Energy Storage 01

lable at ScienceDirect

Renewable Energy 74 (2015) 158e169

Contents lists avai

Renewable Energy

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

A comprehensive power loss, efficiency, reliability and cost calculationof a 1 MW/500 kWh battery based energy storage system forfrequency regulation application

Md ArifujjamanBroomfield, CO, USA

a r t i c l e i n f o

Article history:Received 21 April 2014Accepted 24 July 2014Available online

Keywords:Conduction lossCostEnergy storage systemSwitching lossEfficiencyReliability

E-mail address: [email protected].

http://dx.doi.org/10.1016/j.renene.2014.07.0460960-1481/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

Battery based energy storage system (ESS) has tremendous diversity of application with an intense focuson frequency regulation market. An ESS typically comprised of a battery and a power conversion system.A calculation of performance parameters is performed in this research. The aim is to formulate an in-depth analysis of the ESS in terms of power losses of the semiconductor and electrical devices, effi-ciency, reliability and cost which would foster various research groups and industries around the globe toimprove their future product. In view of this, a relation between the operating conditions and powerlosses is established to evaluate the efficiency of the system. The power loss calculation presented in thispaper has taken into account the conduction and switching losses of the semiconductor devices. Af-terwards, the Arrhenius Life Stress relation is adopted to calculate the reliability of the system byconsidering temperature as a covariate. And finally, a cost calculation is executed and presented as apercentage of total cost of the ESS. It has been found that the power loss and efficiency of the ESS at ratedpower is 146 kW and 85% respectively. Furthermore, the mean time between failures of the ESS is 8 yearsand reliability remains at 73% after a year. The major cost impact observed is for battery and PCS as 58%and 16% respectively. Finally, it has been determined that further research is necessary for higher efficientand lower cost system for high penetration of energy storage system in the market.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Energy storage technologies are emerging as the most prom-ising solutions for augmenting frequency regulation application forutilities. Large scale energy storage solution prefers pumped hydrodue to the maturity of technology as well as requirement posed bythe utility [1]. However, other smaller technologies such as com-pressed air, thermal, batteries, and flywheels are also evolvingrapidly because of near commercial product viability throughvigorous research by the research groups and industries around theglobe. In contrast to the capabilities of other smaller technologies,battery storage technology is forefront owing to the competency oflower power and shorter discharge times, ranging from a fewseconds to 6 h, and easily adaptable at a site without ampleattention on any specific geographical features.

Among various battery chemistries, lead-acid battery remains adominant choice for grid-connected energy storage applications.However, Lithium-ion battery technologies promised enhanced

energy storage densities, greater cycling capabilities, higher safetyand reliability, and lower cost and have reached production levelsas necessary to meet market cost and quality control requirements[2]. In a package level, a significant advantage of lithium-ion systemas compared to the lead acid is that, a lead acid systemmust have alarger nameplate energy capacity than the lithium-ion system tohave the same amount of available energy and is a favorable choicein the industry. In addition to the battery system, an efficient PowerConversion System (PCS) is one of the most crucial parts of anyEnergy Storage Systems (ESSs). It serves as the interface betweenthe storage devices and the utilities that distributes electricity to itscustomers. While a 2-level PCS for DC to AC power conversion forutilities is readily available, a 3-level PCS could be an optimumchoice [3]. Adaption of low voltage switches, multiple level ofoutput voltage, and lower total harmonic distortion are some of theadvantages of a 3-level PCS. The power losses due to switching andconduction losses are lower due to the use of lower switchingfrequency and low forward voltage drop of the semiconductorswitches. Furthermore, when several voltage levels are used, thedv/dt of the output voltage is smaller thus the stress in cables andbatteries is smaller.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 159

Recently, many industries around the globe are developingbattery based ESS including an integrated PCS for frequency regu-lation application. Industries, such as ABB, Saft, Dynapower, Parker,Bosch, Princeton Power System is the top notch brand namesamong others around the globe. The basic product line from theindustries ranges from 1MW to 4MW system for a time duration of30 min to 1 h. Unfortunately, a detail of the information of theproducts is not publicly available due to the non-disclosureagreement nature of the R & D program. These possess a grimconstraint to other emerging industries of the same domain as theyhave a very minimal to no information about any in-depth initialconsiderations of the ESS product. The non-disclosing nature of theproduct information also discourages a high penetration andfurther development of the energy storage systems to the market.In order to enhance a profound understanding of the internal na-ture of the ESS, an in-depth study is performed and findings arepresented to encourage others for further development of theirproduct. Although author has investigated an in-depth analysiswith the best of his knowledge, however, author does not take anyresponsibility in any circumstances' if applying this research doesnot meet the performance expectation of an ESS by any individual,research groups or any industries. In addition, the research pre-sented in this article is Author’s own and don’t represent anycompany positions, strategies, or opinions.

Among various performances and design criteria for the ESS, theoverall power losses, efficiency, reliability and cost are the mostsignificant factors that needs extensive investigation because of agrowing concern regarding the energy savings, efficiency and cost.However, a considerable lack is observed in the previous literaturesthat practically discusses with the investigation on calculation ofpower loss, efficiency and reliability that varies with the operatingpoints for an energy storage system. Furthermore, a detail costbreakdown for an ESS is almost null in the previous literature byeither any research group or industries which is an essence fordeveloping a product.

An extensive literature review has been performed and foundthat there is a considerable need to comprehensively calculate thepower losses of the semiconductor and other electrical devices forthe ESS. A calculation of power losses of a PCS for a given operatingcondition is performed in Refs. [4e19] in terms of the total semi-conductor power losses. However, calculating individual semi-conductor power loss lacks a considerable valid justification. This isbecause, firstly, a non-linear loss calculation approach is unable toreflect the switching losses of the semiconductor devices, whichcould be a dominant factor during the high switching state [4e11].Secondly, power loss calculation based on the data provided by themanufacturers is ambiguous and pessimistic [12e16]. Thirdly,physics-based simulation models of semiconductor devices powerlosses requires implicit integration methods, leading to anincreased simulation time. Furthermore, it requires detail knowl-edge of the dimensions of the devices [17e19]. There have beenvery limited efforts found on modeling of the PCS as well as batterypower losses that constitute an ESS for frequency regulationapplication.

In addition, most of the reliability calculations for electroniccomponents are based on the accessible data provided by the mil-itary handbook for reliability prediction of electronic equipmentwhich is criticized for being obsolete and pessimistic [20,21]. Acomparative reliability calculation of different PCS has been carriedout based on themilitary handbook by Aten et al. [21]; however, theabsence of environmental and current stress factors can pose grimconstraints on the calculated reliability value. Rohouma et al. [22]provided a reliability calculation for an entire PV unit which canbe consideredmore useful, but the approach lacks valid justificationas the data provided by the author is taken from themanufacturers'

published data which is somewhat questionable. This is due to thefact that reliability calculations using purely statistical methods[12], manufacturers data [22,23,28], or military handbook data [24]neglect the operating point of a component. Moreover, the totalnumber of components could vary for two systems (which have thesame objective) in order to meet a certain criterion of the overallsystem. Although higher components in the ESS will exhibit lessreliability and vice versa, the effects of the covariates could bedifferent and consequently could lead to a variation in the reliability[25]. Furthermore, a reliability evaluation for the ESS of a gridconnected application is essential in order to optimize the systemperformances as well as system cost [26]. Another important pointtomention is that reliability analysis based on the covariate factor isstrongly influenced by the standard reliability data book also. Forexample, it is shown in previous research that different values ofcovariate factor for a same covariate is possible by using a differentreliability standard data book [27]. This variation in covariate factoralso varies the reliability of an integrated systemwhich is composedof numerous semiconductor devices. Moreover, it is well under-stood that an error in reliability prediction for a system could proveto be fatal for the high penetration of ESS.

Based on the above discussions, it can be asserted that most ofthe attempts for the power loss and reliability calculation havebeen developed so far based on several assumptions and oftenneglected a fraction of the entire power losses as well as could notconvey the actual reliability data of the system. Furthermore, apower loss and reliability calculation in the energy storage domainis difficult to find. This discrepancy could affect the preference of anefficient grid-connected ESS that is in a great need for high pene-tration of frequency regulation application. As a consequence, thisresearch aims at advancing the use of grid-connected ESS bycalculating the power losses and reliability of the semiconductorand other electrical devices of ESS for varying operating conditions.Based on the power generation and loss with operating points,efficiency is calculated for the system. A novel approach has beenpresented to relate the power loss to the reliability calculationthrough Arrhenius Life Stress relation and consequently mean timebetween failures of the ESS is quantified, which can be consideredthe most widely used parameter in reliability studies [20]. Theresearch then extended the scope by calculating the cost of theenergy storage system thus helps other individuals, researchgroups or industries to gain a preliminary assumption on the cost ofthe system.

This paper is organized as follows: Followed by a detail litera-ture review in the first section, the configuration of the ESS ispresented in the second section. The third section describes thepower loss calculation in the semiconductor and electrical devicesfor considered operating conditions and corresponding efficiencycalculation is presented in fourth section. The fifth and sixth sectiondescribes the approach to calculate reliability and a module basedcost calculation approach of the ESS. The calculation results anddiscussions are presented in seventh section and finally, the find-ings of the investigations are highlighted in the conclusions.

2. Energy storage system description

Fig. 1a shows a functional block diagram of the ESS connected toa low voltage bus that consists of a combination of four BatteryStrings (BS) and two-parallel-operated 3-level PCS. Each BScomposed of a series connected battery modules (battery modulesare formed by the individual battery cells in series) and a 3-levelPCS which transfers energy to the local low voltage ac bus. TwoBS are protected by a single Battery Management System (BMS)that has a bi-directional communication with the Energy Man-agement System (EMS). The EMS is the supervisory controller that

Fig. 1. Energy storage system functional a) Block diagram, b) Detail component level connection.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169160

Fig. 2. Thevenin equivalent representation of the BS.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 161

accepts frequency signal from the dispatch center and communi-cates that with the BMS and local controller of each 3-level PCS. ThePCS is based on P and Q control and the PCS couples to the Point ofCoupling (PCC) through a delta-wye transformer, acting as a sourceof leading or lagging active/reactive current. Each PCS shouldmaintain the frequency at the PCC using only local information. ThePCS can not only convert the input dc voltage to a three-phase ACvoltage with desired magnitude, frequency and phase angle at thePCC, but also capable to supply bidirectional controllable active andreactive power to limit the fluctuation of the frequency and voltageto an allowable range if required. However, it should be mentionedthat the primary objective is to ensure the injection and absorptionof active power depending on the frequency signal and if required,the PCS is capable to perform to regulate the voltage. The system iscapable providing 1 MW output of 480VAC/60 Hz, three phase lowvoltage power. The initial energy capacity is 500 kWh. The systemalso adopts LiFePO4 battery technology with long cycle life andlarge cell capacity tomeet theMW-scale energy storage output. Theswitchgear and step up transformer is neglected due to the out ofscope of this research. Fig. 1b shows the detail of the electricalcomponent level connection that forms the ESS.

3. Power loss calculation

A mathematical model of the power losses in the internalresistor of the battery and semiconductor devices (diodes/IGBTs)for the 3-level PCS is required in order to calculate the efficiency ofthe ESS. The losses for the resistor and semiconductor devices arestrongly dependent on the voltage and current waveforms.Simplified analytical derivation of voltage and current equationsassociated with the individual semiconductor devices are derivedto determine the power losses. The power loss calculation pre-sented in this investigation focus on the losses generated duringthe conduction and switching states of the semiconductor devices.

3.1. Battery

In an ideal world, a battery cell can be represented as an idealvoltage source. However, a more practical approach but still ideal isto represent battery using a voltage in series with a resistor. Thisform of representation is the simplest types of battery cell modelsand has beenwidely accepted in electric circuit analysis and design[29e31]. However, it needs to be mentioned that they are over-simplified and cannot give any detailed and accurate informationabout the battery operation and performance such as the batterySOC, thermodynamics, etc. More advanced battery circuit modelswill be considered and left for future research.

The battery string modeling is performed based on the Theve-nin's equivalent circuit. Fig. 2 shows the Thevenin's equivalentmodel of one of the BS, where Req is the equivalent series resistanceof series combination of battery resistances which is calculatedbased on the Thevenin's equivalence.

It has been considered that the battery will be charged anddischarged at the same 2C rate. In such a situation the batteryterminal voltage due to internal resistance can be expressed as

Veq�t ¼ Veq�b � Req�b � Ieq�b (1)

where Veq-t is the terminal voltage of the Thevenin equivalentvoltage of a BS, Veq-b is the Thevenin equivalent open circuit voltageof the battery string, Req-b is the Thevenin equivalent resistance ofthe BS and Ieq-b is the DC current from the battery string and servesas the input to the PCS. It should be mentioned that each batterystring is composed of several battery modules that is essentiallymade up of battery cells.

The power loss of the battery then can be calculated as.

Pl�eq�b ¼ Veq�t � Ieq�b (2)

3.2. 3-Level power conversion system

The conduction losses Pc are comprised of losses in the IGBTsand diodes. The conduction losses for each switch can be calculatedby (3) [32,33]

Pc ¼ U0Iavg þ rf i2rms (3)

where U0 is the forward voltage drop with zero current, rf is theforward resistance, Iavg is the average current and irms is the root-means-square of the current.

Figs. 3 and 4 summarize all possible power paths and switchingstates in the 3-level PCS. The load current Iom(t) ¼ Iomsin(ut � f)and phase leg voltage as Vo(t) ¼ Vomsinut, and the duty cycle acrossthe switching devices as:

dT11 ¼�M sin ut 0 � ut � p

0 p � ut � 2p (4)

dT12 ¼�

1 0 � ut � p1þM sin ut p � ut � 2p (5)

dT13 ¼ 1� dT11 (6)

dT14 ¼ 1� dT12 (7)

The average and rms currents in IGBTs T11 and T14 of the 3-levelPCS are calculated as follows [32]

I3�lvl�PCST11;avg ¼ I3�lvl�PCS

T14;avg ¼ 12p

Zp

0

dT11iomdut

¼ MI3�lvl�PCSom4p

½sinj4j þ ðp� j4jÞcos 4� (8)

I23�lvl�PCST11;rms ¼ I2

3�lvl�PCST14;rms ¼ 1

2p

Zp

0

dT11i2omdut

¼ MI23�lvl�PCSom4p

�1þ 4

3cos 4þ 1

3cosð24Þ

�(9)

Fig. 3. Current direction in one leg of 3-level PCS.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169162

where I3�lvl�PCSom is the peak current of the output current; 4 is the

phase difference between output voltage and current; M is themodulation index

The average and rms currents in IGBTs T12 and T13 of the 3-levelPCS are calculated as follows [32]

Fig. 4. The switching st

I3�lvl�PCST12;avg ¼ I3�lvl�PCS

T13;avg ¼ 12p

64Zp

iomdutþZpþ4

dT12iomdut75

0 p

3

¼ I3�lvl�PCSom

p�MI3�lvl�PCS

om4p

½sinj4j� j4jcos4� (10)

ates of 3 level PCS.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 163

23�lvl�PCS 23�lvl�PCS 1

26Zp

2Zpþ4

2

37

2p4

0

iomdut þp

dT12iomdut5

¼ I23�lvl�PCSom

4�MI2

3�lvl�PCSom4p

�1� 4

3cos 4þ 1

3cosð24Þ

�

(11)

In principle the diodes from D11 to D14 don't carry any current,because the current of T11 commutes to D15, the current of T14commutes to D16 and the current of T12 commutes to T13. This isdemonstrated in Ref. [21].

The average and rms currents in diodes D15 and D16 of the 3-level PCS are calculated as follows [32]

I3�lvl�PCSD15;avg ¼ I3�lvl�PCS

D16;avg ¼ 12p

264Zp

0

dT13iomdut þZpþ4

p

dT12iomdut

375

¼ I3�lvl�PCSom

p�MI3�lvl�PCS

om4p

½ðp� 24Þcos 4þ 2 sinj4j�(12)

I23�lvl�PCSD14;rms ¼ I2

3�lvl�PCSD15;rms

¼ 12p

264Zp

0

dT13i2omdut þ

Zpþ4

p

dT12i2omdut

375

¼ I23�lvl�PCSom12p

½3p� 6M � 2M cosð24Þ� (13)

By substituting the current through T11 or T14 of the PCS into(3), the conduction loss for T11 and T14 becomes [34]

P3�lvl�PCSc;T11T14 ¼ 2�

�Ui0I

3�lvl�PCST11=T14;avg þ rif I

23�lvl�PCST11=T14;rms

�(14)

where Uio and rif is the forward voltage and resistance of the IGBTrespectively.

In a similar manner the conduction losses of T12 and T13 of thePCS is

P3�lvl�PCSc;T12T13 ¼ 2�

�Ui0I

3�lvl�PCST12=T13;avg þ rif i

23�lvl�PCST12=T13;rms

�(15)

Similarly, the conduction losses of D15 and D16 of the PCS is

P3�lvl�PCSc;D15D16 ¼ 2�

�Ud0I

3�lvl�PCSD15=D16;avg þ rdf I

23�lvl�PCSD15=D16;rms

�(16)

where Udo and rdf is the forward voltage and resistance of the dioderespectively.

Using (14)e(16), the total conduction losses can be determinedby

P3�lvl�PCSc ¼ 3

�P3�lvl�PCSc;D15D16 þ P3�lvl�PCS

c;T11T14 þ P3�lvl�PCSc;T12T13

�(17)

where fsw is the switching frequency of the PCS; ER is the recoveryenergy of the switch.

The major switching losses of a pn-diode are primarily due tothe turn-off losses since the turn-on losses are negligible ascompared to the turn-off loss. The energy dissipation at turn-off isdependent on the charge stored in the depletion region and not lostdue to internal recombination. During the reverse recovery, thecurrent flows in the reverse direction while the diode remainsforward biased, and this results in a high instantaneous power loss

in the diode. Under the assumption of a linear loss model for thediodes, the switching loss of the diodes D15 and D16 is given by(18).

P3�lvl�PCSsw;D15D16 ¼ 2fSWESR

p$Vdc2Vref ;d

$Idc1Iref ;d

(18)

where fsw is the switching frequency of the PCS, ESR signifies therated switching loss energy given for the reference commutationvoltage and current Vref,d and Iref,d, while Vdc2 and Idc1 indicate theactual commutation voltage and current respectively.

In a similar manner, the switching loss of the IGBTs T11 and T14is given by

P3�lvl�PCSsw;T11T14 ¼ 2fSWðEON þ EOFFÞ

p$Vdc2Vref ;i

$Idc1Iref ;i

(19)

The reference commutation voltage and current for the IGBT isVref,i and Iref,i respectively. EON and EOFF signifies the turn-on andturn-off energies of the IGBT as can be found in the datasheet.

The total switching losses can be calculated as

P3�lvl�PCSsw ¼ 3

�P3�lvl�PCSsw;D15D16 þ P3�lvl�PCS

sw;T11T14

�(20)

There the total power loss of the 3-level PCS can be found as

P3�lvl�PCSl ¼ P3�lvl�PCS

sw þ P3�lvl�PCSc (21)

So the total power losses of the of the battery and 3-level PCScan be determined by using (2) and (21) as.

PESSl ¼ Pl�eq�b þ P3�lvl�PCSl (22)

4. Efficiency calculation

The power condition for grid connected ESS typically does notrequire a DCeDC converter for the grid-connected PCS. Because ofthe high voltage output of the lithium e ion battery that is capableto supply enough voltage to the PCS input for a proper injection ofsinusoidal voltage and current in the grid.

In order to calculate the efficiency of the systems, the relationbetween the operating point and power generation/loss is needed.Each of the battery string is composed of 35 battery module con-nected in series, hence the power generation, Pg of the ESS isexpressed as.

Pg ¼ ViIi¼w1�wni (23)

where wi represents a particular battery module for and Pgi rep-resents the power generation for wi battery module. Vi and Iirepresent the voltage and current for wi module respectively. Thepower loss of each system can be found as described in Section IVand the total power loss is mathematically expressed as

Pl ¼�PESSli

�i¼w1�wn(24)

The global efficiency, h of the ESS is then calculated as,

h ¼ Pgi � PliPgi

� 100% (25)

5. Reliability calculation

Reliability is the probability that a component will satisfactorilyperform its intended function under given operating conditions.The average time of satisfactory operation of a system is the Mean

M. Arifujjaman / Renewable Energy 74 (2015) 158e169164

Time Between Failures (MTBF) and a higher value of MTBF refers toa higher reliable system and vice versa. As a result, engineers anddesigners always strive to achieve higher MTBF of the powerelectronic components for reliable design of the power electronicsystems. The MTBF calculated in this paper is carried out at thecomponent level and is based on the life time relationship wherethe failure rate is constant over time in a bathtub curve [35]. Inaddition, the system is considered repairable. It is assumed that thesystem components are connected in series from the reliabilitystandpoint. The lifetime of a power semiconductor is calculated byconsidering junction temperature as a covariate for the expectedreliability model. The junction temperature for a semiconductordevice can be calculated as [36].

TJ ¼ TA þ PlRJA (26)

Pl is the power loss (switching and conduction loss) generatedwithin a semiconductor device and can be found by replacing the Plfrom the loss calculation described in Section 3 for eachcomponent.

The life time, L(Tj) of a semiconductor is then described as

L�TJ� ¼ L0 exp

��BDTJ�

(27)

where,

L0 is the quantitative normal life measurement (hours) assumedto be 1 � 106

B ¼ EA/K, K is the Boltzman's constant which has a value of8.6 � 10�5 eV/K, EAis the activation energy, which is assumed tobe 0.2 eV, a typical value for semiconductors [37].DTJ is the variation of junction and ambient temperature and canbe expressed as

DTJ ¼ TA1 � TJ1 (28)

The failure rate, l is described by

l ¼ 1L�TJ� (29)

The global failure rate, lsystem is then obtained as the summationof the local failure rates, li as:

lsystem ¼XNi¼1

li (30)

The Mean Time Between Failures, MTBFsystemand reliability,Rsystem of the system are given respectively by

MTBFsystem ¼ 1lsystem

(31)

Rsystem ¼ e�lsystemt (32)

In addition to the above mentioned method, a partial stressprediction method is used to calculate the reliability of the batteryresistor. Themethod calculates the failure rate of any component bymultiplying a base failure rate with operational and environmentalstress factors (electrical, thermal etc). It is assumed that the batterycarries a continuous duty cycle operation. The power loss in resistorcan be found from (2) and based on this value, a commerciallyavailable resistor is selected and the stress ratio, S is calculated asthe ratio of the operating power to the rated power of the resistor.The base failure rate, gb is than calculated as [24]

gb ¼ 4:5� 10�9 exp12

Tv þ 273

343

exp

S0:6

Tv þ 273

273

(33)

where Tv is the ambient temperature (�C) and S is the stress factor.The failure rate for a wire wound resistor is given by Ref. [24]

gR ¼ gbpRpEpQ � 1� 10�6 failure=hour: (34)

where the resistance factor, pR is 1 as the external resistance is lessthan 1 MU. The environmental factor, pE is 1 due to the fact that aharsh environment is not considered, and the quality factor, pQ isconsidered to be 15 due to the use of a commercial resistor.

6. Cost calculation

The preliminary cost of the energy storage system is calculatedbased on the available market price of each equipment. The ESS isconsidered to build on a module concept where each of the mod-ules would perform a specific assigned task. Moreover, the moduleswill be connected to each other through a detail integration plan.The cost of ESS is subdivided into 7 sections based on modules andwork load for integration. A short description of each of the sectionis presented below:

1. PCS: The PCS mainly comprised of inverter and related switch-gear. It is assumed that the inverter supplied by a manufacturerthat includes related circuit breaker, fuse and other accessoriesthat is required for proper protection of the utility and personal.

The cost of the PCS, PCScost is calculated from individual com-ponents and expressed as a percentage of the total system cost, Tscand given by (35)

PCScost½%� ¼ 16:1Tsc (35)

2. Battery: The battery section holds the battery string, BMS andnecessary DC fuse and breaker and expressed by

Bcost½%� ¼ 57:4Tsc (36)

3. Electrical System Module (ESM): The ESM integrates PCS, su-pervisory and local controller, fans and etc. The ESM also in-tegrates the auxiliary power and other components as requiredto perform the essential task. The cost of the ESM is calculated as

ESMcost½%� ¼ Filtercost þ Aux Transcost þ Contr:cost þ Contac:cost

þ Power_supcost þ Fusecost þ Conncost ESMcost½%�¼ ð0:2þ 0:27þ 6:05þ 1:15þ 0:35þ 0:18þ 0:92ÞTsc

(37)

4. Harness: Each of the sectionwithin the ESS would be connectedto each other using harness assembly for ease of integration andtesting purposes

Hcost½%� ¼ 1:8Tsc (38)

5. HVAC: The HVAC section includes the gas suppression system,ventilation and others as required for proper safety of apersonal.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 165

HVcost½%� ¼ 4:6Tsc (39)

6. Mechanical: The mechanical system comprised of any base thatis required to place the ESM, battery, panel doors and others asrequired for the ESS

MCcost½%� ¼ 4:6Tsc (40)

Labor: The labor considered for a personal to integrate the point

LAcost½%� ¼ 6:4Tsc (41)

Fig. 5. Power loss as a percentage of rated power for

7. Results and discussions

The analytical calculations illustrated in the preceding sectionswere carried out to determine the total power generation/losses,efficiency, MTBF and consequently reliability of the ESS undervarying operating conditions. The rated power for the ESS isassumed to be 1 MW/500 kWh. The PCS switching frequency isconsidered as 3 kHz and to investigate the worst-case scenario ofthe power loss in this numerical calculation study, the modulationindex is assumed unity and load current is assumed in phase withthe output voltage. In addition, it is well understood that typicallyan ESS operates based on the frequency signal from the dispatchcenter and it is very difficult to pre assume a well operating

a) Battery system, b) Power conversion system.

Fig. 6. Efficiency as a percentage of rated power for battery, power conversion system and energy storage system.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169166

condition. However, in order to achieve economic feasibility, it isextremely important to investigate the reliability at rated powerlevel. Generally rated power of an ESS is considered beforedeployment of an ESS even though the ESS may operate fraction ofthe rated power for most of the time of the year. As a result, toemulate the worst case scenario, reliability at rated power level isan important aspect from a system for high penetration of energystorage to the utility. This realistic assumption leads to determinethe reliability for a power level of 1 MW/500 kWh. The thermalmodel of the battery and PCS is neglected provided that the heatsink is adequate enough to maintain the battery/semiconductorsproper working. Powerwasted in the power supplies for the controlof the converters is also ignored (It may be between 1 kW and5 kW). The analytical calculation is based on the Semikron IGBTmodule SKiiP 1213 GB123-2DFL V3 [38].

The power loss of the battery for 10%e100% of rated power ofthe ESS is presented in Fig. 5a. Higher values of power results in

Fig. 7. Component reliability of batter

high power losses and vice versa while charging and dischargingstate of the battery. It is assumed that the resistance is unchangedduring charging and discharging state. The corresponding con-duction and switching losses as well as the total power loss of thePCS is presented in Fig. 5b for a similar operating condition. Theresults of the power losses for both battery and PCS is higher assoon as the ESS shifts the operating point from low to high regime.It has been found that the maximum power loss at rated powerlevel for battery and PCS of the ESS are 130 kW and 16 kWrespectively, while the total power loss of the ESS at rated power is146 kW.

A comparison of efficiency for the battery and PCS as well asoverall efficiency of the ESS is presented in Fig. 6. The operatingconditions are the same to make a fair assumption between powerloss and efficiency. It is obvious that the battery efficiency degradesas soon as the operating level shifts from 10% to 100% of ratedpower, however, remains in the vicinity of 87% which is similar to

y and power conversion system.

Fig. 8. Reliability of the battery, power conversion system and energy storage system a) Over a year, b) Over time.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 167

other previous literature. However, the PCS efficiency remains inthe level of 98% which is obviously justifies the appropriate use of a3-level PCS. It is found from the previous literature that the totalharmonic distortion of the 2-level PCS is high compared to the 3-level PCS. This is understandable as more voltage level can be ob-tained from a 3-level PCS that would generate fewer harmonics tothe utility. The output filter requirement of a 2-level PCS is highcompared to a 3-level PCS. This requirement can also be validatedfrom the harmonics assumption. Dimension and cost of a 2-levelPCS is high compared to a 3-level PCS. Even though a 2-level PCShas lower device count, nevertheless, lower rating devices can beused for the same voltage level compared to a 2-level PCS wouldmake a 3-level PCS less costly and could be an optimum choice foran energy storage system. Finally, the overall ESS efficiency is found85% at 100% power level which is a considerable efficiency for

moving forward. Nevertheless, further research is absolutelynecessary to enhance the efficiency and the work is in progress bythe author.

Afterwards, the reliability calculation is performed following theprocedure outlined and the results are presented in Fig. 7. Thecalculation reveals that the battery failure rate for the ESS is1.39 � 10�5 and the MTBF is 7.17 � 104 h (8 years). The corre-sponding figure for the PCS is 2.16� 10�5 and 4.64 � 104 h (5 years)respectively. It is well understood that the ESS needs to be afford-able, reliable and most importantly, almost maintenance free forthe average qualified personal to consider installing one. As can beseen, the need to replace the ESS corresponds to the MTBF value of8 years. However, it should be kept in mind that typically; a com-plete checkout would occur in each year by the ESS manufacturerand in such a scenario, without any maintenance the ESS is capable

Fig. 9. Cost of the modules of an energy storage system as a percentage of the total cost.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169168

of a continuous duty cycle operation for around 8 years. Moreover,the reliability calculation assumed that all the components areconnected in series, which is a very conservative estimation ofreliability. In addition, the PCS reliability is found to around 5 yearsbased on the previous literature which was primarily computedform the field data [20,28,39e41]. This study confirms the resultsthrough quantitative calculation which can be a useful tool toextend the calculation for other PCS configuration.

It should be mentioned that besides the switches, the DC linkcapacitors contribute significantly to cost, size and failure of the PCSon a considerable scale. However, the present research work as-sumes that the energy storage requirement to the DC link will bereduced in such a quantity so that Aluminum capacitors could bereplaced by Metallized Polypropylene Film Capacitors to achievehigher level of reliability without considerably increase the costand volume. Nevertheless an effort is undertaking by the author toinclude a better design of the DC link capacitor and include thereliability with the overall system in near future.

Fig. 8a shows the reliability of the battery and PCS for a period ofone year (8760 h) for the ESS. The result reveals that the reliabilityof the battery and PCS for the ESS drops to 88% and 83% after oneyear, while the reliability of the ESS drops to 73% after one year. Thereliability of the battery and PCS as well as the ESS time is presentedin Fig. 8b. It is easily noted that the reliability of the battery reachesless than 50% at 50,000 h (5 years), while PCS maintains a lowervalue of 35% in the same time frame. The reliability of the ESS re-mains 17% which combines both performances of battery and PCS.In both scenarios, the ESS illustrates a reasonable reliability and iscertainly a hopeful direction for the ESS manufacturer around theglobe. This would also enhance to work further on the reliability asthis research quantifies a parameter which could be a good startingpoint for further research in the energy storage domain.

Afterwards, the cost calculation is performed as described inSection 6. An initial design of individual module is performed at thebeginning with a consideration that the proper control and oper-ation of the ESS is achieved. Fig. 9 reflects that the battery and PCSconstitute a major portion of the cost (16% and 58% respectively),while mechanical and harness assembly (5% and 2% respectively)

has the lowest impact on the total cost. The ESM, labor and HVACsystem remains in between higher and lower end of the ESS cost. Avigorous cost reduction of the ESS is being undertaken by theauthor and left for future publication.

8. Conclusions

The power loss, efficiency, reliability and cost calculation of agrid-connected energy storage system for frequency regulationapplication is presented. Conduction and switching loss of thesemiconductor devices is used for power loss and efficiencycalculation and temperature is used as a stress factor for the reli-ability calculation of the energy storage system. In addition, amodule based approach for the energy storage system cost calcu-lation is presented. It is found that the system ensures lower lossand consequently higher efficiency. Moreover, the mean time be-tween failures is in an acceptable agreement and battery and PCShas the highest impact on the cost of the system. It is expected thatmore research will be undertaken for a more efficient and reliableas well as lower cost system in near future.

References

[1] Grid energy storage. US Department of Energy; December 2013. p. 1e67.[2] Qian H, Zhang J, Lai J, Yu W. A high-efficiency grid-tie battery energy storage

system. IEEE Trans Power Electron 2011;26(3):886e96.[3] Rodriguez J, Lai J, Peng FZ. Multilevel inverters: a survey of topologies, con-

trols, and applications. IEEE Trans Ind Electron 2002;49(4):724e38.[4] Hoffmann R, Mutschler P. The influence of control strategies on the energy

capture of wind turbines. In: Proceedings of the IEEE Industry applicationsConference; 2000. p. 886e93.

[5] Polinder H, Van der Pijl FFA, De Vilder GJ, Tavner PJ. Comparison of direct-drive and geared generator concepts for wind turbines. IEEE Trans EnergyConvers 2006;21(3):725e33.

[6] Abrahamsen F, Blaabjerg F, Pedersen JK, Thoegersen PB. Efficiency-optimizedcontrol of medium-size induction motor drives. IEEE Trans Ind Appl2001;37(6):1761e7.

[7] Li H, Chen Z. Design optimization and site matching of direct-drive permanentmagnet wind power generator systems. Renew Energy 2009;34(4):1175e84.

[8] Qiao W, Zhou W, Aller Jos�e M, Harley GR. Wind speed estimation basedsensorless output maximization control for a wind turbine driving a DFIG.IEEE Trans Power Electron 2008;23(3):1156e69.

M. Arifujjaman / Renewable Energy 74 (2015) 158e169 169

[9] Aarniovuori L, Laurila L, Niemela M, Pyrhonen J. Loss calculation of a frequencyconverter with a fixed-step circuit simulator. In: Proceedings of the EuropeanPower Electronics and Applications Conference; 2007. p. 1e9.

[10] Whitaker C, Newmiller J, Bower W. Converters performance certification:results from the Sandia test protocol. In: Proceedings of the PhotovoltaicEnergy Conversion Conference; 2006. p. 2219e22.

[11] Soltani F, Debbache N. Integration of converter losses in the modeling ofhybrid photovolatic-wind generating system. Eur J Sci Res 2008;21(4):707e18.

[12] Zeng Z, Chen Z, Blaabjerg F. Design and comparison of full-size converters forlarge variable-speed wind turbines. In: Proceedings of the European PowerElectronics and Applications Conference; 2007. p. 1e10.

[13] Chen Z, Spooner E. Wind turbine power converters: a comparative study. In:Proceedings of the International Power Electronics and Variable Speed DrivesConference; 1998. p. 471e6.

[14] Blaabjerg F, Jaeger U, Munk-Nielsen S. Power losses in PWM-VSI inverterusing NPT or PT IGBT devices. IEEE Trans Power Electron 1995;10(3):358e67.

[15] Helle L, Munk-Nielsen S. Comparison of converter efficiency in large variablespeed wind turbines. In: Proceedings of the IEEE Applied Power ElectronicsConference and Exposition; 2001. p. 628e34.

[16] Blaabjerg F, Pedersen JK, Jaeger U. Evaluation of modern IGBT-modules forhard-switched AC/DC/AC converters. In: Proceedings of the IEEE IndustryApplications Conference; 1995. p. 997e1005.

[17] Kraus R, Turkes P, Sigg J. Physics-based models of power semiconductor de-vices for the circuit simulator SPICE. In: Proceedings of the IEEE Power Elec-tronic Specialist Conference; 1998. p. 1726e31.

[18] Azar R, Udrea F, De Silva M, Amaratunga G, Wai Tung N, Dawson F, et al.Advanced SPICE modeling of large power IGBT modules. IEEE Trans Ind Appl2004;40(3):710e6.

[19] Nikzad MR, Radan A. Accurate loss modelling of fuel cell boost converter andtraction inverter for efficiency calculation in fuel cell-battery hybrid vehicles.In: 1st Power Electronic & Drive Systems & Technologies Conference(PEDSTC); 2010. p. 218e23.

[20] Ristow, Begovic M, Pregelj A, Rohatgi A. Development of a methodology forimproving photovoltaic inverter reliability. IEEE Trans Ind Electron2008;55(7):2581e92.

[21] Aten M, Towers G, Whitley C, Wheeler P, Clare J, Bradley J,K. Reliabilitycomparison of matrix and other converter topologies. IEEE Trans AerospElectron Syst 2006;42(3):867e75.

[22] Rohouma WM, Molokhia IM, Esuri AH. Comparative study of different PVmodules configuration reliability. In: Proceedings of the Ninth Arab Interna-tional Conference on Solar Energy (AICSE-9); 2007. p. 122e8.

[23] Sameer V, Michel T. Performance and reliability analysis of wind turbinesusing Monte Carlo methods based on system transport theory. In: Proceedingsof the AIAA Structural Dynamics and Materials Conference; 2005. p. 1e8.

[24] Military handbook MIL-HDBK-217F, reliability prediction of electronicequipment. Washington, DC: U.S. Dept. Defense, [415]; Dec. 2, 1991.

[25] Calleja H, Chan F, Uribe I. Reliability-oriented assessment of a DC/DC converterfor photovoltaic applications. In: Proceedings of the Power Electronics Spe-cialists Conference; 2007. p. 1522e7.

[26] Carrasco JM, Franquelo LG, Bialasiewicz JT, Galvan E, PortilloGuisado RC,Prats MAM, et al. Power electronic systems for the grid integration ofrenewable Energy sources: a survey. IEEE Trans Ind Electron 2006;53(4):1002e16.

[27] Guidelines to understanding reliability prediction. 24th ed. European PowerSupply Manufacturers Association (EPSMA); Jun. 2005.

[28] Pregelj BM, Rohatgi A. Impact of inverter configuration on PV system reli-ability and Energy production. In: Proceedings of the IEEE Photovoltaic Spe-cialists Conference; 2002. p. 1388e99.

[29] Lee S, Kim J, Lee J, Cho BH. State-of-charge and capacity estimation of lithium-ion battery using a new open-circuit voltage versus state-of-charge. J PowerSources 2008;185:1367e73.

[30] Johnson VH. Battery performance models in ADVISOR. J Power Sources2002;110:321e9.

[31] Idaho National Engineering & Environmental Laboratory. Battery test manualfor plug-in hybrid electric vehicles. Idaho Falls, ID, USA: Assistant Secretary forEnergy Efficiency and Renewable Energy (EE) Idaho Operations Office; 2010.

[32] Gjermund T, Nielsen R. Analytical equations for three level NPC converters. In:9th European Conference on Power Electronics and Applications; 2001. p. 1e7.

[33] Alemi P, Dong-Choon L. Power loss comparison in two- and three-level PWMconverters. In: IEEE 8th International Conference on Power Electronics andECCE Asia (ICPE & ECCE); 2011. p. 1452e7.

[34] Mestha LK, Evans PD. Analysis of on-state losses in PWM inverters. IEE Proc1989;136(4):189e95.

[35] Lewis EE, Hsin-Chieh C. Load-capacity interference and the bathtub curve.IEEE Trans Reliab 1994;43(4):470e5.

[36] Oettinger FF, Blackburn DL, Rubin S. Thermal characterization of powertransistors. IEEE Trans Reliab 1976;23(8):831e8.

[37] www.siliconfareast.com.[38] www.semikron.com.[39] Tirumala R, Imbertson P, Mohan N, Henze C, Bonn R, Efficient An. Low cost DC-

AC inverter for photovoltaic systems with increased reliability. In: Pro-ceedings of the Industrial Electronics Conference (IECON); 2002. p. 1095e100.

[40] Maish B, Atcitty C, Hester S, Greenberg D, Osborn D, Collier D. Photovoltaicsystem reliability. In: IEEE Photovolatic Specialist Conference (PVSC); 1997.p. 1049e54.

[41] Bower W. Inverters-critical photovoltaic balance-of-system components:status, issues and new-millennium opportunities. Prog Photovoltaics ResAppl 2000;8(1):113e26.