Energy Conversion and Management 48 (2007) 30013008 [Link].

com/locate/enconman

Super-capacitors fuel-cell hybrid electric vehicle optimization and control strategy development
Vanessa Paladini, Teresa Donateo, Arturo de Risi *, Domenico Laforgia
` Universita degli Studi di Lecce, Dipartimento di Ingegneria dellInnovazione, Via Monteroni 73100 Lecce, Italy Available online 28 August 2007

Abstract In the last decades, due to emissions reduction policies, research focused on alternative powertrains among which hybrid electric vehicles (HEVs) powered by fuel cells are becoming an attractive solution. One of the main issues of these vehicles is the energy management in order to improve the overall fuel economy. The present investigation aims at identifying the best hybrid vehicle conguration and control strategy to reduce fuel consumption. The study focuses on a car powered by a fuel cell and equipped with two secondary energy storage devices: batteries and super-capacitors. To model the powertrain behavior an on purpose simulation program called ECoS has been developed in Matlab/Simulink environment. The fuel cell model is based on the Amphlett theory. The battery and the super-capacitor models account for charge/discharge eciency. The analyzed powertrain is also equipped with an energy regeneration system to recover braking energy. The numerical optimization of vehicle conguration and control strategy of the hybrid electric vehicle has been carried out with a multi objective genetic algorithm. The goal of the optimization is the reduction of hydrogen consumption while sustaining the battery state of charge. By applying the algorithm to dierent driving cycles, several optimized congurations have been identied and discussed. 2007 Elsevier Ltd. All rights reserved.
Keywords: Hybrid vehicle; Fuel cell; Super-capacitor; Optimisation

1. Introduction The large interest around fuel cells technology is motivated by two big issues currently the world is facing with: the ghost of the petroleum reserves depletion and consequent oil dependence; the earth global worming due to vehicles emissions. In order to address these problems a strong input has been given to fuel cells development. Fuel cells represent a big promise of clearness and equal energy distribution. However, fuel cells vehicle are still far, mainly in terms of costs, to achieve current vehicles. Many researchers have demonstrated that to achieve performance similar to internal combustion engines a single electric system (like a fuel cell) is not enough [1]. Many eorts are devoted to the study of fuel cell hybrid vehicles, where the main power is provided by the fuel cell and a battery
*

Corresponding author. Tel.: +39 0832 297757; fax: +39 0832 29756. E-mail address: [Link]@[Link] (A. de Risi).

is utilized as buer device. However, the battery presents some drawbacks, such as low cycle-life, long recharging time and low power density. These issues can be overcome adding a super-capacitors that boast an excellent cycle-life, in comparison with batteries. Moreover, they can provide high power for accelerations and can recover more eciently power during regenerative braking. Due to these properties, they are well suited to be used as buer energy sources together with batteries. The simultaneous use of battery and super-capacitor as secondary energy storage systems is a promising solution because both the high energy density of batteries and the high power density of super-capacitors can be exploited. Nevertheless, such a system can compete with internal combustion engine in terms of specic energy consumption but still not in terms of power density. Considering the tank to wheel process, the fuel cell system, compared to a 1.3 L Diesel engine, needs around half of the energy required by the Diesel engine, as it will be shown later on in the paper. The power density

0196-8904/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/[Link].2007.07.014

3002

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008

of the Diesel engine, on the contrary, is still higher compared to the fuel cells system. 2. Vehicle conguration The considered fuel cell powered hybrid vehicle is congured as shown in Fig. 1. The energy system is composed by a Proton Exchange Membrane Fuel Cell (PEMFC) fed with compressed hydrogen, a nickel metal hydrides Ovonic battery [2] and a Maxwell [3] super-capacitor. All the above three elements are connected in parallel and power is drawn according to an appropriate control strategy. The energy system is coupled to the electric motor through a DC/DC converter (which adapts the voltage to the motor controller voltage) and a DC/AC inverter. The energy system has been dimensioned to be able to approach a speed of 60 km/h starting from zero in 10 s. The fuel cell is used both to power the electric motor and to recharge the battery or the super-capacitor when the stored energy becomes
Hydrogen Storage Battery Fuel Cell Supercapacitor

lower than a specied threshold. Moreover, the battery and the super-capacitor are recharged via regenerative braking. 3. The ECoS code The vehicle conguration has been simulated with a Matlab/Simulink code schematized in Fig. 2. The instantaneous power requested by the vehicle is calculated using a backward paradigm for an arbitrary pre-dened driving cycle. To calculate the power request the aerodynamic resistance, the rolling resistance, the grade force and the inertial term are taken into account. During deceleration, the negative power on the vehicle is assumed to be recovered with an ideal regenerative brake; the brake system is not modeled, whereas in the model a battery recharging eciency of 80% is considered. According to the control strategy, the supervisor, in the motor controller, manages the energy ows between fuel cell, battery and super-capacitor. Table 1 reports the input and the output variables used in the simulation code and the control strategy parameters. The same inputs were also taken into account in the optimization process described below. 4. Energy system The electric vehicle considered in the present investigation is powered by a fuel cell system, as primary power source coupled with a battery and a super-capacitor to face peak power requests. Thus, the total electrical power available to the vehicle is given by the sum of the fuel cell power Pfc, super-capacitor power Puc and battery power Pbatt: P em P fc P uc P batt 4.1. Fuel cell Fuel cells are electrochemical devices which convert chemical energy into electrical energy directly by oxidizing 1

Motor Controller Motor/ Gen. Gear

Fig. 1. Vehicle conguration.

FC Pwr Req. Battery Pwr Req.

H2 Req.

FC

IF >0 Cycle Car Pwr Req. Total Pwr Req. FC Aux. Pwr Req.

Control Strategy

Battery SOC Battery MC

Super Cap. Pwr Req. IF <0 Breaking Power Super Capacitor SOC

Super capacitor

Fig. 2. Flowchart of the simulation code ECoS.

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008 Table 1 Code input and output variables Code input h N A B_SOC_min B_SOC_max UC_SOC_min UC_SOC_max Acc Code output H2 B_energy FC_use B_use Fuel cell operating eciency Number of fuel cell stacks Fuel cell active area Minimum battery SOC Maximum battery SOC Minimum super-capacitor SOC Minimum super-capacitor SOC Percentage of acceleration Cycle hydrogen consumption Finally battery energy Fuel cell use Battery use

3003

present in the cathode channels, an excess of air (e) has been used. In the present investigation a value of e = 2 has been used as suggested by Larminie et al. [5]. The result of the chemical reactions inside a fuel cell is the reversible single electrode potential or open circuit voltage Er, which can be calculated by means of the Nernst Eq. (3): "  !# p H2 O RT Er E0 ln 3 p nF p H2 x p O 2 where p represents the partial pressure and R is the universal constant. However, mainly because of activation polarization, fuel crossover, internal current and mass transport losses, the theoretical value Er cannot be reached. For each electrode, the relation between the activation over voltage and the current density is described by the Tafel equation:   i in AT ln 4 i0 The constant AT is higher for slow electrochemical reactions. The constant i0 is the exchange current density and it is higher if the reaction is faster. The constant in is the crossover current. The voltage drop that corresponds to the ohmic loss is proportional to the current: i Rohm 5

hydrogen without intermediate thermal or mechanical processes. Proton exchanges membrane fuel cells (PEM-FC) are usually preferred in automotive applications because they are compact and they have a short start up. The utilization of precious metal as catalyst allows to increase the reaction rate, reducing as consequence the fuel cell operation temperature at about 353 K (compared for instance to the 1000 K required by a solid oxide fuel cell). The fuel cell is modeled with a phenomenological model which uses empirical and semi-empirical equations based on the theory of Amphlett [4] to determine the currentvoltage characteristics. For the fuel cell the following assumptions were taken: Steady-state operation and one-dimensional conditions. Uniform temperature. Oxygen and hydrogen behave like ideal gases. No humidity changes. No gas diusion resistances at the electrodes. No chemical kinetics limitations.

The ohmic resistance Rohm, varies with the porosity of the gas diusion layer and the contact area between the bipolar plates and the gas diusion layer. The ohmic resistance is a function of the membrane conductivity, and it is obtained by integration over the membrane thickness. Concentration overvoltage is the result of resistance to mass transport of gases to the reaction sites. An equation that approximates the voltage drop from concentration losses due to water management issue is given by Kim [6]: mtrans expntrans i
5

In Table 2 the operating parameters selected for the fuel cell system are reported. Hydrogen and oxygen ow rates are calculated as function of the current requested by the power demand as follows: I 4F 2 I _ m H2 M H2 2F where M O2 and M H2 are, respectively, oxygen and hydrogen molar mass and F is the Faraday constant. Then, the stoichiometric air ow rate is calculated. To ensure that sucient air is present and to aid in the removal of the water _ m O2 M O2
Table 2 Fuel cell operating conditions Operating fuel cell temperature Electrodes relative humidity Number of cells in the stack Stack voltage Platinum catalyst 353 K 90% 80 48 V 0.4 g/cm2

The value of mtrans is typically about 3 10 V and ntrans about 8 103 cm2/mA. Both mtrans and ntrans relate to water management issues: a partially dehydrated electrolyte membrane leads to a decrease in conductivity, which can be represented by mtrans, whereas an excess in liquid water leads to a reduction in porosity and hence to an early onset of mass transport limitations, which can be captured by the parameter ntrans. Considering the above-mentioned losses, the fuel cell potential can be expressed by the following equation:   i in V fc Er AT ln i Rohm mtrans expntrans i i0 7 Accordingly to Eq. (7), the gross electrical power provided by the cell is calculated as follows: P TOT 2V fc F _ mH2 M H2 8

3004

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008

However, in a fuel cell system this power is never provided entirely to the user, due to the power required by auxiliary components, Paux. The auxiliary components considered in the present investigation are: compressor Pcompressor, coolant pump Ppump, radiator fan Prad and condenser fan Pcond. P aux P compressor P pump P rad P cond 9 The compressor is the most power demanding auxiliary component of the fuel cell system. The theoretical compressor power is given by the isentropic gas compression and it is denes as following: " # c1 T atm pair c _ P compressor C p 1 mair 10 gcp patm where Cp is the specic heat capacity, Tatm is the atmospheric temperature, gcp is the compressor eciency, p is the partial pressure, c is the specic heat ratio. On the basis of literature results [7], the other auxiliary components (coolant pump, radiator fan and condenser fan) are assumed to take up 1150 W of constant power. The net power provided to the vehicle by the fuel cell is then given by: P fc P TOT P aux 11 The fuel cell voltage and power versus current density calculated by the proposed model are reported in Fig. 3. The current density is given by the following relation: i I A 12

4.2. Battery Nickel metal hydride battery has been considered in this investigation. In Table 3 the battery characteristics are reported. The internal resistance maps have been obtained by literature [9], while the current was calculated as function of the requested power using the following equation: q Er E2 4Rint Rt P b r 13 I bat 2Rint Rt where Ibat is the battery current, Rint is the internal battery resistance and Rt is the thermal ohmic resistance. The battery state of charge (SOC) is calculated using the Current Integration Method implemented in a similar fashion as in Advisor [10]. A nickel metal hydride battery has been chosen instead of a lead-acid battery due to the higher gravimetric energy content of the former compared to the last one (6090 kW/kg for the nickel metal hybrid against 3550 kW/kg of the lead acid battery). Moreover, a nickel metal hydride battery presents a higher volumetric energy in comparison to a lithium ion battery (about 350 W h/L for a nickel metal hydride and about 300 W h/L for a lithium ion battery) [1]. 4.3. Super-capacitor The simulated super-capacitor is a Maxwell PC2500 whose characteristics are reported in Table 4. The reason for considering a super-capacitor in the vehicle setup is its high specic power rate and its ability to accept a full recharge in a very short time, which leads to an improvement in the vehicle eciency and energy economy. The use of super-capacitor is especially benecial if energy braking has to be recovered. 4.4. Control strategy The control strategy has to assure that the power delivered by the electric motor meets, at any time, the power demand. For the sake of simplicity, the response of the energy system to the power request has been assumed instantaneous. The power demand can be satised either by the fuel cell or by batteries/super-capacitors. A control strategy where a fuel cell is used as primary source and the battery is used as buer energy source has been already shown to perform better than the reverse situation, where the fuel cell meets only peak loads [11]. Thus, this strategy

In the formula here I is the total current and A is the membrane active area. Temperature and pressure eects are not included in this model. Moreover, according to the results of Hamelin et al. [8] the fuel cell system is assumed to respond almost instantaneously to load variations.

Table 3 Battery specications [7] Package voltage Package capacity Dimension Weight Power density Energy density V Ah mm kg W/kg W h/kg 48 2 340 74 135 33.6 230 64

Fig. 3. FC system power (green line) and cell potential (solid line).

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008 Table 4 Super-capacitor specications Package voltage Package capacity Weight Power density Energy density V Ah kg W/kg W h/kg 48 45 13.5 1838 7

3005

has been adopted in this study. The additional analysis of the present investigation is the exploitation of two secondary energy devices: a battery together with a super-capacitor. However, the use of two secondary energy devices requires a more complex control strategy with respect to the previous investigation [11]. The implemented strategy for power split and charge sustaining is schematized in Fig. 4 and it was derived on the basis of the larger energy density of batteries and the higher power density of supercapacitors. The power requested by the electric motor(s) is supplied by the power system in the following order: (1) fuel cell, (2) super-capacitor, (3) battery. As primary source the fuel cell is used, because employing it directly for traction instead of recharging the battery is more energy ecient. In addition the super-capacitor, compared with battery, can be easily recharged and discharged; while the battery is invoked only in high power demand situations. In other words, the following three conditions are identied. As long as the required power is lower than the available fuel cell power, the fuel cell acts as the only power source; when the power demand becomes greater than

the fuel cell power, it supplies its rated power and the remaining part is supplied by the super-capacitor. If the requested power exceeds the sum of fuel cell and supercapacitor power, the battery is also employed. Note that the battery and the super-capacitor are restricted to work in a pre-dened range of state of charge (SOC). Out of these ranges unfeasible situations may happen if these elements are not properly recharged. Both the battery and the super-capacitor can be recharged by the fuel cell or by the energy recovered during vehicle braking. While the vehicle is in traction and the power request is below the fuel cell rated power, the two secondary storage systems can be recharged by the fuel cell surplus energy. In braking, the choice between recharging the battery or the super-capacitor depends on the degree of deceleration. Weak decelerations are exploited for recharging the battery, while strong decelerations are employed for recharging the super-capacitor due to its higher power density. By an analysis of electric vehicles characteristics and the four selected driving cycles (NEDC, UDDS, HWFET, 10 15) it has been observed that the deceleration of the vehicle usually falls in the range of [1/5;0] m s2. Thus, the control strategy selects a value in this range as threshold which discriminates if either the battery or the super-capacitor has to be recharged by braking energy recuperation. This threshold is one of the control parameters to be optimized. In designing a hybrid electric vehicle powered by fuel cell the main goal to be reached is the reduction of fuel

Preq

yes
If Preq<0 brake

no
If Preq<PFC_rated

yes

PFC=Preq

no
If Preq<PFC_rated+PUC

yes

If SOCUC<SOCUC_min

yes

yes
If SOCB<SOCB_min unfeasible

no no
PFC=PFC_rated PUC=Preq-PFC

no
PFC=PFC_rated PB=Preq-PFC

yes
If Preq<PFC_rated+PUC+PB If SOCUC<SOCUC_min If SOC_B<SOCB_min

yes
unfeasible

no
unfeasible PFC=PFC_rated PUC=Preq-PFC PB=Preq-PFC-PUC

Fig. 4. Traction control strategy.

3006

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008

consumption of a pre-dened driving schedule. However, a charge sustaining strategy is usually chosen to avoid the battery to be fully discharged, at the end of the cycle. In this way, the user does not need to recharge the system like in the pure electric vehicle since the battery state of charge is always kept higher than a minimum level. To fulll this specication, the nal state of charge of the battery is controlled in the present investigation and compared with the initial one. 5. Optimization method In order to achieve the reduction of fuel consumption and allow for battery charge sustaining, the vehicle congurations and the control strategies have to be contemporary optimized. In the present investigation the input parameters of Table 5 have been considered and allowed to vary in a certain range. The vehicle conguration and the control strategy obtained with each combination of the input parameters are evaluated with the developed Matlab/Simulink code and the output parameters (fuel consumption and nal state of charge of the battery) are calculated. The aim of the present optimization is to nd a combination of the input parameters that minimizes both the fuel consumption and the dierence between initial and nal battery state of charge. Since these two targets are in competition to each other, this can be treated as a multi-objective problem which can be optimized using a Multi-Objective Genetic Algorithm (MOGA) [1214]. The algorithm is called genetic because, just like in nature, the dierent input parameters are crossed to generate the outputs. The input parameters that better than other contribute in meeting the output requirements have the highest probability to be further selected in the [Link] allows the best combination of the parameters to be preserved in the next generation. At some point an abrupt change in the input parameters selection is introduced, to explore combinations of parameters that may not occur otherwise. After several iteration, the set of parameters with the best outputs is obtained. Genetic algorithms have been developed for singleobjective optimization but they can be easily applied to multi-objective problems due to their intrinsic parallelism. Since genetic algorithms work with a population of soluTable 5 Parameters range Minimum g n A B_SOC_min B_SOC_max UC_SOC_min UC_SOC_max Acc % cm2 % % % % % 37 1 350 55 71 10 70 10 Maximum 46 4 500 69 85 30 90 90

tions, multiple optimal outputs can be captured in a single run. The optimization result in the case of multi-objective problems is not a single optimal solution, but a group of design parameters which give a dierent compromise between fuel consumption and battery state of charge. For each possible solution (i.e. combination of the input parameters of Table 5) a vector of tness values is assigned. In the present case, the components of the tness vector are the fuel consumption and the variation of the battery state of charge which have to be minimized. F 1 H2 F 2 B SOCfinal B SOCinitial 14 15

The tness vectors of the possible solutions are compared and the non-dominated solutions are selected. In minimization problems, the tness vector of a possible solution is said to dominate another solution if all its components are inferior or equal to the other possible solutions and at least one of the components is absolutely inferior. Possible solutions whose tness vectors are non-dominated are called Pareto optimal and they form the so called Pareto front. Thus, the Pareto front in multi-objective optimization problems represents the optimal set of solutions, where each solution leads to a dierent compromise in the achievement of the optimization goals, so it represents the group of solution that provides the best compromise between the two outputs. Genetic algorithms are one of the most widespread methods used to dene the Pareto front in practical applications. In fact, they are very easily combined with existing simulation codes and can be applied to all kind of optimization problems. Genetic algorithms also allow for the use of feasibility constraints. In the present analysis, the optimization has been carried out with two constraints. The rst limited the maximum fuel consumption: only the designs with hydrogen consumption smaller than 200 g have been considered. This limit has been set considering the length of the investigated driving cycles; higher hydrogen consumption would lead to unacceptable vehicle eciency. The second constraint, instead, is used to preserve the total amount of energy stored in the battery. Therefore, only the designs with a nal battery state of charge at least equal to its initial value have been considered feasible (the initial battery state of charge is set equal to 70%). The optimization has been carried out on four dierent driving cycles which are usually exploited for emission regulation in Japan (e.g. 1015 modes), USA (e.g. UDDS and HWFET) and Europe (e.g. NEDC). The design space is the same for all the cycles and is dened by the variation range of the input variables in Table 5. 5.1. Results of the optimization In Fig. 5 the Pareto fronts obtained with respect to the four driving cycles are reported. Looking at the gure from left to right, the cycles are located as following: 1015,

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008

3007

NEDC UDDS HWFET 1015

50

100 150 hydrogen (g)

200

250

Fig. 5. Pareto front for each driving cycle.

UDDS, NEDC and HWFET. This order follows the average speed of the cycles. It is worthwhile to note that even if the energy requested to complete an UDDS cycle is greater than the required energy for the NEDC, the results of Fig. 5 show that a better fuel economy can be achieved in the UDDS cycle compared with the NEDC. This is due to the peculiarity of the UDDS cycle which is characterized by a uctuating velocitytime trace very wavy that allows larger energy recovery with the battery and the super-capacitor. The optimal solutions in each Pareto front dier mainly as function of the following input parameters: number of stacks n, fuel cell eciency g, battery maximum state of charge B_SOC_max, and super-capacitor minimum state of charge UC_SOC_min. In particular, hydrogen consumption increases by increasing the number of stacks n and decreasing the fuel cell eciency g. At the same time lower fuel consumption
Table 6 Optimal set of input congurations for each driving cycle NEDC (new European driving cycle) Powertrain components inputs Fuel cell average eciency Fuel cell # of stacks Fuel cell active area Minimum battery SOC Maximum battery SOC Minimum super-capacitor SOC Maximum super-capacitor SOC Electric motor maximum output power System power density Simulations outputs Total fuel consumption (H2) Cycle length Specic fuel consumption Specic energy consumption (120 MJ/kg H2)

is obtained for low B_SOC_max and low UC_SOC_min. Therefore, the battery should be forced to work in a narrow range of SOC (this is also benecial to the battery life-cycle) while the super-capacitor should be exploited until very low level of its state of charge. For the HWFET a larger fuel cell active area is obtained in all the optimal solutions, which resulted in higher fuel consumption. This is because the HWFET is the most demanding cycle in terms of power. From the Pareto fronts shown in Fig. 5, for each driving cycle curve, the point of minimum fuel consumption has been selected. In Table 6, the output fuel consumption for each cycle is reported; moreover, for each of this value the input parameters which have allowed getting these gures are reported. Note that the optimal input parameters do not dier much among the cycles. The average eciency of the fuel cell for all tested driving cycles has been found to be about 46%, while the optimized active area of the fuel cell ranges between 350 and 374 cm2. The most favorable range of SOC for the battery is between 62% and 81%, and for the super-capacitor is between 21% and 76%. Concerning the deceleration threshold which denes the amounts of the braking energy to be used to recharge, respectively, the battery and the super-capacitor, the optimum has been found for a deceleration of about 0.7 ms2. Knowing the energy uxes for the fuel cell, the battery and the super-capacitor, during the entire driving cycle, the utilization percentage for the three components can be calculated and the results are reported in Table 7:
R Utilizationi % R
cycle

P FC tdt

R cycle
cycle

P i tdt P B tdt

R
cycle

P SC tdt

100

16

UDDS (urban dynamometer driving schedule) 46 1 350 66.6 85.0 19.0 74.5 60 275 84.53 11.99 7.05 84.6

HWFET (high way fuel economy test)

1015 (Japanese driving schedule)

% cm2 % % % % kW W/kg g km g/km MJ/ 100 km

45.55 1 365 67.6 76.4 20.0 82.0 60 275 76.66 11.36 6.75 81

45.55 2 363 59.6 81.2 18.0 74.5 60 275 139.85 16.51 8.47 101.6

46 1 374 55.0 82.0 29.0 75.0 60 275 37.09 4.16 8.92 107

3008

V. Paladini et al. / Energy Conversion and Management 48 (2007) 30013008

Table 7 Percentage of utilization of each storage system NEDC Fuel cell Battery Super-capacitor % % % 53 12 34 UDDS 52 4 44 HWFET 69 2 29 1015 66 2 32

where i may be either the fuel cell (FC) or the battery (B) or the super-capacitor (UC) and Pi(t) is the instantaneous power that the energy element (either fuel cell, or battery or super-capacitor) provides to the vehicle. From the data reported in Table 7 it appears that the optimal utilization for the fuel cell, the battery and the super-capacitor changes signicantly by changing the driving cycle. This is a clear signal that there is not silver bullet solution for the utilization strategy of the energy elements, but it strictly depends on the driving schedule. For instance, in the HWFET, that is a high way driving cycle, the fuel cell is more used that in the other cycles, because in this case, due to lack of strong braking phases the storage units do not have much change to be recharged. However, as shown by the results of Table 6 the powertrain conguration, as well as the control strategy parameters are not very sensitive to the driving cycle. Finally, in Table 6 two gures related to the entire power system (fuel cell, hydrogen tank, battery, supercapacitor and electric motor) have been reported: the system power density and the specic energy consumption. The power density is the power that the system is able to generate over 1 kg of system. In the system under investigation it is around 275 W/kg. Considering a 1.3 L Diesel engine with 80 Hp and a tank with enough fuel to run a vehicle for around 500 km, this conventional system has got a power density of almost 400 W/kg, that means that conventional engines have got power density about 30% higher than fuel cell system. On the contrary, if the same comparison is done calculating the specic energy consumption, that is the energy required to cover the same distance the gures are quiet the opposite (considering the tank to wheel process, so without considering the costs from fuel production until distribution at the lling station). A conventional vehicle powered by a 1.3 L Diesel internal combustion engine requires about 5 L of fuel per 100 km. Assuming for Diesel fuel a lower heating value of 44 MJ/kg with a density of 0.78 kg/L, the specic energy consumption for a standard car is 171.6 MJ/100 km. In the case of the fuel cell system this value waves between 71 MJ/100 km and 107 MJ/ 100 km. In the worse case the specic energy consumption of a fuel cell system is about 40 % lower than a conventional vehicle. 6. Conclusions A Matlab/Simulink code has been written to analyze fuel cells based hybrid electric vehicles. The conguration considered in the present investigation uses a PEM fuel

cell as primary power source while both batteries and super-capacitors are used as secondary energy buers. The aim of the investigation was to optimize both the vehicle design and the control strategy parameters in order to minimize fuel consumption while preserving the battery state of charge with respect to dierent driving cycles. From the results of the optimization, the super-capacitor was found to capture a signicant portion of the braking energy and thus to improve fuel economy for all cycles. The optimal values of the control and design parameters, identied with the use of genetic algorithms, were found to be quite independent of the driving cycle, whereas the energy elements utilization should be adapted according to the particular driving schedule. This is an important result because the optimal congurations can assure good fuel consumption for all the investigated cycles and thus can be expected to work well on the real driving conditions of the vehicle. Finally, the fuel cell system appears to have power density 30% lower compared to internal combustion engine, but also specic energy consumption 40% lower. Acknowledgement This research has been supported by the Ministry of Instruction, University and Research (MIUR). References
[1] Winter M, Brodd RJ. What are batteries, fuel cells and supercapacitors? Chem Rev 2004;104:424569. [2] Ovonic website, [Link] [3] Maxwell website, [Link] [Link]. [4] Amphlett C, Baumert RM, Mann RF, Peppley BA, Roberge PR. Modeling of automotive fuel cell operation in driving cycle. J Electrochem Soc 1991:142. [5] Larminie J, Dicks A. Fuel cell systems explained. 2nd ed. Wiley; 2003. [6] Kim J, Lee SM, Srinivasan S, Chamberlin CE. Modeling of a proton exchange membrane fuel cell performance with an empirical equation. J Electrochem Soc 1995;142(8):26704. [7] Gurski SD, Nelson DJ. Cold start fuel economy and power limitations for a PEM fuel cell vehicle. SAE 2003-01-0422. [8] Hamelin J, Agbossou K, Laperriere A, Laurencelle F, Bose TK. Dynamic behavior of a PEM fuel cell stack for stationary applications. Int J Hydrogen Energ 2001;26:6259. [9] Schell A, Peng H, Tran D, Stamos E, Lin C, Kim MJ. Ann Rev Contr 2005;29:15968. [10] NREL. Vehicle system analysis homepages-Advisor Website [Link]/analysis/[Link]. [11] de Risi A, Paladini V, Donateo T, Laforgia D. Control strategy optimization of a fuel-cell electric vehicle. In: ASME EFC05 technical conference; 2005. p. 146. [12] Goldberg DE. Genetic algorithms in search, optimization and machine learning. Addison-Wesley Longman Inc; 1999. [13] Donateo T, de Risi A, Laforgia D. Optimization of high pressure common rail electro-injector using genetic algorithms. SAE 2001-011980. [14] de Risi A, Donateo T, Laforgia D. A new advanced approach to the design of combustion chambers in diesel engines. Int J Vehicle Des 2006;41:16587.