Control Strategies for Active Power Sharing in a Fuel Cell Powered ...

Control Strategies for Active Power Sharing in a FuelCell Powered Battery Charging StationZhenhua Jiang, Student Member, IEEE, and Roger A. Dougal, Senior Member, IEEEDepartment of Electrical EngineeringUniversity of South CarolinaColumbia, SC 29208, USAjiang@engr.sc.eduAbstract— This paper presents a system design for a fuel cellpowered battery charging station and control strategies for activepower sharing among the batteries. This battery charging stationallows multiple batteries to be charged simultaneously. Threecontrol strategies were investigated to coordinate the active powerdistribution among the battery charging branches. The baselinecontrol strategy was equal rate charging. Two advanced controlstrategies, proportional rate charging and pulse current charging,were compared to the baseline strategy. These control strategieswere realized in MATLAB/Simulink, and the current and voltageregulations were implemented using the classical proportionalintegralcontrol approach. The system simulation was conductedin the VTB by embedding Simulink objects of the controller andco-simulating with MATLAB. The experimental tests wereperformed by compiling Simulink codes of the controller anddownloading onto the dSPACE platform to control real hardware.The simulation and experimental results are given. Experimentaltests validate the control strategies.Keywords— active power sharing; battery; charging station;control strategy; dSPACE; fuel cell; Simulink; VTBI. INTRODUCTIONRechargeable batteries, such as lithium ion cells, are playingan increasingly significant role in the utilization of portableelectronic devices such as portable computers, cellular phonesand camcorders [1]. Their limited usable time makes it necessaryto develop some kind of portable battery charging system. Thefield application of a portable charging system may be far awayfrom the utility power systems. In this case, the fuel cell, which isemerging as one of the most promising technologies for thefuture power sources [2-3], may provide a good solution forpowering the portable battery charging station [4]. However, thelimited power supply of the fuel cell and the nonlinearity of bothfuel cells and lithium ion batteries [5-8] present difficulties forthe system design. Power converters are thus needed in this fuelcell/battery system and should be controlled properly for the bestsystem efficiency.In general, the battery charging station should allow multiplebatteries to be charged simultaneously. In order to meet thesimultaneous requirements of multiple users, power convertersare connected in parallel, each for one battery pack. The initialstates of the batteries may be different. A battery with a lowerinitial state-of-charge may require a larger charging current orotherwise a longer charging time. Furthermore, when chargingadvanced technology batteries such as lithium ion cells, it ishazardous to exceed some certain current or voltage limits.Therefore, the power from the fuel cells should be distributedefficiently among these charging branches and the powerconverters should be regulated appropriately. Whereas, thepower distribution from a nonlinear and current-limited powersource presents some challenges for control algorithm design.Obvious difficulty also arises from the controller design ofpower converter because the source and load of the powerconverter are strongly nonlinear and dynamic.This paper presents a fuel cell powered battery chargingstation and control strategies for active power sharing among thebatteries. The remainder of this paper is organized as follows.The system architecture of the fuel cell powered battery chargingstation is proposed in Section II. Section III describes threecontrol strategies that were investigated to coordinate the powerdistribution among the battery charging branches. The baselinecontrol strategy was equal rate charging. Two advanced controlstrategies, proportional rate charging and pulse current charging,were compared to this baseline strategy. Section IV presents theimplementation of the charge controllers, which were designed,based on these control strategies, in MATLAB/Simulink for bothsystem simulation and experimental tests. The system simulationwas conducted in the VTB by embedding Simulink objects of thecontroller and co-simulating with MATLAB. The simulationresults are given in Section V. The experimental tests wereperformed by compiling Simulink codes of the controller anddownloading onto a dSPACE platform to control real hardware.Section VI demonstrates the experiment results and validates thecontrol strategies. Conclusions are made in Section VII.II.SYSTEM ARCHITECTURE DESIGNSince the battery charging station is configured for multiplebatteries to be charged simultaneously, many individualcharging channels can be built in the charging station. Here weconsider three charging channels, which can represent thegeneral solution of many charging channels. Assume here thatall the batteries are put in the charger. The case that somebatteries are inserted or retrieved during charging is investigatedin [9].The block diagram of the fuel cell powered battery chargingstation is shown in Figure 1. A fuel cell stack, which is the powerThis work was supported in part by the US Army CECOM under contractNRO-00-C-0134, and by US ONR under contract N00014-00-1-0368 and undercontract N00014-00-1-0131.

generation system, is used to charge up to three lithium ionbattery packs through three separate buck converters. Eachbattery contains four series-connected cells. The buck converterscontrol the charging current and voltage supplied to each battery,and allocate the available power among the batteries accordingto one of several control strategies that are described here.Figure 1. Block diagram of the system architecture.As shown in Figure 1, three buck converters are connected inparallel to a single fuel cell power source. The power availablefrom the fuel cell is distributed among the three batteries,according to (1).P fc P1 + P2+ P3= (1)where P 1 , P 2 , and P 3 are the power to three charging channelsrespectively, and P fc is the power from the fuel cell.In practice, the power distribution among the batteries isrealized by regulating the charging currents of the batteries. Thefollowing equation relates the current from the fuel cell to threecharging currents, assuming no power loss in these powerconverters.I fc d1I1+ d2I2 + d3I3= (2)where I 1 , I 2 , and I 3 are the charging currents to three batteriesrespectively, I fc is the current from the fuel cell, and d 1 , d 2 and d 3are the duty cycles of three buck converters, respectively, andthey have values between 0 and 1.The sum of the right hand side in (2) should be less than themaximum current available from the fuel cell. Considering thatvariations of the voltages of the fuel cells and batteries are notlarge, the duty cycle will vary within a limited small range,which means, for instance, values of d 1 , d 2 and d 3 are between0.70 and 0.75. Based on this assumption, we can take thefollowing expression as a criterion for power distribution amongthe batteries.I1 I2+ I3≤ Ilim+ (3)where I lim is a preset limit for the total charging current that canbe estimated according to the maximum current available fromthe fuel cell and the average duty cycle of power converters.Equation (3) gives a basic requirement for the design of controlstrategies for active power sharing.III. CONTROL STRATEGIES FOR ACTIVE POWER SHARINGDifferent users may have different requirements for chargingtheir batteries. Some people may require that the batteries befully charged within the shortest period of time, while otherswish to maximize the cycle life of their batteries, or minimize thefuel consumption during the charging process. This paper aimsto discover the appropriate control schemes for minimizing thecharging time. Three control strategies were investigated tocoordinate the power distribution among the battery branchesand they are equal rate charging, proportional rate charging andpulse current charging. The first one is the baseline controlstrategy, and the other two are compared with it.Among these control strategies for active power sharing, twokinds of charging protocols may be used. They are constantcurrent-constant voltage (CC-CV) charging protocol and pulsecurrent charging protocol. CC-CV charging protocol can help toprotect the battery from overcharging. Under this protocol, thebattery is charged to an end potential using a constant current.The potential is then held constant after this potential is reached,and the charging current gradually decreases. The chargingprocess terminates when the current reaches a preset small valueduring the constant voltage mode. Under pulse current chargingprotocol, a pulse current with a period of T and pulse-duration ofTon is applied to the battery. Pulse current charging protocol hasbeen shown to enhance charging rate capability and to preventthe increase of internal impedance of the battery, thus reducingthe total charging time [8].A straightforward control strategy for charging all of thebatteries simultaneously using constant currents is to equallydistribute the charging current among them. Due to differencesin the initial states of the batteries, some batteries may reach thereference voltage earlier than the others may. When one batteryreaches the voltage limit, its voltage will be held constant and itscharging current will taper. The rest of the total available currentwill be distributed equally between the other two batteries. Thenthe same scheme is followed for the remaining batteries until allof the batteries are held in the constant voltage mode. Thisstrategy, called equal rate charging, is the baseline controlstrategy for active power sharing. This control strategy isillustrated in Figure 2-a, and it can be easily implemented on thehardware. The charging time will be as long as it takes to chargethe most depleted battery, and, in particular, it may be muchlonger than the charging time for the less depleted battery tobecome fully charged when the initial states of charge of thebatteries are widely disparate.A more time-efficient control strategy can take into accountthe fact that the charge that any battery will need to become fullycharged is proportional to its depth of discharge where the depthof discharge is defined as unity minus the state of charge of thisbattery. As implied in the previous control strategy, if continuouscurrents of the same magnitude are applied to charge differentbatteries, the charging times will be approximately proportional2

to the depth of discharge (neglecting nonlinearity of thebatteries). On the other hand, if we want all the batteries tobecome fully charged during the same period of time, thecharging current of each battery can be made proportional to thefraction of the depth of discharge of this battery, which can becalculated according to (4)1−SOCiIi = Ilim ⋅, i = 1,2,3 (4)3∑ (1 − SOC )i=1where I i is the charging current of the ith battery, I lim is the totalavailable charging current, and SOC i is the initial state-of-chargeof the ith battery.This control strategy is called proportional rate charging,which is shown in Figure 2-b. Under this strategy, the batteriesmay become fully charged almost simultaneously. However, it isdifficult to estimate the state of charge. For lithium ion batteries,an approximate relationship between the state-of-charge andopen circuit voltage can be found when the state-of-charge is notwithin the extreme range, i.e., if the state-of-charge is between0.1 and 0.9. When the state of charge is greater than 0.9, thebattery charger is usually working under constant voltage mode.The charging current is not determined by the charge controlleritself, but by the internal potential and terminal voltage of thebattery. It is not necessary to estimate the state-of-charge sincethe information about state-of-charge is useful only when thecharger works under current regulation mode. When the state ofcharge is less than 0.1, the estimation is cut off to 0.1. In this case,the estimated value for depth of discharge is 0.9, and the fractionof the depth of discharge is very close to that using a moreaccurate value of the state of charge. Therefore, the initialstate-of-charge can be estimated by measuring the battery opencircuit voltage and linearly fitting the open circuit voltage to thestate-of-charge. As we will show, even this simpleapproximation is sufficient for defining the current sharing.In addition to DC charging, the third control strategy is pulsecurrent charging which uses the pulse current charging protocolwhile the previous two control strategies use CC-CV chargingprotocol. With this strategy, three pulse currents with the sameperiod of T and different pulse-durations are applied to threebatteries alternatively. The sum of the pulse-durations of thethree pulse currents is equal to the period. This control strategy isillustrated in Figure 2-c. A strategy similar to proportional ratecharging can be defined for pulse current charging by making theduty cycle of each pulse current proportional to the fraction ofthe depth of discharge of this battery. This duty cycle can beestimated according to the following equation.i1−SOCidi =, i = 1,2,3(5)3∑ (1 − SOCi)i=1where d i is the duty cycle of the charging current of the ithbattery.Under this strategy, the charging current can be relativelylarge because only one battery draws this current at any time.Therefore, it can be the limiting current available from the fuelcell as long as it does not exceed the current limit for any battery.With this strategy, it is also possible for all the batteries tobecome fully charged almost simultaneously. Nevertheless, thedisadvantage is that the control algorithm to implement thisstrategy on the hardware is more complicated compared to DCcharging because the output currents of three power converterschange frequently resulting in frequent changes in the voltagesof the batteries, an undesired fluctuation in the output voltage ofthe fuel cell, and high dynamics of the system.(a) Equal rate charging(b)Proportional rate charging(c) Pulse current chargingFigure 2. Illustration of three control strategies for active power sharing.The detail of implementing the three control strategies asshown in Figure 2 is explained as follows. For the convenienceof demonstration, assume that initially V1>V2>V3 where V 1 , V 2 ,V 3 are the voltages of three batteries respectively. In thesestrategies, I lim is the limit of the total charging current; V ref isbattery end potential, which is usually 4.2V for each cell; and I 1 ,I 2 , I 3 are the charging currents of three batteries respectively.(Here we assume that the maximum current available from thefuel cell, I lim , is less than the sum of the maximum safe chargingcurrents for all of the batteries.)A. Strategy 1: Equal rate charging (Baseline)• If V 1

• If V 2 = V ref , I 2 is tapering, and I ref,3= I lim - I 1 - I 2• If V 3 = V ref , I 3 is tapering• If I i

V. SIMULATION RESULTSIn order to compare the performances of three controlstrategies for active power sharing, simulation studies were firstconducted in the VTB [10]. Figure 4 shows the VTB schematicview of the system shown in Figure 1. The power source is a25-cell PEM fuel cell stack. Each battery is a 4X1 (series byparallel connections) array of lithium ion cells. The capacity ofeach battery is 1500mAh. The initial states of charge of thebatteries are 0.7, 0.5, and 0.3, respectively. Each power converteris implemented by a switching-average buck converter model inseries with a low pass filter.Figure 4. Schematic view of the fuel cell powered battery charging station.The controller is implemented in the Simulink models asshown in Figure 3. It is embedded in VTB through aVTB-Simulink interface and it can co-simulate with VTBinteractively. The charging currents and battery voltages aresensed and fed to the controller. The controller outputs three dutycycles to the power converters and three commands for theswitches. The proportional and integral gains for currentregulation are 0.05 and 0.002 respectively. The proportional andintegral gains for voltage regulation are 0.01 and 0.001respectively. In the studied DC charging strategies, the sum ofthe total available charging currents is limited to 2 Amperes.While the reference charging currents are set 0.66A for all of thebatteries under the baseline strategy, they are 0.38A, 0.66A, and0.96A, respectively, under proportional rate charging strategy.The limit of the charging currents is 1.5A for pulse currentcharging strategy taking the safe charging current of the batteriesinto consideration. The duty cycles of the charging currents are20%, 33.3% and 46.7%, respectively, for three batteries. Thecharging process stops when the battery voltage during thelow-pulse-duration reaches the reference voltage.This system is simulated with the three different chargingstrategies for 2 hours (7200 seconds). The simulated chargingcurrents and states of charge of the batteries under three chargingstrategies are shown in Figures 5-a, b, 5-c, d, and 5-e, frespectively.Obviously, with the baseline strategy, the state-of-charge ofthe battery with highest initial voltage increases more rapidlythan the others. It is seen from Figures 5-d and 5-f that, withproportional rate charging strategy and pulse current chargingstrategy, the state-of-charge of the battery with lowest initialvoltage increases more rapidly than the others. Therefore, thesetwo strategies can help reduce the total charging time. From thesimulation results, the following conclusions can be drawn forthe three control strategies.1) While the battery with the highest initial state-of-chargebecomes full fastest with the baseline charging strategy, the totaltime to charge all the batteries is the longest if there are apparentdifferences in the initial state-of-charge. It is seen from Figure5-a that it takes about 4050 and 6900 seconds for the first and lastbatteries, respectively, to become full.2) Compared with the baseline strategy, the total chargingtime with proportional rate charging strategy is shorter. It isshown in Figure 5-c that it takes about 4050 and 6900 secondsfor the first and last batteries, respectively, to become full.3) Among three control strategies, pulse current chargingstrategy achieves the shortest total charging time because theindividual charging current is relatively large. It takes about5400 seconds for all the batteries to become full, about 500seconds earlier than the baseline strategy and 1500 secondsfaster than proportional rate charging strategy.4) It can be seen that all of the batteries can become fullycharged almost simultaneously when they are charged with anappropriate set of proportional-rate continuous currents or pulsecurrents. Therefore, it is true that proportional rate chargingstrategy and pulse current strategy are effective for charging thebatteries.From above, we can see that proportional rate chargingstrategy is superior to the baseline strategy in respect to reducingthe total charging time. In order to give a wider view of thisadvantage, the simulation is run five times each at nine discreteinitial states to compare the total charging time with these twostrategies under different initial conditions. Figure 6 shows theplots of the total charging time against the averagestate-of-charge of three batteries under equal rate chargingstrategy and proportional rate charging strategy for differentinitial battery states. The states of charge of the three batterieshave equal difference and the average state-of-charge is theiralgebraic mean value. It is seen from Figure 6 that the totalcharging time decreases with the average state-of-charge in spiteof the control strategy or initial states of the batteries. For everyinitial average state-of-charge, the total charging time withproportional rate charging strategy is shorter than that with equalrate charging strategy. The total charging time increases with thedifference in the initial state-of-charge with both strategies. Thereason for proportional rate charging strategy is that the error ofthe initial state-of-charge estimation is bigger when theirdifferences increase. For equal rate charging strategy, this isbecause it takes a longer time to charge the most depleted batterywhen the difference is bigger. When all batteries have theidentical initial state-of-charge, both strategies have the sameeffect on the current sharing and the total charging time isidentical.5

(a) charging currents(b) states of charge(c) charging currents(d) states of charge(e) charging currents(f) states of chargeFigure 5. Charging currents and states of charge under three control strategies: (from top to bottom) equal rate charging, proportional rate charging, and pulsecurrent charging.6

#3 were 0.75, 0.83, and 0.92 respectively. Their initial opencircuit voltages were 15.9V, 16.1V, and 16.4V, respectively.The initial state-of-charge of the battery was measured using thefollowing approach. Each battery was discharged to fulldepletion. A constant current was then applied to charge thebattery until it was full and the total charging time was recorded.After fully depleting this battery, charging this battery with thesame current for a proportion (equal to the initial state-of-chargein magnitude) of the total charging time could obtain a desiredinitial state-of- charge.Figure 6. Plots of the total charging time against the average state-of-chargeof three batteries with equal rate charging strategy and proportional ratecharging strategy under different initial battery states: 1. with proportional ratecharging strategy and a maximum initial state-of-charge differennce of 20%, 2.with proportional rate charging strategy and a maximum initial state-of-chargedifferennce of 40%, 3. with equal rate charging strategy and a maximum initialstate-of-charge differennce of 20%, 4. with equal rate charging strategy and amaximum initial state-of-charge differennce of 40%, 5. with either of thecharging strategies and the same initial state-of-charge.VI. EXPERIMENT VALIDATIONNext, these control strategies were validated with realhardware. A prototype of the fuel cell powered battery chargingstation was built using an H-Power D35 PEM fuel cell stack asthe power source. This stack has a nominal power capacity of35W and nominal 24V open circuit voltage. Three 4-cellPanasonic lithium ion batteries were used. The nominal capacityof each battery is 1500mAh. Three buck converters were builton one single board to distribute the charging current. The blockdiagram of the experiment environment is shown in Figure 7.The charging algorithms implementing the control strategiesreside on a general-purpose dSPACE real-time controller board,which also houses the hardware interface consisting ofmulti-channel A/D and D/A converters. The charging controlalgorithms are designed and implemented using MATLAB/Simulink and the codes are then compiled and dropped onto adSPACE DS1103 PPC controller board to control the realhardware. The charging currents and battery voltages aremonitored and input to the dSPACE controller board throughthe A/D converters mounted on it. The power source bus voltageis also an input variable for monitoring purpose. The real-timecontroller provides the switch duty commands to each buckconverter. The circuit protection function is also implementedwithin the software.Tests were conducted using the proportional rate chargingstrategy. In order to ensure that each charging current wouldnever exceed the safe maximum charging current (that is800mA for the batteries used in the experiment), the totalcharging current was scaled down by half to 1.0 Ampere. For theconvenience of demonstration, three batteries were numbered#1, #2, and #3. The initial state-of-charge of batteries #1 throughFigure 7.Block diagram of experiment environment.According to the initial states-of-charge of the batteries, thecontroller selected the charging currents as 0.5A, 0.35A, and0.15A for batteries #1 through #3 respectively. The measuredbattery voltages and charging currents are shown in Figure 8,where the simulated voltages and currents are also plotted.Small differences in the voltages between simulation results andexperiment data were observed and this was due to the fact thatmore detailed transients were not characterized in the batterymodel. The ripples in the currents were a little large and this wasbecause current transducers with low resolution were used tomeasure the charging currents. Although then, it can be said thatthe simulation results matched experiment data very well. FromFigure 8, it is seen that the battery with the lowest initial voltage(and thus the least initial charge) was charged with the highestcurrent, and its voltage increased more rapidly than the others.This was also predicted by the simulation. This feature impliesthat proportional rate charging strategy can help to reduce thetotal charging time. After a little while of charging, the voltagesof batteries #1 and #2 exceeded the voltage of the third battery.The voltages of these two batteries reached the reference voltagealmost simultaneously, and then their currents tapered. Whenthese currents approached to 0.15A, the third battery reached thereference voltage and its current tapered. It is also seen that all ofthree batteries became fully charged almost simultaneously.7

Voltage(V)16.816.716.616.5Simulation Result (#1)Simulation Result (#2)Simulation Result (#3)Experiment Datawere used for both system simulation and experimental runtime.Simulation results show that, with equal rate charging strategy,the battery with the highest initial state-of-charge becomes fullfastest but the total charging time is longest. Proportional ratecharging strategy is superior to equal rate charging strategy inrespect to the total charging time. The total charging timeincreases with the difference in the initial state-of-charge of eachbattery with both strategies. The total charging time is minimumwith pulse current charging strategy. It is possible for all thebatteries to become full almost simultaneously when they arecharged with appropriate proportional constant currents or pulsecurrents. Experimental tests validated these control strategiesand simulation results.16.40 1000 2000 3000 4000 5000 6000Time(s)(a) From top to down (on the left vertical axis): voltages of batteries #3, #2, and#1Current(A)0.50.40.30.20.1Simulation Result (#1)Simulation Result (#2)Simulation Result (#3)Experiment Data00 1000 2000 3000 4000 5000 6000Time(s)(b) From top to down: charging currents for batteries #1, #2, and #3Figure 8. The measured battery voltage voltages and charging currents:(a) voltages, (b) currents.VII. CONCLUSIONSThis paper presents an effective system design for a fuel cellpowered battery charging station. Three control strategiescoordinating active power sharing among multiple batterybranches were developed and compared. TheMATLAB/Simulink definitions of the battery charge controllersACKNOWLEDGMENTThis work was supported in part by the US Army CECOMunder contract NRO-00-C-0134, and by US ONR under contractN00014-00-1-0368 and under contract N00014-00-1-0131.REFERENCES[1] R. J .Brodd, “Overview: rechargeable battery systems”, WESCON'93.Conf. Record, pp. 206 –209, 1993.[2] B. Rohland, J. Nitsch, and H. Wendt, “Hydrogen and fuel cells ¯¯ theclean energy system”, Journal of Power Sources, Vol. 37, No. 1-2, pp.271-277, January 1992.[3] A. Heinzel, C. Hebling, M. Müller, M. Zedda and C. Müller, “Fuel cellsfor low power applications”, Journal of Power Sources, Vol. 105, No. 2,pp. 148-153, March 2002.[4] Z. Jiang and R. Dougal, “Control design and testing of a novelfuel-cell-powered battery-charging station”, Proceeding of IEEE AplliedPower Electronic Conference, Miami, Florida, Feb. 10-14, 2003.[5] M. M. Bernardi, and M. W. Verbrugge, “A mathematical model of thesolid-polymer-electrolyte fuel cell”, Journal of Electrochemical Society,Vol. 139, No. 9, pp. 2477-2491, September 1992.[6] J. Kim, S. M. Lee, and S. Srinivasan, “Modeling of proton exchangemembrane fuel cell performance with an empirical equation”, Journal ofElectrochemical Society, Vol. 142, No. 8, pp. 2670-2674, August 1995.[7] L. Song and J. W. Evans, “Electrochemical-thermal model of lithiumpolymer batteries”, Journal of Electrochemical Society, Vol. 147, No. 6,pp. 2086-2095, 2000.[8] J. Li, E. Murphy, J. Winnick and P. A. Kohl, “The effects of pulsecharging on cycling characteristics of commercial lithium-ion batteries”,Journal of Power Sources, Vol. 102, No. 1-2, pp.302-309, December2001.[9] Z. Jiang and R. Dougal, “Design, rapid prototyping, and testing of afuel-cell-powered battery-charging station”, unpublished.[10] T. Lovett, A Monti, E. Santi, R. Dougal, “A multilanguage environmentfor interactive simulation and development of controls for powerelectronics”, Proceedings of IEEE 32nd Annual Power ElectronicsSpecialists Conference, Vol. 3, pp. 1725 –1729, 2001.8