Design of a high voltage input - output ratio dc-dc converter dedicated to small power fuel cell systems

Transcription

1 Design of a high voltage input - output ratio c-c converter eicate to sall power fuel cell systes O. Béthoux, J. athelin To cite this version: O. Béthoux, J. athelin. Design of a high voltage input - output ratio c-c converter eicate to sall power fuel cell systes. European Physical Journal: Applie Physics, EDP ciences,, (), <./epjap/48>. <hal-64646> HAL I: hal ubitte on Nov HAL is a ulti-isciplinary open access archive for the eposit an isseination of scientific research ocuents, whether they are publishe or not. The ocuents ay coe fro teaching an research institutions in France or abroa, or fro public or private research centers. L archive ouverte pluriisciplinaire HAL, est estinée au épôt et à la iffusion e ocuents scientifiques e niveau recherche, publiés ou non, éanant es établisseents enseigneent et e recherche français ou étrangers, es laboratoires publics ou privés.

2 DEIGN OF A HIGH OLTAGE INPUT OUTPUT RATIO D-D ONERTER DEDIATED TO MALL POWER FUEL ELL YTEM O. Béthoux, J. athelin Laboratoire e Génie Electrique e Paris (LGEP) / PEE-Labs, NR UMR 87; UPELE; Université Pierre et Marie urie P6; Université Paris-u ; rue Joliot urie, Plateau e Moulon F99 Gif sur Yvette EDEX Keywors: fuel cell, power electronics, D-D power conversion,, efficiency, seiconuctor stress, soft switching, experiental valiation. Abstract onsuing cheical energy, fuel cells prouce siultaneously heat, water an useful electrical power [] []. As a atter of fact, the voltage generate by a fuel cell strongly epens on both the loa power ean an the operating conitions. Besies, as a result of any esign aspects, fuel cells are low voltage an high current electric generators. On the contrary, electric loas are coonly esigne for sall voltage swing an a high /I ratio in orer to iniize Joule losses. Therefore, electric loas supplie by fuel cells are typically fe by eans of an intereiate power voltage regulator. The specifications of such a power converter are to be able to step up the input voltage with a high ratio (a ratio of is a classic situation) an also to work with an excellent efficiency (in orer to iniize its size, its weight an its losses) []. This paper eals with the esign of this essential ancillary evice. It intens to bring out the best structure for fulfilling this function. everal D-D converters with large voltage step-up ratios are introuce. A topology base on a coupleinuctor or tappe-inuctor is closely stuie. A etaile oelling is perfore with the purpose of proviing esigning rules. This oel is valiate with both siulation an ipleentation. The experiental prototype is base on the following specifications: the fuel cell output voltage ranges fro a open-voltage to a rate voltage while the loa requires a constant voltage. The stuie couple-inuctor converter is copare with a classic boost converter coonly use in this voltage elevating application. Even though the voltage regulator faces severe specifications, the easure efficiency reaches 96 % at the rate power whereas conventional boost efficiency barely achieves 9. % in the sae operating conitions. Introuction ince the 9 th, inustrial expansion has always leane on intensive use of fossil resources: coal, oil an natural gas [4]. At the beginning of the st century, governents graually becoe aware of the liit of the hyrocarbon reserves []. On top of a secure an sustainable long-ter energy supply, the huge use of hyrocarbon resources alreay prove to have negative ipact on the earth environental equilibriu. As a atter of fact, greenhouse gases such as carbon ioxie an ethane have eeply an suenly increase leaing to a previous unknown situation [6]. For these two ain reasons, fining out new energy alternatives has becoe a key issue of this century. In this context, hyrogen has attracte great attention in the recent years as a new an clean energy carrier. Hyrogen avois the epenency an epletion of fossil fuels because it can be prouce by ifferent eans an especially thanks to renewable resources (win, sun ) [7]. Furtherore, hyrogen has a goo energy specific gravity ( MJ/kg); no toxic gas is generate uring its cobustion, particularly in the case of the electrocheical oxiation an reuction reactions taking place in fuel cells; resiue is only water an heat [8]. Therefore, fuel cells (s) are the ost iportant fiel of the "hyrogen-energy" use. Theirs areas of application are various. It provies electric power to portable electronic evices, counication equipents, spacecraft power systes (electricity an water cogeneration), transport systes (cars, boats, an planes), an also builings (electricity an warth cogeneration), in a range of power fro a few watts to hunre kilowatts. However, is not an ieal voltage generator an the voltage becoes saller as far as the loa power ean increases. Inee, the voltage-current relationship is inuce by the three ain irreversible losses that occur in the processes. Activation polarization ( act ) is the irreversible voltage loss associate with overcoing the energy barrier to the electroe reaction an cathoe (air) kinetic liitations oinate in this act ter. Ohic losses are basically ue to ionic current in the electrolyte whereas collector resistance an contact resistance (electronic current) are negligible. It inuces an ohic polarization ( oh ). Transport echaniss within the gas iffusion layer an electroe structure cause a reactant concentration ecrease at

3 the electroe surface an prouce an irreversible voltage loss nae concentration polarization ( con ) [9]. As a conclusion, the actual fuel cell voltage ( ) at any given current (I ) can be represente as the reversible voltage (roughly the open-circuit voltage) inus the activation, ohic an concentration voltage losses. E () act Fro a practical point of view, the electric loa is strongly affecte by an iportant voltage swing. In the worst case (age an /or ba operating conitions), the fuel cell voltage can be ivie by a factor of when the require power rise fro zero to a critical level. To face this huge voltage constraint, a power converter usually regulates the voltage elivere by the to the electric loa (figure ) [] []. H Air Fuel ell control I I oh con oltage Regulator H Heat ontrol value ontrol - oltage regulation - urrent protection voltage setpoint Figure. Block iagra of a electric generator In aition, the voltage regulator specifications are quite special. As a atter of fact, the change in Gibbs free energy associate to the reuction/oxiation reactions lea to a.8 stanar theoretical cell potential, assuing a vapour water prouct. Even when no current is rawn fro a fuel cell, there is irreversible voltage loss ue to parasitic reactions which eans that the practical open circuit voltage never excees. Now, in orer to obtain hoogeneous cells conitions (teperature, partial pressure, water raining, seal pressure ), esigner rarely stacks ore than a thousan cells together. This practical choice inuces that the is a low voltage an high current electric generator. As loas coonly behave the opposite way, the voltage regulator connecting an loa has to raise the voltage value with a high ratio an to withstan high input voltage swing. The step-up voltage conversion ratio is usually about, an the voltage swing is typically about / with respects to open circuit voltage []. This paper focuses on the specific voltage regulator integrate in a syste. The current investigations are liite to sall power systes (fro a few watts to kw). This article is ivie into four parts. First, it analyses various topologies base on a unique controlle power switch in orer to bring out the ore appropriate structure (section ). Then, a etaile survey of this latter is carrie out; taking coponents real characteristics into account (section ). Thir, an experiental stuy is evelope to valiate the effectiveness of this analysis. Last, coparative experients allow confiring an quantifying the iportance of the propose structure as far as classic boost converter is concerne. Finally, the conclusion rearks en the paper in section 6. Power converters topology. General purpose In orer to obtain both a goo efficiency an a safe operation, current ensity is liite. Hence, their electroes areas efine their rate current. With regar to their rate voltage, the voltage is liite because of the restricte nuber of eleentary cells that can be connecte in series. This optial nuber epens on the application case but oes not excee uch ore than one hunre. On the contrary, the loa is usually esigne to iniize the Joule losses while taking voltage isolation constraints into account an thus it works with low currents. For sall power, D voltage bus coonly ranges fro to 6. Our experiental setup is base on a Nexa oule esigne by Ballar. It is ae of 46 cells. Therefore, the propose requireent is base on a rate voltage an on a loa voltage reference. In aition, the voltage is assue to fluctuate between an. Base on these specifications, a specific D power converter, aapte to sall power requireents, is esigne. As far as this scope of appliance is concerne, the key issues are cost an easy ipleentation. That is the reason why the power structure of this voltage regulator is restricte to single controlle switch topologies. The topology selection criteria are base on two key points. The first one relies on the switch voltage an current constraints. In orer to copare each topology, a switch coefficient (F ) is efine: I F () P Where is the axiu switch voltage; I is the axiu switch current; P is the power supplie by the D converter. As conuction an switching losses have to be iniize, the lower the switch coefficient F is, the better the topology is. onsequently, in any converter, current ripple has to reain sall copare to axiu switch current; the current ripple will hence be neglecte in the following survey. On the other han, soe topologies are base on transforer or couple inuctors. In this case, it is also iportant to keep the transforer ratio close to one. As a atter of fact, greater or saller voltage conversion ratio inuces larger leakage inuctance an also larger parasitic capacitance. Both eleents significantly egrae the syste perforance [].. D-D converter without transforer In this application, galvanic isolation is not require. o, the conventional boost converter is a very straightforwar way to

4 ipleent the voltage regulator function (figure ). For this topology, the switch coefficient (F ) boost is: ( ) I F () boost I F + (6) ( ) quara As a result, (F ) quara. is even higher than (F ) boost which akes this solution out of purpose. I L I i v D u (t) or O Figure. onventional boost D-D converter As this application requires extree step-up voltage conversion ratios, (F ) boost is high ((F ) boost ) which eans a poor utilization of power coponents ( an D) an a egraation of the regulator efficiency. Furtherore the boost converter operates with the following uty-cycle () boost : ( ) boost (4) () boost.9 is very close fro its upper liit which iplies that it ipairs transient response. onsiering this issue, two classic boost converters in cascae is no way to cope with the specifications since this solution nees two power switches which, on top of that, are ifficult to rive properly because of a high orer uner ape resonant circuits [4] []. Alternatively, this iea was taken up in esigning quaratic converters ae of a single power switch an three ioes with D I synchronize to switches (figure ) [6] [7] [8]. I L I D I D I I L O i v D O O u (t) or Figure. quaratic boost D-D converter onsequently, its uty-cycle () quara is: ( ) quara () which signifies a oerate uty cycle () boost.684 leaing to a uch better transient tracking behaviour. But, the switch coefficient (F ) quara is in this case:. D-D converter with transforer In orer to solve the proble of D-D converters with extree step-up voltage conversion ratio, this sub-section now consier the avantages an isavantages of using a transforer. With respects to agnetic coupling, two kins of topologies can be taken into consieration [9]. In the forwar topology (figure 4), the transforer allows irect power transfer. The flyback topology uses the transforer as two couple inuctors (figure ). In both case, the transforer purpose is to ecouple electric constraints linke to the source (low voltage / high current) fro those linke to the loa (high voltage / low current). It allows using suitable power switches in priary an seconary sies. I HF Filter D M N i N N v D O u (t) or L O D OD N N O Figure 4. forwar D-D converter o as to avoi agnetic saturation, the forwar structure uty cycle ust not excee. []. Assuing N / N, the uty cycle () for an the switch coefficient (F ) for can be figure out as follows: (7) for ( ). ( ) ( )( I ) F (8) for I The saller the transforer ratio is, the saller the switch coefficient is. As the forwar uty cycle is boune to., the coefficient can not go below /. An so this forwar topology reveals a uch enhance switch coefficient than the conventional boost one : (F ) for > 4. Even so, this topology has a transforer with a very large step-up ratio. It ranges in our specifications case. As a atter of fact, the seconary part of this topology is a buck converter. onsequently, as this latter part behaves as a stepown converter, the transforer nees to be an extree stepup voltage converter. As a result of this high ratio, the transforer ipleente in the forwar topology entails an iportant parasitic capacitance. Utilizing a transforer with such a flaw generates voltage an current spikes an increases

5 raatically power loss an noise. This rawback akes this solution unfeasible. I HF Filter i N v N u (t) or D N N O Figure. Flyback D-D converter The flyback structure features a basic switching cell ( & D) an its uty cycle () fly can range fro to []. Assuing N / N, the uty cycle () fly an the switch coefficient (F ) fly can be expresse as: ( ) fly (9) + ( ) ( + )( I ) F fly ( )( I ) ( )( ) fly () As a result, (F ) fly is iniize for a. uty cycle. In this optial esign case, the switch coefficient (F ) fly gets own to 4 an the transforer ratio is only. Despite that, the flyback transforer requires an air gap in its core aterial causing leakage inuctance between the couple wining. With this ipleentation, the flyback topology switches suffer fro voltage spikes which reuce the structure efficiency. Thus, the flyback topology requires a non-issipative ancillary syste able to recover the energy store in the leakage agnetic fiel. To achieve this function, one takes avantage of the fact no isolation is eane. onsequently, we suggest the new schee epicte in figure 6. It is a variation of the high step-up clap-oe converter propose by []. I HF Filter N i v D u (t) or N D O fly fly O Figure 6. Flyback D-D converter with non-issipative active clap with galvanic isolation A etaile stuy of the selecte D-D converter In the forer section, several topologies have been analyse with respect to the set of requireents of a voltage regulator eicate to a fuel cell electric generator. Aong the D-D converter caniates, one esign has been own-selecte. In this paragraph a oelling of this specific schee is perfore to eterine how well the caniate satisfies requireents. Moreover oelling helps the esigner to perfor an excellent final prototype. In this stuy, we only consier the non ieal agnetic coupling of the transforer wining. Hence, the two-wining transforer equations are oelle with topologically equivalent structure incluing an ieal transforer, a agnetizing inuctance referre to priary (calle L ) an a sall serial leakage inuctance locate at the priary sie (calle L l ) []. Figure 7 illustrates the electrical equivalent circuit oel use in the following converter steay state analysis. In a oelling iteration (for accurate switches analysis), the parasitic capacitance will be taken into account in the next section. I HF Filter u (t) i l T L l L i i i i Dc D i c i D O v O Figure 7. electe D-D converter along a switching perio. teay state successive stages The steay state behaviour analysis shows that the converter features four ifferent topologies along switching perio T. Figure 8 epicts the ain electrical wavefors: seconary voltage v (t), capacitance current i (t) an the priary an seconary currents i (t), i (t).. During the first stage, the power switch current i (t) rises while the seconary ioe current i (t) ecreases. For a tiny leakage inuctance value, the current slope is roughly liite to A/µs rate with respect to MOFET technology. But whatever the transforer we further experient, the rising slope is actually liite by the leakage inuctance. As a consequence, the power evice switches on softly with alost no losses ue to negligible switch voltage. Meanwhile, the seconary voltage v (t) is equal to (v - loa ). The length v 4

6 of stage is linke to the transforer leakage inuctance by the following expression: v (t) T Ll I + v () loa t ubsequently, to experient high switching frequency, L l value has to reain sall in coparison with L value. During the secon stage, the transfers its power to the transforer an the seconary voltage is now equal to (L l << L ). The length of stage is onitore by the close loop eicate to the converter voltage control. o it lasts ( T. ) At the en of the secon stage, the power evice switches off. At the very beginning of the thir stage, its voltage increases proptly as a result of its parasitic capacitance loaing. It brings about the on-state of both ioes D O an D. Hence, the power evice voltage v is clape to v uring the stage. Therefore, the switch-off losses are shortene all the ore since tens to ( + ( loa )/). Next paragraph will give the expression of an next section will show its experiental easureent. The voltage claping also allows choosing a power evice with saller rate voltage. This point is ostly significant for MOFET topology because the conuction resistance R DON ecreases as far as the rate voltage increases. R DON is inee ore or less proportional to / [] [4]. On top of switching losses reuction, the switch benefits fro conuction losses ecrease. During the thir step, leakage inuctance transfers its energy to the clap capacitor. The capacitance value of this latter is high in orer to obtain sall v voltage ripple an hence low switch voltage constraint. That is the reason why the clap phase, which involves a resonant circuit (L l an ), shows a quite linear behaviour. onsequently the length of stage is linke to the transforer leakage inuctance as follows: T L I l + v () During the fourth an last stage, the energy store in the transforer as well as the energy store in the capacitance are restore to the electric loa. The seconary voltage is equal to (v - loa ).. teay state key values The selecte D-D converter has two esigning egrees of freeo: its rate uty cycle an the transforer voltage ratio. In orer to optiize this choice, key values have to be copute. The previous section explaine the converter operation an figure 8 illustrates the significant wavefors in steay state. In this conition, one can state: i (t) i (t) i (t) T T I I T I (-)T Figure 8. Main wavefors of the D-D converter along a switching perio that the transforer average voltage equals zero. Accoring to figure 8, the seconary average voltage is copute as follows: ( ) ( )( v ) v () that the capacitance average current equals zero. This value is calculate using the previous escription: i I I + (4) an that the output power equals the input power. For this purpose, we consier that the input an output filters eliver a constant voltage. Hence, this power balancing epens only on the average values of i (t) an i (t): I + I + () Assuing that the is negligible with respect to the three other stages lengths, one finally gets: t t

7 + (6) (7) + (8) As a result the switch coefficient (F ) can be evaluate as follows: ( ) F (9) ( )( ) Thus the optial uty cycle is ½ which leas to the best possible switch coefficient (F ) 4 associate to the transforer voltage ratio 8 an the theoretical switch axial value I. The above oel was siulate in atlab-siulink environent using its eicate ipoweryste application library. This software is convenient to esign a converter fro the topological point of view to the control esign concern. As a atter of fact, the next step of this present work is to elaborate a siple an efficient control algorith to regulate the output voltage an prevent over-currents. Table lists the ain siulation paraeters. The solver is base on an Euler etho using fixe-step set to ns. Figure 8 epicts ost iportant electric variables: the three currents i l (t), i (t) an i Dc (t) are rawn on the first subplot; the two switch variables i (t) an v (t) are plotte on the secon graph. It can be notice that the above assuption ( << ) is confire. Moreover, the axial switch voltage occurs inee uring the stage an harly reaches 6 which is rather close to the theoretical an very sall copare to the loa voltage. The ifference between both values is ue to the finite clap capacitance value leaing to a I (t) voltage swing. Table : siulation paraeters L 44 µh L l µh P W 8 F khz µf 4 Ipleentation an experiental results of the selecte D-D converter The previous section brought out the theoretical iportance of the selecte structure. It also expose its optiize ain paraeters. The current section presents its ipleentation an associate experiental results. i l (A) ie (A) i (A) i (A) i (A) i (A) i (A) ik (A) v () vk () x x teps (s) x x Tie (µs) teps (s) Tie (µs) x - Figure 9. Main variables of the selecte converter siulation 4. electe converter ipleentation A W / khz prototype was built using coponents with characteristic siilar to table. The transforer was constructe with an ETD9 core ae of 9 aterial. The input, output an clap capacitors ( I,, O ) are ceraic capacitors ( I µf /, 4.7 µf /, O.47 µf / 6 ). The input capacitor is reinforce with an electrolytic capacitor (68 µf / 6 ). The switch is a MOFET transistor IRF46PbF (, 7 A, R DON 4 Ω), the output ioe D O is a DD46 (6, 4 A, schottky technology) an the clap ioe D is a WQFNPbF (, A, schottky technology). In orer to avoi aitional voltage stress to the switches an EMI probles, it is essential to esign a switching cell length as sall as possible (figure ). Actually the switch off overvoltage spike is ue to the prouct of the switch current slope an the inuctance value of the switching cell []. i MOFET Lcell () t MOFET As it is iportant to lower the MOFET voltage constraint, the spike axiu value has to be less than ten percent of its theoretical axiu value, which in our case eans no ore than to 6. With respect to (- A/µs) current MOFET 6

8 slope, this leas to realize a switching cell with a self inuctance lower than: to nh. haracterizing precisely the connections an the P boar is quite coplicate an can be successfully investigate with Partial Eleent Equivalent ircuit Metho [6-7]. However in our case, the switching cell can be approxiate by a square P boar (figure ), with epth b µ sall copare to the other size: the trace D an the trace with w. In that conition, the self inuctance of the loop can be approxiate by the following expression [8]: µ D w L w + ( D w) sinh π w µ ( ) ( ) + + w sinh D w w π () This leas to a square sie saller than D MAX (figure ). I I N D N D O Figure. The switching cell of the selecte D-D converter D D w w Figure. quare coil with rectangular cross section inctance value (nh) quare Loop Inuctance with trace with 4. Prototype tests Figure illustrates prototype behaviour at rate power. Input current I l (t) an switch voltage (t) are plotte on the sae tie graph. In general, these wavefors fit well the siulation ones. In particular, no overvoltage occurs when switches off, confiring that the switching cell is well esigne. However parasitic ringing affects input current I l (t) uring phase. Moreover, when phase starts, the input current value I l (T + ) is saller than expecte. The previous oel cannot anticipate these two ringing phenoena. In an attept to better fit the transforer behaviour, we ust inee take its parasitic capacitor PO into account (figure 4). In figure 4, R PO represents the ringing aping effect ue to value agnetic core loss an wiring conuction loss. This new schee allows preicting the etaile prototype behaviour. Ipeance-eter analysis was conucte with a 49A ( Hz MHz) ipeance analyser HP evice. It confirs a agnetizing inuctance L an leakage inuctance L l close to 44 µh an a µh respectively. It also gives PO about a hunre pf value. It also valiates the ringing frequency of MHz. R P resistance value is set to kω after a fitting process while coparing experiental results an siulation ones. Figure 4 illustrates a rate power siulation with the following paraeters values: PO pf an R PO kω. In brief, phase is ivie in two phases: phase - an phase -. During phase -, output ioe D O oes not yet conuct an for that reason the parasitic capacitor voltage can evolve freely. As L >> L l, one can assue that L behaves as a current source I. iilarly, >> PO iplies can be consiere as a voltage source. onsequently, a resonant circuit takes place with L l in the priary sie an PO in the seconary sie. Phase - ens when ioe D O switches on. At that specific tie, the current rop I l can be copute as: PO I l () Ll This eans that a part of the energy store in the leakage inuctor is not transferre to the non-issipative clap syste but to the parasitic capacitor PO. This energy is lost uring the phase when the sae resonant circuit oscillates uring several perios: energy issipation occurs in winings an core aterial. As a conclusion, these two parasitic ringing phenoena have a etriental effect on the converter efficiency. Accoring to equation (), the saller PO value is, the ore efficient the claping syste is. Hence the transforer esign has to reuce the parasitic capacitance value. That is the reason why the priary an seconary inuctors are wine with a wining spacer (figure 6). trace length () Figure. quare coil inuctance 7

9 W I l P W Phase -- 4 I ik (A) Figure. Experiental wavefors of the prototype I i (A) v () vk () i l T L l L I i i I Dc u (t) D I c PO D O R PO v i O Figure 4. Detaile oel of the selecte converter P 8 W x teps (s) Tie (µs) x - Figure. iulate wavefors of coplete oel Wining spacer oil Forer Figure 6. The transforer wining builing Experiental coparison of the selecte topology an the classic boost converter The analysis of the selecte D converter experiental wavefors contribute to optiize the transforer esign which is one of the key coponents. In this section, the selecte D converter is copare with the conventional boost converter. For that purpose, two W / khz prototypes were built, using M coponents, ceraic capacitors an the sae core aterial ETD9 / 9 for the transforer an inuctor. Table suarizes the ain features of each prototype. Figure 7 shows key wavefors of the. Power losses were analyse in etail. The selecte D converter efficiency is far greater than the classic boost one. (Table ). At rate power, conventional boost has twice ore losses than the selecte converter. As shown on figure 7, the switch blocking voltage of the selecte converter never excees 6 which allows choosing a MOFET with low rate voltage an hence low on-state resistance R DON. On the contrary, boost converter switch suffers fro high voltage ( ) an thus has large conuction loss. On figure 7, i l (t) wavefor also confirs that the selecte structure avois the transistor switch-on losses. The active switch has sall switch-off losses because its voltage is clape to a oerate voltage ( ). Quite the opposite, the transistor of the boost converter faces iportant switch losses because of har switching an large voltage stresses. In aition, a general picture of selecte converter valiates the four epicte phases an corroborates equation (8). This equation proves that phase length is constant an uch saller than switch off-state length (-)T. onsequently, phase 4 always exists an assures the soft switching of phase. oponents atasheets enable to calculate switch losses, an teperature sensors perit to easure the heat sink teperature to valiate these evaluations. As a atter of fact, active switch losses of selecte structure are eight ties saller than boost MOFET losses. It points out that the selecte converter has an efficient topology with respect to switch stresses. This fact is also confire when easuring switch cases teperatures. On the other han, the transforer agnetic core has two to three ties ore losses than the inuctor agnetic core. The ringing phenoena escribe in section 4 explain this result. But nevertheless, this rawback oes not offset the other avantages of the selecte structure. Table : prototypes paraeters Boost paraeters: I 68 µf / 6 an µf / O.47 µf / 6 L µh, ETD9 / 9 IXFNP ( / A/66 Ω) MOFET D DD46 (6,.6 A, schottky technology) electe converter paraeters: I 68 µf / 6 an µf / O.47 µf / µf / Transf. L 44 µh, 8, ETD9 / 9 IRF46PbF (, 7 A, R DON 4 Ω) D DD46 (6,.6 A, schottky technology) 8

10 Figure 6. Prototypes key wavefors Table : efficiencies coparison P 68 W P W Boost converter η 9. % η 9. % electe converter η 97. % η 96. % 6 onclusion This paper focuses on the voltage regulator as a key supporting equipent of a fuel cell electric power supply. As a atter of fact, this evice faces severe specifications. The present investigations are liite to sall power systes (fro a few watts to kw) an allow exhibiting an attractive topology. It is ae of a single active switch an couple inuctors. This latter coponent as a esign egree of freeo. It allows reucing the switch voltage stresses. onsequently, the MOFET technology fits uch better the switch requireents than in other stuie topologies. Moreover, the transforer has obviously leakage inuctance. In this topology, the agnetic leakage oes not inuce further voltage spikes but perit the active switch to have soft coutations. Last bus not least, this structure is siple an can be easily controlle. The paper also takes care of esigning rules. Experiental prototypes were built to valiate the theoretical behaviour. Experiental results prove that the propose converter is inee far ore efficient than a conventional boost converter. References [] J.M. Anújar an F. egura, Fuel cells: History an upating., Renewable an ustainable Energy Reviews, olue, Issue 9, Deceber 9, Pages 9-. [] J. Larinie an A. Dicks, Fuel ell ystes Explaine (n Eition) John Wiley & ons, IBN X [] A. hahin, B. Huang, J.P. Martin,. Pierfeerici, B. Davat, New non-linear control strategy for non-isolate D/D converter with high voltage ratio, Energy onversion an Manageent, olue, Issue, January, Pages 6-6 [4] International Energy Agency, Worl Energy Outlook, 9. [] European oission (), /77/E Directive on the Prootion of ity fro Renewable Energy ources in the Internal ity Market. [6] Intergovernental Panel on liate hange, The fourth Assessent Report, 7 [7] OED Publishing, Energy Technology Analysis, Prospects for hyrogen an Fuel ells, International Energy Agency, Dec.. [8] Encyclopeia of Energy, eite by John Zuerchik, Macillan,, IBN [9].M. ishnyakov, Proton exchange ebrane fuel cells, Ph D Thesis, Université e ergy-pontoise,. "Proton exchange ebrane fuel cells", acuu, olue 8, Issue, August 6, Pages -6 [] F. Barbir, PEM Fuel ells, New York: Elsevier Acaeic Press,. [] O. Bethoux an J. athelin, Etue coparative e régulateurs e tension éiés à es générateurs pile à cobustible, EF9 onference, Electrotechnique u Futur, opiègne 9, France [] G. Kovacevic, A. Tenconi, R. Bojoi, Avance D D converter for power conitioning in hyrogen fuel cell systes, International Journal of Hyrogen Energy, olue, Issue, June 8, Pages -9 [] R. W. Erickson an D. Makisovic, Funaentals of power electronics, n eition, UA: Kluwer Acaeic Publishers;, IBN [4] L. Huber an M. M. Jovanovic, A esign approach for server power supplies for networking, in Proc. IEEE- APE onf.,, pp [] X. G. Feng, J. J. Liu, an F.. Lee, Ipeance specifications for stable c istribute power systes, IEEE Transactions on Power Electronics, vol. 7, pp. 7 6, Mar.. 9

11 [6] D. Maksiovic,. uk, witching converters with wie D conversion range, IEEE Transaction on Power Electronics, 99;6(): 7. [7] L. Barbosa, J.B. ieira, L.. e Freitas, M.. ilela,.j. Farias, A buck quaratic PWM soft-switching converter using a single active switch. IEEE Transactions on Power Electronics 999; 4():44. [8].M. Pacheco, A.J. Do Nasciento,.J. Farias, J.B. ieira, L.. e Freitas, A quaratic buck converter with lossless coutation, IEEE Transactions on Inustrial Electronics ; 47():64 7. [9] N. Mohan, T. M. Unelan, an W. P. Robbins, Power Electronics, John Wiley & ons,, r Eition. [] L. alazar; P.D. Ziogas, A high frequency forwar D/D converter topology with transforer flux balancing capability, Inustry Applications ociety Annual Meeting, 988., onference Recor of the 988 IEEE -7 Oct. 988 Page(s):78-79 vol. [] K. W. Ma an Y.. Lee, An integrate flyback converter for c uninterruptible power supply, IEEE Transaction on. Power Electronics, vol., pp.8 7, March 996. [] Q. Zhao, F.. Lee, High-efficiency, high step-up D- D converters, IEEE Transactions on Power Electronics, olue 8, Issue, Part, Jan. Page(s):6 7 [] Muhaa H. Rashi, Power Electronics Hanbook, n eition, Elsevier Acaeic Press, 7, IBN [4]. Lefebvre an F. Miserey, «oposants à seiconucteur pour l électronique e puissance», Lavoisier, (4). [] N. Piette, E. lavel an Y. Marechal, «Optiization of cabling in power electronics structure using inuctance criterion», Inustry Applications onference, oct Thirty-Thir IA Annual Meeting, vol., pp: [6] E. lavel, Y. Marechal an J. Rouet «Moélisation es câblages en électronique e puissance: apport et copléentarité es ethos éléents finis et e circuits équivalents», The European Physical Journal Applie Physics, EPJAP, pp:, 998. [7] J-L. chanen,. Guérin, J. Rouet, an Gérar Meunier, Influence of a conuctive plane on loop inuctance, IEEE Transactions on agnetic, vol. n, ay 99. [8] M. Zahn, «Electroagnetic fiel theory: a proble solving approach», John Wiley, 979.