Characterisation, analysis and technical solutions of a biomass-gas fed PEM fuel cell system

Transcription

1 MASTER Characterisatin, analysis and technical slutins f a bimass-gas fed PEM fuel cell system Geers, M.P.A. Award date: 2006 Link t publicatin Disclaimer This dcument cntains a student thesis (bachelr's r master's), as authred by a student at Eindhven University f Technlgy. Student theses are made available in the TU/e repsitry upn btaining the required degree. The grade received is nt published n the dcument as presented in the repsitry. The required cmplexity r quality f research f student theses may vary by prgram, and the required minimum study perid may vary in duratin. General rights Cpyright and mral rights fr the publicatins made accessible in the public prtal are retained by the authrs and/r ther cpyright wners and it is a cnditin f accessing publicatins that users recgnise and abide by the legal requirements assciated with these rights. Users may dwnlad and print ne cpy f any publicatin frm the public prtal fr the purpse f private study r research. Yu may nt further distribute the material r use it fr any prfit-making activity r cmmercial gain

2 U.le!echnische universi!ei! eindhven Capaciteitsgrep Elektrische Energietechniek Electrmechanics & Pwer Electrnics Master fscience Thesis Characterisatin, Analysis and technical slutins fa bimass-gas fed PEM fuel cell system M.P.A. Geers EPE.26.A.7 The department Electrical Engineering fthe Technische Universiteit Eindhven des nt accept any respnsibility fr the cntents fthis reprt Caches: r. P.J.H. Wingelaar Dr. J.L. Duarte r. M.A.M. Hendrix Octber 2006 faculteit elektrtechniek

3 Abstract Plymer Electrlyte Membrane (PEM) fuel cells can be used as an alternative pwer surce. These fuel cells ften use platinum as catalyst and require hydrgen t prduce energy. Hydrgen can be prduced in varius ways, such as gasificatin f bimass and via electrlysis. A byprduct f gasificatin is carbn mnxide (CO). CO adsrbs easily t platinum, causing an bstructin fthe catalyst in the fuel cell, thus reducing its perfrmance. The gal f the research is t regenerate a CO pisned PEM fuel cell by means f electrlysis. This is dne by applying a negative vltage pulse n a pisned cell. A cnverter is designed and built t generate these pulses and perfrmed within design specificatins. t is measured that tw ut f fur cells f ne PEM cartridge can be regenerated by applying negative vltage pulses t the cells. Hwever, it is uncertain whether the prcess respnsible fr the regeneratin is electrlysis. The ther cells remained unaffected. t is als determined that perids fn lad help in the regeneratin fpisned cells.

4 Table f cntents Abstract Table fcntents V ntrductin. Prject backgrund.2 Electrchemistry basics 2.2. Ande and Cathde Open circuit vltage / electrmtive frce (emt) 2.3 The PEM Fuel cell. 3.4 CO pisning 5.5 Electrlysis 6 2 Pwer electrnic cnverter 7 2. System descriptin Cmpnent calculatin The transfrmer rati n The magnetising inductr L m The utput capacitr C Leakage inductance Snubber Bi-directinal pwer transfer Design cnsideratins fr the transfrmer Cnverter cmpnent design and specificatin Cnverter simulatin 20 V

5 2.6 Cntrl Additinal circuitry Relay switching set-up Cnverter implementatin, tests and results Cnverter testing methds and results Output characteristics fr different lads Output pwer characteristic Reversed cnverter peratin 30 3 Fuel cell regeneratin The measurement setup Reversing plarity Vlt. Secnd mismatch Vltage current characteristics, CO pisning Pulsing with negative vltages Timed pulses n cell Timed pulses n cell Timed pulses n cell Timed pulses n cell Timed pulses at 6A Cnclusins and discussin 52 4 Cnclusins and recmmendatins Cnclusins Recmmendatins 53 V

6 5 Acknwledgement 54 6 Literature 55 Appendix A : Derivatin ftransfrmer equatins A. Appendix B : PSM schematic fr cnverter simulatin B.2 Appendix C : Additinal circuitry C. Gate drivers C. C. C.2 : Measurement circuitry C.2 C.2. : Current measurement C.2 C.2.2 : Vltage measurement C.2 C.2.3 : Anti-aliasing filters C.4 C.3 Relay circuitry C.6 Appendix D : List fequipment.. D. Appendix E : Synthetic gas specificatin E.2 Appendix F : Bundary cnditins and limitatins F. Appendix G : Measurement results G.2 Addendum : Pssible explanatin fr cunter pulse V

7 ntrductin This reprt describes the subject and results f my master thesis: Characterisatin, Analysis and technical slutins fa bimass-gas fed PEMfuel cell system. The wrk has been dne in the grup Electr mechanics and Pwer Electrnics (EPE) in the department f Electrical Engineering fthe Eindhven University f Technlgy (TU/e). Within the grup, research is cnducted n mdelling steady-state and dynamic behaviur f a Plymer Electrlyte Membrane (PEM) fuel cell. Recently, small- and large signal behaviur was cmbined with steady-state characteristics in ne mdel. The parameters fthis mdel can be fund using Electrchemical mpedance Spectrscpy ( S), step-respnses and steady-state characteristics []. The research prject includes the use f synthetic gas, simulating the utput f a bimass gasificatin system. The main gas cmpnent f this gas mixture is hydrgen (H 2 ), the fuel fr a PEM fuel cell. Hwever, it als cntains trace amunts f carbn mnxide (CO). One f the disadvantages f using PEM fuel cells is that the platinum catalyst is extremely sensitive t CO pisning, causing a reversible deteriratin fthe verall perfrmance [7]. The gal f this master prject is t regenerate a PEM fuel cell frm CO pisning by means f peridically electrlysing the water cntained inside the membrane, with a pwer electrnic cnverter. The existing electrical mdel f the system shuld be adjusted based n the results gathered.. Prject backgrund With the rise f il prices and the call fr sustainable energy surces, hydrgen is cnsidered as a prmising alternative fuel. Electricity can be generated frm hydrgen and xygen in a fuel cell with water as by-prduct, hence this technlgy has the ptential t be very envirnmentally friendly [25]. The fuel cell is starting t rival the internal cmbustin engine (CE) mre and mre in terms f pwer density. This makes it a viable alternative pwer surce in decentralised pwer generatin units and in transprtatin applicatins. Hwever, there is n infrastructure available fr large scale hydrgen generatin, distributin and strage at the present. Therefre, it is unlikely that pure hydrgen will be used as a large scale fuel n the shrt term. Hwever, hydrgen can be prduced with n-bard hydr-carbn refrmers. The majr prducts will be carbn dixide (C0 2 ) and hydrgen (H 2 ) [2].

8 .2 Electrchemistry basics The fuel cell is an electrchemical device. n rder t understand sme f the thery presented in this reprt, briefdescriptins fseveral electrchemical cncepts are presented here..2. Ande and Cathde n electrchemistry there is a cnventin where the electrde at which electrns are released (where xidatin takes place) is called the ande. The electrde where electrns are used (where reductin takes place) is called the cathde [3]. This translates int electrical engineering terms as: the ande being the electrde where the electrical current is flwing int and the cathde the electrde where the current is riginating frm..2.2 Open circuit vltage electrmtive frce (emf) n an electrchemical cell, an electrical ptential is created between tw dissimilar metals. This ptential is a measure f the energy per unit charge which is available frm the xidatin/reductin reactins t drive the reactin. The cell ptential (emf) has a cntributin frm the ande, which is a measure f its ability t lse electrns, its "xidatin ptential". The cathde has a cntributin based n its ability t gain electrns, its "reductin ptential". The cell ptential is defined as the sum fthe xidatin and reductin ptentials (.) [4). (. ) Hwever, the individual electrdes cannt be determined in islatin. Therefre, the ptential f the reactin at an electrde is measured with respect t a standard hydrgen electrde (SHE) [4). This is cnsidered t be the electrchemical equivalent f "grund" r "mass" in electrical engineering terms. The reactin that takes place at the SHE is (.2) 2H+ (aq) + 2e- p. H 2 (g). (.2) fthe reactin is reversed, this ptential is necessary t let the reactin ccur. 2

9 .3 The PEM Fuel cell A fuel cell is a device that can prduce electricity frm chemical reactins. A plymer electrlyte membrane (PEM) fuel cell, als knwn as a prtn-exchange membrane fuel cell, cnsists primarily ffive parts. Figure - shws these parts schematically. Metal plate a (ande) is the negative terminal f the fuel cell, metal plate e (cathde) is the psitive terminal f the fuel cell. Plates band d are diffusin layers, cnsisting f carbn with a thin layer f platinum catalyst. Finally, plate c is the plymer membrane in a rubber gasket [5J. Tw chemical equatins describing the ande (.3) and cathde (.4) reactin result in the verall reactin fthe PEM fuel cell, (.3) (.4) (.5) Figure -2 shws these reactins in a schematic representatin f a fuel cell. Fur prtns (H+) and fur electrns (e-) are created frm tw hydrgen mlecules, under the influence f the platinum catalyst [6J. The membrane has the prperty that nly the hydrated prtns (r hydrnium ins) are let thrugh. The electrns will thus mve thrugh the ande and via an electrical circuit t the cathde side. The prtns will attach t water mlecules, frming hydrnium ins (H 3 0+) which transprt the prtns thrugh the membrane t the cathde side, thus freeing a catalyst site fr cntinuatin f the reactin. On the cathde side, xygen (0 2 ) is split int tw ins which attach t the platinum catalyst. After that, the electrns and prtns cmbine with the xygen ins t frm tw water mlecules [6J. A practical fuel cell device cnsists ut f stacks f individual fuel cells. The individual cells are placed in series t btain a higher utput vltage. Several f these stacks can again be placed in series t get an even higher vltage. During this prject, cartridges are used which cntain a stack ffur cells. 3

10 Figure -: Schematic representatin fa PEM fuel cell. Electrn-+ Flw t Lad Hydrgen ~",e 0 ~:.00 Ande Electrlyte Cathde Figure -2: A schematic verview f a PEM fuel cell 4

11 .4 CO pisning The majr prducts f a hydr carbn refrmer are COz and Hz. Hwever, it als cntains carbn mnxide (CO) (a few percent). The CO adsrbs well t the platinum catalyst. Due t which, the fuel cell will be mre and mre hindered in its peratin until n pwer can be drawn frm it, hence the term "pisning" [7]. This hindrance is caused by ccupatin f free catalysis sites by the CO. The adsrbed CO als lwers the hydrgen reactivity via diple interactins and electrn capture [7]. Several methds have been devised t free the catalysis sites f adsrbed CO. These include the xidatin f the CO and raising the temperature. Raising the temperature is nt very practical. A higher cell temperature leads t greater water lss t the exhaust gas streams as mre water evaprates, causing drying fthe prus media inside the cell [8]. This in turn increases the inic resistance, which gives rise t the vltage drp due t hmic lsses [9]. There are, hwever, several ther knwn methds t xidise adsrbed CO. The first is by adding pure xygen t the fuel stream [0], [], [2]. This intrduces a gas mixture in which chemical xidatin fhydrgen and CO can take place. Only a small amunt fthe xygen in the fuel stream reacts with the adsrbed CO, causing the fuel cell t heat up as ther parts f the added xygen react with the hydrgen, thus reducing efficiency even mre. The secnd methd is by raising the ande ptential. [f the ande ptential is high enugh an electrchemical xidatin f CO takes place. t is hwever unpractical t perate a PEMFC at such high ande ptentials, as it has a very lw efficiency at that perating level [7], [2]. t is, hwever, pssible t lwer the ptential required t induce this reactin. This can be accmplished by adding ruthenium (Ru) t the platinum catalyst. This lwer activatin ptential can have the additinal effect f self-xidatin [3]. This is a phenmenn where the ande ptential rises due t the CO pisning, causing the xidatin f CO. An scillatry behaviur will becme visible as the ptential will return t nrmal values after xidatin and rises again till the next xidatin takes place. This phenmenn has been the basis fr anther methd t xidise the adsrbed CO. By briefly applying a pulse in the lad f the PEMFC, the ande ptential will be briefly raised high enugh t xidise the CO. The perfrmance f a PEMFC fed with Hz/CO fuel can be greatly increased by peridically pulsing the lad. The cnsequences f this might be that the ttal system cmplexity can be reduced, as the refrmat gas des nt have t be as pure as befre [4], [5], [6], [7]. 5

12 .5 Electrlysis Electrlysis is an electrchemical redx reactin. This implies that it cnsists ut f tw linked reactins. Electrns will be released at ne electrde and used at the ther. These tw redx half reactins are (.6) and (. 7) (.6) E: ed =-O.83V, (.7) (.8) Here (.6) and (.7) are respectively the ande and cathde reactins [8]. t is clear that xygen is prduced at the electrde where the electrical current is entering the system. The verall reactin (after an acid-base reactin has taken place) is then (.8). With the fuel cell in nrmal peratin (generatr mde) the reactins (.3) and (.4) take place at respectively the ande and cathde. The hydrgen fuel is used at the ande (.3), which implies that the CO cntaminatin will als take place at the ande. n rder t generate xygen at this ande by means f electrlysis, the electrical current shuld keep the same directin, whereas the cell vltage shuld be reversed. This can be accmplished by putting a vltage r current surce in parallel with the affected cell. There shuld thus be the ability t switch this surce in parallel with a specific cell and this surce shuld be able t pulse the current r vltage. The electrlysis reactin described abve is based n a base-reactin. The fuel cell membrane, hwever, is assumed t wrk via acid reactins. t is therefre expected that the electrlysis will take place via the inverted fuel cell reactins. Thus reactins (.3) and (.4) but frm the right t the left. 6

13 2 Pwer electrnic cnverter The thery presented in this chapter related t the design f the fly-back cnverter is primarily basedn the therypresented in [9J and[20]. A pwer electrnic cnverter can be used t peridically apply pulses t the PEM fuel cell, in rder t induce brief perids f electrlysis. This cnverter can be placed in parallel with ne f the cells f a fuel cell stack. t is therefre necessary that the utput f this cnverter is nt grunded. An islated fly-back tplgy cmplies with these bundary cnditins. The fly-back is a ppular tplgy as it requires very few pwer electrnic cmpnents. A fly-back cnverter is a mdified Buck-bst cnverter, where the inductr is replaced with a pair f cupled inductrs, r fly-back transfrmer as is shwn in Figure 2-. Here, the fly-back transfrmer is depicted by an ideal transfrmer with a magnetising inductr L m in parallel with the primary transfrmer cil. The cnverter wrks by turning the primary switch Sl n, causing the vltage Us t be applied ver the inductr L m This causes a current t flw with a cnstant nn-zer slpe. During this time, the dide n the secndary side will be negatively biased and n current will flw thrugh it. As a result, energy will be stred in the transfrmer. After a time, the switch will be turned ff and the dide at the secndary side will becme psitively biased. The stred energy will then be released via the secndary side f the transfrmer t the utput capacitr and the lad. Nl N2 Us C -J; lad S:J D Figure 2-: deal islated fly-back cnverter 7

14 The cmpnent values f the cnverter can be derived frm a set f desired vltages and currents. The analysis fthe circuit is split up int tw parts. The first is when the primary switch is n and the secnd when it is ff. Cntinues Cnductin Mde (CCM) is assumed. This means the current thrugh the magnetizing inductr will never reach zer. Furthermre, the input and utput vltages will be assumed cnstant and the circuit t be at steady-state. 2. System descriptin The perid where the primary switch is turned n, will be referred t as "Mde " and the perid where it is turned ff as "Mde ". The duty-cycle D describes the lengths f Mde and as a functin fthe switching perid, with tn the n-time and ', the switching perid. D = tn. ', (2.) The vltage and current wavefrms described belw are shwn in Figure 2-2 and Figure 2-3 tgether with the time spans D', and (- D) ',, which indicate the different mdes. The current thrugh the primary switch is indicated by s,prim and the ne thrugh the secndary switch (D) r dide by s, sec' During Mde, the vltage ver the magnetizing inductr L m will be equal t the surce vltage Us' The resulting current as a functin f time (2.6) can be derived frm the inductr relatinship (2.4). m(o) is the current level at the start f the switching perid. The maximum current reached at the end fmde (t =D'"m~) will be equal t (2.8). During Mde, the dide will cnduct and the vltage ver L m will be (2.3), as the vltage n the secndary side is transfrmed with a factr n t the primary side. This results in a decrease fthe current accrding t (2.5) and (2.7). Bth (2.6) and (2.7) are referred t the primary side. The energy stred during Mde shuld be equal t the energy released during Mde. This results in equatin (2.9). By slving this equatin, a relatinship can be fund between U ' Us' D and n (2.0) [9J. 8

15 Mdel Mde (2.2) (2.3) U =L dim s m dt (2.4) di -nu =L ---!!!- m dt (2.5) (2.6). ( ) - _ nu m t - m,m"" L t m (2.7) m,rrmx = m (0)+ ~s DT, m (2,8) nu (-D)T, _ UsDT, L m L m (2.9) U _ Us D (l-d)n (2.0) -<' 20 ' -. m ~ di _ s,pnm U l:;!!:la~~ 'l~,---'---.om!l-.-~t"" _-_~-..;.-----s~,..;;.se;",;c-_l DT (-D)T ~.. S s... _,_",... _... D -- -".,.. -- ) ~ max > - 0 5,- U s -n*u 0 ~ Time (Jls) ~ Figure 2-2: Current and vltage wavefrms fan ideal islated fly-back cnverter 9

16 > 5-0 =» <C " ca 7.8 J;! 4.6 DT s (--D)T' s Time (ls) Time (ls) Figure 2-3: Output wavefrms f an ideal islated fly-back cnverter 2.2 Cmpnent calculatin Several bundary cnditins are needed befre the cmpnents f the cnverter can be calculated. These bundary cnditins include values fr U ' Us, D and 0, The circuit cmpnents can be calculated frm these parameters The transfrmer rati n With a given Us> U and D the transfrmer rati n can be determined frm equatin (2.0): (2. l) The magnetising inductr L m n steady-state peratin the charge transprted thrugh the dide during Mde shuld be equal t the charge used by the lad during a full switching perid l/ ad ', =Tn ( - D)',. (2.2) 0

17 Accrding t the transfrmer relatinship in =i 2 N 2 the average dide current will be n times the average magnetising current. Therefre, the required average magnetising current T", will be [2] - im= ( ) f'd' n l-d (2.3) t is shwn in equatin (2.6) that the slpe f the magnetising current is determined by the size f the magnetising inductance. Therefre, the ripple current 2!i m shuld be chsen s that it satisfies the specific design. The maximum currents will then be f m.max =T", +!im, (2.4) fd,max =n fm,max n rder t accmplish the 2!i m ripple, (2.4) can be used t determine the size f the magnetising inductance [2] (2.5) The utput capacitr C The utput capacitr size is dependent n the allwed vltage ripple!u. This vltage ripple is dependent n the lad current f'd and the duty-cycle D: i = C!V ~!U =!t f'd = DT/'d!t 0 C C' (2.6) Leakage inductance The cupling between the primary and secndary cil f the fly-back transfrmer may nt be ideal, resulting in a leakage flux which des nt cntribute t the energy transfer frm the primary t the secndary side. t des hwever stre a small amunt f energy. As is shwn in Figure 2-4, the energy stred in this leakage inductance L, des nt have a path t grund r the surce when the switch is turned ff. The result is that the parasitic capacitance f the switch is charged rapidly t a high vltage, causing the switch t breakdwn.

18 Ll Nl N2 Us C ~ lad D Figure 2-4: Fly-back cnverter with leakage inductance Snubber T prevent this breakdwn, a cnductin path shuld be available at tum-ff. A methd t prvide this path is t include a snubber in the circuit. The energy in the leakage inductance can be redirected by r dissipated in this snubber. This allws the excess energy t be dissipated at a predefined area in the cnverter. The snubber used in this cnverter des nt have t be regenerative, as it is nt a gal t reach a very high efficiency. Therefre a passive dissipative snubber is chsen. There are basically three types f passive snubbers, shwn in Figure 2-5: the rate f rise cntrl, the vltage clamp and the damping snubbers [22]. The damping snubber, Figure 2-5a, will put a cnstant drain n the system with every vltage change and therefre hinder circuit peratin and efficiency. The vltage clamp snubber, Figure 2-5b, will limit the vltage spike caused by the release f the energy stred in the leakage inductance. During the release f this energy, the vltage will be effectively clamped t the cnstant vltage ver the snubber capacitr. Finally, the rate-f-rise cntrl snubber, Figure 2-5c, can be used t bth clamp the drain vltage t a specific value and limit the rate at which this vltage rises [22]. a b c Figure 2-5: a) damping snubber b) vltage clamp snubber c) rate f rise cntrl snubber 2

19 n first instance, the vltage clamp tplgy was chsen. The disadvantage f the vltage clamp snubber is that scillatins will ccur when the vltage drps belw the clamp vltage. These scillatins are unwanted, as these might cause EM prblems. The snubber vltage and the vltage ver the primary switch, as indicated in Figure 2-6, are shwn in Figure 2-7. The first graph shws the vltages f a vltage clamp snubber with a lng RC-time and the secnd ne with a shrt RC-time. A lng RC-time will allw fr large scillatins. A shrt RC-time can be chsen t damp ut the scillatins, but will als cause a cntinuus capacitive lad fthe circuit nde at the drain f the MOSFET. This can be seen in the first graph where the switch vltage and the snubber vltage are the same during the time the switch is ff. L.6434 C 0 + Figure 2-6: slated fly-back cnverter with a vltage clamp snubber. 3

20 60-- Shrt RC-time -V; sw ---V snub f 60 :> -; 40 C).B 20 > Lng RC-time ~ ~ : Time (ls) Figure 2-7: Typical wavefrms when a vltage clamp snubber is used in a fly-back cnverter withut damping. The snubber has anther functin, which is preventing parasitic switching r increased switching times f the MOSFET. The rapid vltage rise at the drain might have a large enugh dv / dt which causes a rise f the gate vltage via the parasitic gate-drain capacitance C gd Therefre, the rate at which the drain vltage rises, shuld be limited by the snubber. snubber capacitr size will thus be a cmprmise between a high acceptable clamp vltage and a slw rate f rise t ensure prper turn-ff f the MOSFET. The rate f rise cntrl snubber will therefre be used in the circuit, as is shwn in Figure 2-8. The Ll N N2 Us + C.J, lad Csnub 8 D Figure 2-8: Fly-back cnverter with rate f rise cntrl snubber and parasitic elements 4

21 With this cnfiguratin, the vltage spike ver the MOSFET will becme a relatively lw frequency scillatin. The snubber resistr R snub shuld be chsen such that this scillatin is damped quickly, preventing high frequency scillatins, such that high frequency nise is eliminated. Furthermre, the RC-time shuld be shrt enugh t ensure the ability t use small duty-cycles. Shrt RC-times and scillatin reductin cincide with relatively high clamp vltages and quick vltage rises and thus the pssibility f a unfavurable MOSFET tum-ff. Lnger RC-times and gd scillatin reductin cincide with slw rate f rise and a lwer clamp vltage and thus an increase f the minimum duty cycle. An example f the effects f a rate frise snubber n the drain vltage f the primary switch is shwn in Figure Bi-directinal pwer transfer As is discussed in the intrductin, pulsing the lad f the fuel cell culd be beneficial t the perfrmance f the fuel cell. t is als indicated that this des nt wrk with the catalyst available in the PEM under test. t is, hwever, nt knwn if this technique will be beneficial in cmbinatin with the electrlysing technique. Therefre the ptin f feeding back energy t the input f the cnverter is implemented. This is dne by replacing the secndary dide by a secnd MOSFET. The MOSFET in this set-up has a parasitic capacitance similar t the ne at the primary side. This will intrduce an scillatin between the leakage inductance divided by N 2 and the parasitic capacitance as sn as the cmbinatin f the MOSFET and its anti-parallel dide stps cnducting. A secnd snubber is added t damp this scillatin. The resulting cnverter tplgy is shwn in Figure > ;- 40 C) ns ~ 20 ~. _V sw ~ ~.. ---V \ \ \ \.,..0' ~... snub,.. ---' \, Time (Jl5) Figure 2-9: Typical wavefrms when a rate-f-rise-cntrl snubber is used with a fly-back cnverter. 5

22 Ll Nl N2 Us + Dsnub2 C -J, lad Csnub Cpl S2 J Rsnub02 Cp2 Figure 2-0: Fly-back cnverter with a dual (active) switch tplgy 2.3 Design cnsideratins fr the transfrmer n the previus sectin, equatins have been derived fr the winding rati and the magnetisatin inductance f the transfrmer. Hwever, a transfrmer exists nly f tw cils and a cre. Additinally, an air gap in the cre is needed t stre the energy being transferred by the cnverter during each switching perid. The physical aspects f the cre put a limit n the maximum allwable flux density B max. Furthermre, the size f the air gap determines the maximum amunt fenergy that can be stred in the transfrmer. By cmbining these limitatins with the desired L m and n a result fr the number fwindings is fund. The required L m, n and maximum current max thrugh the transfrmer, are determined t cmply with the desired functinality fthe cnverter. Afterwards an apprpriate cre is selected. n this case, hwever, an available bbbin and cre are selected and the design is adjusted accrdingly. The equatins needed fr dimensining the transfrmer are derived in Appendix A. Given the design parameters frm Table, the numbers f turns needed in cmbinatin with the RM-2 cre are: N :=: imaxlm = 9.3l 2.44 O--{j =5.06:=:5 l.max B A ' max e (2.7) (2.8) 6

23 2.4 Cnverter cmpnent design and specificatin Befre the cmpnents fthe cnverter can be dimensined, values fr the input (Us Yin) and utput ( U ) vltages, duty cycle (D), switching frequency (is) and utput current ( ) 0 shuld be chsen. Hwever, a prblem arises with the specificatin f the utput parameters. N literature indicates the vltage-current characteristics f a PEM fuel cell used as an electrlyser, nr is it knwn hw much pwer is needed. A vltage-current characteristic f a PEM electrlyser has been fund [23], which indicates a dide like curve. Furthermre, the cnverter shuld be dimensined t frce vltage nt the fuel cell. This has led t the utput specificatin f =8A and V ut =5V. The chsen switching frequency is 00 khz. This allws fr the use f a small transfrmer cre and bbbin, fr which the RM-2 has been chsen. The input vltage and current shuld be within reach f the available equipment. A maximum input current f loa has been specified accrding t the available vltage surces. t has been nted earlier in this reprt that the cnverter shuld als have the pssibility t feed back energy int the surce, in rder t fulfil the system requirement, D shuld nt exceed 0.5. This will prvide a better symmetry in bth mdes f peratin. Bth the input vltage and duty cycle can be used t keep the input current within the specified limit. t is nw pssible t design the cnverter. First, the transfrmer rati is calculated with (2.). Secnd, the different aspects f the magnetising current can be calculated, based n desired values. t is imprtant nt t make the maximum and average currents during the n-time t large. The reasn fr this is shwn in (2.4), (2.5). The maximum current thrugh the secndary dide r switch will be n times the maximum magnetising current. Furthermre, the ripple n the magnetising current is inversely related t the magnetising inductance L m, which is directly related t the number fwindings placed n the bbbin. A cpper wire with a diameter f.6 mm is used fr the cils. A wire this thick ensures lw (DC) lsses. t des, hwever, limit the number f windings t apprximately in ne layer. Last, (A. 4) shws that the number f windings allwed t make a certain magnetising inductance is directly related t the maximum current. All the cnsideratins abve have led t the specificatins and cmpnent values shwn in Table. The calculated values are the result f the design parameters. The number f windings has been runded twards the nearest integer, as it is nt practical t make a transfrmer with 5.06 and 2.63 windings. The cnsequence is a slight difference in the values f L m and n. This can be crrected by chsing a lwer duty cycle, thus keeping the currents the same. t als becmes clear frm Table that the realised magnetisatin inductance is abut twice as lw. During the calculatins regarding the air gap nly the middle leg has been taken int accunt, 7

24 whereas all three legs shuld have been taken int accunt, hence the factr tw difference. This can again be cmpensated with an altered duty cycle. The result can be saturatin f the transfrmer cre. The pint fsaturatin can be checked by rewriting (A.3) t (2.9): B 2 A gap max e =4.6A. lll m (2.9) t is bvius that with the realised L m and crrected A e saturatin will nt ccur till a value beynd the riginal design parameters. 8

25 Table : Fly-back cnverter parameters Electrical design parameters Transfrmer design parameters Us 0 V B rrmx 270" mt U 5 V A.# e 46'" mm 2 D 0.49 MLT 6(" mm fld 8 A K u 0.9'" dim 0.4.i,;, A 9Op 0.27 mm s 00 khz du 0.08 U V Usw,max 55' V fsw,max 56' A Calculated values Realisable values Realised values C 98 J.LF J.LF L m 2.44 J.LH J.LH L l.29 J.LH nh n N] N z f m.max 9.3 A i m 8.6 A Us",) V R snub... 5 n 2//2 n Csnub nf nf D * ** *** These include a safety margin t prevent failure f the switch due t t high vltages and currents. B max is given as 300mT; the lwer value specified in the table is a safety margin. Parameters specified in data bk [24]. **** Derived frm simulatins. *# This is the crss sectin f the middle leg f the cre. 9

26 2.5 Cnverter simulatin Pwersim's PSM is used t simulate the cnverter. The vltage and current wavefrms f the ideal islated fly-back cnverter shwn in Figure 2-, can be seen in Figure 2-2 and Figure 2-3. A mre realistic simulatin f the built cnverter can be btained by including the snubbers, additinal parasitical resistrs and the gate driver. These resistrs are estimated based n the resistivity f cpper and the dimensins f the print tracks. Cntact resistances have been neglected. The resulting PSM schematic is given in Appendix B. A mdel f the gate-driver f the primary MOSFET is presented in Figure 2-. The nly MOSFET parameter that can be specified in PSM is its n-resistance Rds,n' Furthermre, it can nly be cntrlled by an "n" r "ff' cntrl signal. The cmparatr (Campi) cmpares the "gate vltage" with a reference f4v (the threshld vltage), The switch is turned n when the gate vltage passes the 4V. The results are shwn in Figure 2-2. The disadvantage fthis tplgy is an instant tum-n f the MOSFET when the gate vltage reaches the 4V. The effects n the switching behaviur are shwn in Figure 2-3. A spike in the current thrugh the switch is visible at tum-n. This is caused by the almst instantaneus discharge f the charge stred in the parasitic capacitance Cds' n practise, the switch will nt tum n s abruptly and the parasitic capacitance will be discharged mre smthly ver the n-resistance Rds,n' This tplgy des shw the effect f the fast rising drain vltage fthe MOSFET, as is indicated in Figure 2-2. Finally, the duty cycle needs t be increased t abut 0.49 t cmpensate fr the lsses in the circuit. Cgd ~470P 2 + GND Figure 2-: Gate driver simulatin in PSM 20

27 - > - en 5 >0' 0 ~, dv/dt disturbance f \ t--- ' \ " -, Time (ls) Figure 2-2: Gate driver utput vltage as simulated in PSM \ -« ~ n -_._ , prin - n ,5ec - _.-._.- _._.-, , -_ ~ ~ f, --', -~, > E 0..J > J Time (ls) Figure 2-3: Realistic wavefrms f an islated fly-back cnverter with tw snubbers Cntrl The methd chsen fr the cntrl f the cnverter is a P cntrller implemented in Matlab Simulink. The cntrller utput is limited by a saturatin blck t ensure the switches will always be switched. This prevents damage t the cnverter. An anti-windup circuit is implemented t prevent the integratr utput grwing t large values as sn as the saturatin limit is reached. A ttal f three cntrllers are implemented. One fr the utput vltage, ne fr the utput current and ne fr the cell vltage, which is shwn in Figure 2-5. t is pssible t switch between these cntrllers at any time. The dwnside fusing multiple cntrllers is that the integratrs might be at relatively high values at the time when cntrllers are switched. This is caused by a changing input and a cnstant reference while a cntrller is nt in use. This is slved by making the inputs 2

28 f unused cntrllers zer and resetting their integratrs. A flw diagram f the cntrller is shwn in Figure 2-4. Selected Cntrller N? Cell Cntrller? Yes? Cntrller reference = Cnstant reference Errr =0 ntegratr reset = Cnverter plarity negative? N? Yes? Errr =Vcell - reference Errr = reference - Vcell aturatin input == saturatin utput. AND cnverteun < limit N? ntegratr input =K* K *Errr ntegratr input = 0 Figure 2-4: Flw diagram f the cell-vltage cntrller 22

29 ... ""'l., JQ : Vcell tll N ~ \ c::j~-.~ (]Cntrl!p type& selectr Relatinal a Cntrller Operatr f reference 4 JQ value tll n =::j ~>-r j0.008-=:> ( K_' _ _-----+_ Jl.!sJ ;- Current limitatin ntegratr h-l~f-,' r---'c::j Duty Cycle Value Out 5!. ;-., Cell Vltage reference ---- Cell Vltage Reference cnstant ~ nput current prtectin

30 2.7 Additinal circuitry The (islated) fly-back tplgy has been chsen fr its simplicity and the lw number f (pwer) cmpnents. Hwever, in rder t make the circuit wrk, additinal circuitry is necessary. This includes the already mentined gate drivers fr the MOSFET's, vltage and current measurement sensrs and the driving circuits fr the relays, which allw the cnverter t switch between individual cells f the cnverter. These circuits are shwn and described in Appendix C Relay switching set-up As the Hz/CO fuel stream enters the fuel cell stack, it will pass thrugh sme cells in the stack sner than thers. Sme evidence has been fund that the CO cncentrates itselfprimarily in ne f the cells. This des nt mean that the ther cells will stay clean ver time. t is necessary t have the ability t switch the cnverter dynamically in parallel with different cells. A relay circuit, Figure 2-6, has been designed t prvide this ability. Tw relays, indicated by the fur sets, are switched simultaneusly t switch the cnverter as a whle in parallel with a certain cell r t discnnect it frm the fuel cell stack. This tplgy des nt allw the cnverter t be switched in parallel t multiple cells thugh. f this functinality is needed, a re-wiring between the relay circuit and the fuel cell is needed. The drive circuitry fr the relay cils can be fund in Appendix C. The thick black lines in Figure 2-6 indicate the current paths when the cnverter is regenerating the bttm cell f the fuel cell stack. Nte that the vltage f the cnverter is ppsite t that f the fuel cell. The vltage ver the cell shuld becme equal t the cnverter's utput, minus the vltage drp ver the cnnecting wires. t shuld be nted that this tplgy des nt allw the cnverter t act as a simple lad by having a lwer utput vltage, but equal in sign, than the fuel cell. This design decisin was made t put pririty t the regeneratin by electrlysis. Of curse, if the switching between the individual cells is nt needed, tw sets f relays can be cnnected t the same cell with ppsite plarities. This will prvide bth lad and reversing plarity functinality. 24

31 Ecnverter + Figure 2-6: Switching scheme fthe relay circuit. 2.8 Cnverter implementatin, tests and results The cnverter and its additinal circuitry have first been built n test-bards. Based n the experiences with this setup, sme mdificatins have been made, which are incrprated in the design steps discussed earlier. The final cnverter, including the measurement and driver circuitry, is implemented n a custm made PCB. The lw-pass anti-aliasing filters are implemented n a circuit bard and the relay circuitry n a separate PCB. A picture f the final circuit is shwn in Figure 2-7. A loy pwer supply shuld be cnnected t V in (+) - V;n(O), separate l2v pwer supplies shuld be cnnected t the indicated cnnectrs, V ut (+) - V ut (0) indicate the utput terminals f the cnverter and the remaining cnnectins indicate the utputs fthe measurement circuits and the inputs fr the gate drivers. 25

32 Primary side gate signal Secndary side gate signal Vin (0) lin.. =Cj ;<:: '" 'y... = Vut (0) = ~ e~.. Vut (+) lut ecndary side 2V 0 2V '" cs ~ E ;<:: '" -=....~..~.....~..=0 Cj,;,r: Cj cs,q» c:: "0 ~... cs '0.. '" r--,..; N ~..=.! ~ Vin Vut.0 f'l

33 2.8. Cnverter testing methds and results n rder fr the cnverter t wrk, several sub-circuits shuld be tested during assembly. The first circuits t be built and tested are the gate drivers and snubbers. Tests f cnducting paths shuld be enugh fr the latter circuit. The gate-driver has slightly different characteristics, depending n the lad fthe circuit r its pwer state. This can be seen in Figure 2-8. The lwer characteristics are recrded after the circuit was cmpletely built. The black line crrespnds t a lad f50 and the grey line t.70. Furthermre, the LEM sensrs, MOSFETs and the transfrmer shuld be placed n the PCB t ensure cnductive paths n the primary and secndary sides f the cnverter. Testing f the flyback cnverter is dne with resistive lads f 5 nand.7 n. This gives the pssibility t use the standard duty-cycle f 0.5 f the used functin generatr. The cmbinatin f this lad and the duty-cycle will prvide a nt t high pwer utput. The snubber is checked by measuring drain vltages. f any disturbing scillatins still exist, the snubber has t be redesigned. Oscillatins did ccur at first. Damaged snubber capacitrs, prbably due t high currents, were the surce fthese scillatins. This prblem has been slved by placing 22.3nF and 33.3nF capacitrs in parallel. This reduced the vltage versht slightly mre and slved the peak-current prblem as it is nw levelled ver tw capacitrs. Primary gate vltage, unladed > h, 8, 5 ct == > > 8, 5 ct == > \ J \ J Primary gate vltage, with lad h {- - ~ r ,. l ~ Time ( ls) Figure 2-8: Primary V gs with the pwer circuit turned ff(tp) and with tw different kind fads (bttm)

34 Primary side V ds - ~20 Q) C) S 0 > -. _5... y0 '\ "\ J secnda& side Vds 5 0 >-Q) 0 C) ~. S 5 :""'. > -0 \ 0 \ Y -5 Time (JlS) \ A Figure 2-9: Primary and secndary drain-surce vltages f the MOSFETs 5 0 The currents thrugh the secndary snubber capacitr were nt as large as n the primary side, causing n failure. t was hwever replaced by a slightly larger mdel, as it heated up. The resulting drain-surce vltages fr the different lads are shwn in Figure 2-9. Finally, the measurement circuits are cnstructed. These circuits must be able t measure the lw frequency «khz) behaviur f the in- and utputs. The anti-aliasing filters will cut-ff anything abve 5kHz. The linearity check and calibratin graphs fr the vltage measurement circuits can be fund in appendix C.2.2. Besides the difference in vltage levels between the 5Q and.7q lad wavefrms, anther difference is shwn in Figure 2-9, indicated by D. The cnverter is perating in discntinuus mde when it is laded with a 5Q lad. This means that the energy in the transfrmer has been cmpletely transferred t the secndary side during (l-d)t s. As a result, the secndary side snubber capacitr as well as the parasitic Cds f the secndary side MOSFET is charged by the utput capacitr, causing the slpe visible at D Output characteristics fr different lads The utput characteristics f the cnverter are recrded with the scillscpe (See Appendix D). The tp graph f Figure 2-20 shws the signal fr a 5Q lad measured directly at the scpe. The straight dtted line shws the respnse f the measurement circuit. Averaging the first 28

35 wavefrm with Matlab, shws that the measurement circuit represents this average quite well. The difference is prbably due t calibratin errr. The secnd graph shws the utput vltage fr a.70. lad. The effect fthe secndary snubber is much mre visible in this situatin Output pwer characteristic Relatinship (2.0) shws that the utput vltage fthe cnverter is related t the duty-cycle. Therefre measurements have been dne t shw the relatinship f the utput pwer and the duty-cycle (Figure 2-2). These measurements have been dne in tw situatins. First, the primary MOSFET is switched and the secndary MOSFET is turned ff, thus nly using the antiparallel dide and the zener-dide. Secnd, bth MOSFETs are switched inversely. The used tplgy causes a delay between the reactin f the primary and secndary MOSFET gate vltages. This ensures that bth MOSFETs are nt n at the same time when the primary MOSFET is switched ff. Hwever, the delay causes a simultaneus n-state at the mment f primary turn-n, which increases the efficiency at higher duty-cycles, but decreases the efficiency at lwer duty-cycles, as is visible in Figure 2-2. The abslute difference between the tw situatins is nly 0.82W in favur f the situatin with bth MOSFETs switching. The maximum pwer is als dependent n the state f the cnverter: it will be slightly higher at a cld start-up than when the cnverter has been perating fr a while and has warmed up. The surce used fr these measurements can supply a maximum f 50W. The decline in pwer fr D > 0.6 is caused by the current limiter fthe surce. _ 5.8 ~ 5.6 () ~ 5.4 ' > 5-5 Output vltage: R = 5,Q - t-"'l"l"_..-...""""~_""" "rrmi~=~ ~ 3.4 & 3.2 CS :!:: 3 > 2.8 ' '- ---l- --'---.~ Output Vltage: R =.7,Q Spike caused by,.,. secndary snubber - \~, 2.6 ' '- ----L --' ' Time (ls) Figure 2-20: Output vltage characteristics f the CODverter

36 Single MOSFET switching -Output Pwer nput Pwer nverted MOSFET switching ' C) ~ 20 D.. - Output Pwer nput Pwer _ L_.:...:..:=-=:..J.::..:..=..;.,~~-~-~-~-~::r::::===- ----l L~ ~40 r::: C) 0 20 t: w Duty Cycle - Single MOSFET Dual MOSFET O ' L------' '------' '------' 0.5 Figure 2-2: Perfrmance characteristics fthe fly-back cnverter Reversed cnverter peratin n rder t minimise the amunt f current that will be flwing thrugh the relay switches at turn-n and turn-ff, the cnverter utput needs t be shrted. The current drawn at such a mment, will be similar t the pulsing technique described in the intrductin, with the exceptin that the pulse is nt adjustable in amplitude. The energy ging thrugh the cnverter at this time can be transferred t the primary side. The surce at the input must be able t cpe with this fr brief perids f time. The lw efficiency f the cnverter can als be useful n this regard, as sme fthe energy will be dissipated befre the input fthe cnverter is reached. A test was dne t check the ability f the cnverter t feed back energy. This was dne by placing a 0ltrnix vltage surce (EM473, see Appendix D) at the utput fthe cnverter. The

37 result is shwn in Figure The surce at the input must be taking in energy, as the vltage n the input capacitr (V in ) is nt increasing rapidly. These results indicate that feeding back energy is pssible with the current set-up. 0 ll.. 'S Pwer feedback test ::;, 0 > 5~ ~ ~ ::;, 0 0 D c::: > Time (5) Figure 2-22: Pwer feedback test results 3

38 3 Fuel cell regeneratin The cnverter is cnnected in parallel t the fuel cell via a fur sets f tw relays, as is described in sectin and is shwn in Figure 2-6. The cnverter vltage must be made as lw as pssible at the time f switching, t prevent large peak currents during switching. This means the cnverter shuld be cntrlled by the utput vltage cntrller set t OV. After which, the cntrller is switched t the cell vltage cntrller and the desired pulse shape is made. The cntrller is then set back t OV, a switch t the utput vltage cntrller is dne and the relays are switched again. A pssible benefit f this switching technique is the creatin f additinal water in the cell befre and after applying a pulse, reducing the risk f dehydrating the fuel cell with a regeneratin pulse. The gas used fr the CO pisning tests is a simulatin f gassified bimass. Such a gas is ften referred t as synthetic gas r "syn-gas". The syn-gas used fr the CO pisning measurements cntains 30ppm CO and the cmplete specificatin is given in appendix E. A linear lad, cnsisting f five pwer MOSFETs cntrlled by a P cntrller in dspace, is used as the adjustable lad fr the fuel cell cartridge. 3. The measurement setup The measurement setup is shwn in Figure 3-. The cnverter r vltage surce is cnnected via relays t a PCB with sets f MOSFETs placed in parallel t each cells. These MOSFETs are used t shrt circuit the cells every nw and then t humidify the membranes. Tw have been put in series, as the dides wuld therwise shrt circuit the applied vltage surce when vltages lwer than their (anti-parallel dide) threshlds are applied. The cell vltages are als measured n this PCB. A direct cnsequence f this is that the actual cell vltages will be lwer than measured as sn as the vltage surce supplies current. Furthermre, during negative pulses, the sum f the cell vltages is nt t get belw the threshld vltage f the dides in the linear lad, as this will reverse the plarity f the current. Finally, the linear lad is used t draw a current frm the cartridge as desired. Figure 3- nly shws the MOSFETs, relays and measurement pints fr cell 3. The actual setup has the shwn circuit fr each cell. There is nly ne surce\cnverter and nly ne linear lad. 33

39 Relay Cell 4, - '-- Cnverter r Xantrex surce + Relay +-_..._ f,~/...lr:_:_::_:_:_:_:_:_:_:~) / V cell - '--- Cnnectin wire : Cell 3 Cell 2... l Cell _

40 3.2 Reversing plarity First, a test was dne t check whether the cnverter culd reverse the plarity f the utput f a vltage surce. The test was successful. Secnd, shrt circuit tests were perfnned with the fuel cell. One test used equal plarities f the fuel cell and the utput f the cnverter. Anther test used ppsite plarities; bth tests were successful. After these tests, reversing the plarity f the fuel cell was tested. t was attempted t reach cell vltages lwer than -2V by lwering the cell vltage with a negative slpe. Hwever, the current prtectin interrupted the measurements, in rder t prevent damage t the cnverter, befre the end f the slpe. Different slpes and pulses have been tested t see whether r nt the shape f the slpe had an effect n the necessary current drawn. The cnclusin was that the used cmbinatin f input vltage and input current was nt sufficient. Therefre, the input vltage was increased t 5V, this did nt slve the prblem; the current prtectin still interrupted the measurements. Then, the input surce was replaced by a mre pwerful ne (Xantrex surce, Appendix D). The new surce prvided the ability t reach a lwer cell vltage. Hwever, it was still nt pssible t reach values lwer than -2V Vlt. Secnd mismatch The nly remaining explanatin fr the necessary high input currents is a saturatin fthe flyback transfnner. t was fund that the vlt secnd prducts f the primary and secndary cnverter sides did nt match. The basic principle f the fly-back cnverter is that the energy stred in the transfnner during ne switching cycle is als released during the same cycle. With a fixed magnetising inductance, the prduct f the vltage ver this inductance and the time this vltage is applied is a measure f the energy put int the transfnner. n this case, the prduct f the primary side is larger than that fthe secndary side. As a result, energy will be built up in the transfnner, causing saturatin. Figure 3-2 shws the saturatin f the transfnner. n rder t clearly identify saturatin, the switching frequency was lwered t 0kHz. The current at which the effect f the saturatin becmes apparent, is fund t be 24A. This is higher than the value calculated in (2.9). This is mst likely due t the cnservative value fr B max chsen during the design fthe cnverter. The slutin t this prblem is using a variable frequency cntrl f the cnverter and if necessary, a redesign f the transfnner. As an alternative, the Xantrex surce is used fr the desired measurements. T acquire the prper bundary cnditins fr a new cnverter, the vltage-current characteristic is measured. 35

41 3.2.2 Vltage current characteristics The vltage-current characteristic is made by switching the relays and immediately afterwards applying a negative slpe t the cell vltage cntrller, as is shwn in Figure 3-3. This figure shws a cnnectin between cell vltages. Every time a negative vltage is applied t cell, cell 2 will respnd linearly and the same applies fr cell 3 and 4. The linear relatinship that describes this effect is fund t be v = surce + O.85V. cupled 42 (3. ) 200 -«00 --l: Q) l-.. 0 :: 0 Nrmal fly-back Current thrugh primary switch J peratin '" ~sfrner..., saturatin... "....,. T v Time (J.ls) 2 Figure 3-2: Transfrmer saturatin due t a vlt*secnd mismatch # -,'" ~" ~'......, >, \ '. - 0 \ Qi \ () \ \ \ > -2 ~60 -s::::: Q) l :: 0 Q) :: 0 en 0, \ Time (s) Figure 3-3: Data frm which Figure 3-4 has been derived, (-) = Vcelll, (-) = V cell2, (-) = V cell3 and (--) = V cell 4 '. " 36

42 > "iii - >u \ " -V cell - - -V 2 cell -V cell V 4 cell surce (A) Figure 3-4: Vltage Current characteristic fr a fuel cell in reversed plarity A psitive increase f a cell vltage can have several knwn causes. First, a reductin f the lad current brings the cell in a different state and increases the cell vltage. As the (stack) lad current is zer during these measurements and the raised vltage is well abve the theretical limit f a fuel cell, it is unlikely this is the case. Secnd, the vltage can increase if a negative current is frced thrugh the cell. The vltage surce is placed in parallel with just ne cell at the time and with a reversed plarity. Therefre, n cnductive path exists t cause a negative current thrugh the adjacent cell. The remaining pssibilities are an errr in the set up, the influence f the cartridge r an electrchemical effect. Figure 3-4 shws the current vltage characteristics gathered frm the data f Figure 3-3. The nn-linear behaviur belw 7.5A is due t the behaviur f the surce and the fuel cell right after the start f the measurements. Neglecting the lw-current behaviur, the slpe starts ff linearly. The hrizntal line indicates -.23V, the vltage required fr the reversed fuel cell reactin (.6). Nn-linearities are clearly visible arund this vltage, indicated by the circle. A steeper slpe with respect t the current fllws after these nn-linearities. This culd indicate the reverse fuel cell reactin is taking place, which is as expected. 37

43 3.3 CO pisning t is assumed that the individual cells in a fuel cell cartridge are placed in series with respect t the fuel flw. As a cnsequence, the hydrgen cntent f the fuel gas will decrease as the gas ges thrugh mre cells. This means that the last cells will have t deal with a relatively higher amunt f CO and will be pisned first and faster. Furthermre, by increasing the lad f the cartridge, the hydrgen will be cnsumed mre quickly, causing a mre rapid pisning, assuming that the flw remains cnstant. Finally, the rate f pisning can be increased by reducing the flw fthe gas. This will prlng the time each gas mlecule stays in the cartridge. A measurement is dne t determine the effect fco n the cell vltage. t is repeated seven times t determine whether the pisning fcells is a predictable prcess. The results are shwn in Figure 3-5. Measurements 3 t 6 shw very similar results as measurement 2 and are shwn in Appendix F, Figure 6-. During these measurements, the fuel cell is fed with syn-gas and the linear lad is turned ff at first (t < 20s). After that, the lad is set t 6A, until the vltage ver cell 2 has reached a value f -.05V. This cell is chsen because test measurements have shwn that this cell is affected mst by CO, resulting in a negative vltage. The vltage f -.05V is chsen because the vltage did nt change during test measurements. Figure 3-5 shws the degradatin fthe cell vltages fcells and 2. As cell 2 gets pisned, the hydrgen reactin decreases. This translates int a lwer cell-vltage, till n vltage is generated at all. At the same time, the ther cells still prvide enugh vltage t deliver pwer t the lad. The current frced thrugh the pisned cell causes a vltage drp ver its internal impedance, causing a negative vltage t appear ver the cell. Each subsequent subfigure f Figure 3-5 and Figure 6- are chrnlgical. The data f measurement shws a lnger delay befre pisning ccurs cmpared t the ther measurements. This can be explained by a lnger rest time (n lad) befre the first measurement starts. Each subsequent measurement has an equal rest time between lad changes. The rise fthe cell vltages fcell 3 and 4 are caused by a decrease fthe lad current, as can be seen in Figure 3-5, due t a lwer stack utput vltage. The last measurement did nt g as expected, as the cell vltage did nt reach the switch-ff criteria. An interesting effect can be seen in the data fmeasurement 7; scillatins start t ccur abut 90 secnds after the lad change. The data presented in Figure 3-5 indicates that perids f n current thrugh the cells have a regenerative effect as the measurements were dne sequentially. Only 20 secnds f data is absent between tw sequential measurements (which equals a time f 70 secnds between lad changes.). Preliminary tests have shwn that pulsing the lad can be beneficial. 38