malaysian grid. In: 2010 IEEEinternational conference on power and energy (PECon); 2010. p. 858–63.

[45] Graham VA, Hollands KGT. A method to generate synthetic hourly solarradiation globally. Sol Energy 1990;44:333–41.

[46] Katiraei F, Agu JR. Solar PV integration challenges. Power Energy Mag IEEE2011;9:62–71.

[47] Mills A, Ahlstrom M, Brower M, Ellis A, George R, Hoff T, et al. Dark shadows.Power Energy Mag IEEE 2011;9:33–41.

[48] Tsai H-L. Insolation-oriented model of photovoltaic module using Matlab/Simulink. Sol Energy 2010;84:1318–26.

[49] Steven WS. The scientist and engineer’s guide to digital signalprocessing. California Technical Pub.; 1997.
























Smart Power Management

Documents

Transcript of Smart Power Management