Nonlinear Control of Fuel Cell and Battery for Electrical Vehicle with Energy Supervision under Source Limitations

M. Becherif,a , D. Hisselb , M.Y. Ayadc

a IEEE

member, University of technology of Belfort-Montbliard, FEMTO-ST/FCLab, UMR e CNRS 6174, 90010 Belfort Cedex, France b IEEE senior member, University of Franche Comte, FEMTO-ST/FCLab, UMR CNRS 6174, 90010 Belfort Cedex, France c IEEE member, in R&D in Industrial Hybrid Vehicles Applications, France

Abstract Electrical vehicle propelled by a hybrid source based on of fuel cell (FC) and batteries is presented. FC does not work properly under transient conditions. The sharp changes in the load power demand can lead to electrochemical and thermal non-uniformities in FC. These non-uniformities increase the degradation rate of FC and reduce its expected life span. To mitigate these detrimental eects, a battery bank is used along with FC. This paper deals with the state space modelling of the hybrid system, and the control of the DC Bus voltage by the Interconnection and Damping Assignment Passivity-Based Control (IDAPBC). IDA-PBC is a useful technique to control systems assigning a desired Port-Controlled Hamiltonian (PCH) structure to the closed-loop. In real working conditions, sources (FC and Batteries) are subject to limitations (Hydrogen for FC and discharge depth for batteries), these source limitations are taken in account using the Fuzzy Logic (FL) techniques allowing the energy supervision of the electrical vehicle. The battery model and parameter equations are given in function of the state of charge and temperature, and are experimentally validated. The global stability proof and simulation results are presented and validated by experimentation. Key words: Passivity-Based Control, Interconnection and Damping, Hybrid vehicles, Fuel Cell, Fuzzy Logic.

1. Introduction Automobiles are one of the major sources of air pollution in urban areas. As air pollution in heavily populated areas becomes unbearable, the search for
author. Tel. +33 (0)3 84 58 33 46. Fax +33 (0)3 84 58 33 42 Email addresses: mohamed.becherif@utbm.fr (M. Becherif), daniel.hissel@univ-fcomte.fr (D. Hissel), ayadmy@gmail.com (M.Y. Ayad)
Corresponding

Preprint submitted to Elsevier

August 1, 2011

cleaner alternatives emerges as an imperative. In an electric vehicle using a single energy source, the necessary power is transferred from the permanent source, a FC for example, to the load. The permanent source must frequently supplies (or absorbs in the case of battery vehicle) the peaks of power resulting from the accelerations and the braking. This double uses of the permanent source, as energy source and as power source, is strongly penalizing: the losses, volume and the weight are increased and the lifetime of the energy source is reduced [1], [2], [3], [4], [5]. One solution to this problem is the hybridization of the source with a battery that manages the power peaks. Hence the permanent source can only supply the average power which ensures the vehicles energy autonomy. The battery has the characteristics of high energy density and relatively low power density (but still 3 to 5 times higher than the power density of a FC). Energy management and fuel consumption minimization have been early studied by researchers [6] and still are an active research area [7], [8], [9], [10] and [11]. The battery provides power to the vehicle during periods of peak power demand such as vehicle acceleration or climbing a mountain. In the studied application, the battery is also considered for braking energy recovery. Hybrid sources allow dissociating mean power sizing from peak transient power sizing, the aim being to reduce in volume and weight of the energy supply source and to ensure a smooth behaviour for the FC for example [12, 13, 14]. To match these aims, the IDA-PBC technique is applied using an open loop control for the FC to regulate the DC bus voltage and a closed loop control for the battery to control the battery current towards its reference. In real conditions, sources are subject to limitations, as for example quantity of the embedded hydrogen for FC and the discharge depth of the battery. These two constrains must be taken to account in the control task to manage energy ows between sources. A Fuzzy Logic technique is used to reach this supervision task. FL has as input the State of Charge (SOC) of the battery and the amount of available H2 and gives as output the battery current reference. A negative sign of the battery current provokes the charge of the battery and a positive sign a discharge phase. The amount of available H2 is provided by the hydrogen fuel tank sensor, and the SOC of the battery is obtained using a lookup table. This lookup table was obtained experimentally by the characterization of a battery pack available in our Lab. and this characterization takes in account the voltage, temperature, ... to deliver the SOC. In [15], authors propose an online strategy based on a fuzzy decision system. Fine tuning of the fuzzy system parameters, mainly the membership functions, is made possible using a powerful optimization tool based on a genetic algorithm. The proposed optimization procedure takes into consideration the minimization of the hydrogen consumption while satisfying the requested power over a given driving cycle. The originality of our work, is that the FL supervision ensures a good behaviour of the system and optimizes the energy ow regarding the remaining H2 and the battery discharge depth; moreover no driving cycle is imposed. Then, an online operation is possible. In [16], authors proposed to apply the fuzzy logic technique to manage the energy for an embedded FC 2

system. In a rst step, a dynamic modelling of the overall system is given. Secondly, this system is written in a PCH form where important structural properties are exhibited. Then a Passivity-Based Control of the system is presented using IDA technique. The global stability proof of the equilibrium with the proposed control laws is given. The energy supervision under hydrogen and discharge depth limitations is presented and solved using FL. Finally, simulation results in presence of DC Bus voltage changes and load resistor disturbances, using Matlab, are presented and validated experimentally using the designed experimental setup. 2. FUEL CELLS The developments leading to an operational FC can be traced back to the early 1800s with Sir William Grove recognized as the discoverer in 1839. A FC is an energy conversion device that converts the chemical energy of a fuel directly into electricity. Energy is released whenever a fuel (hydrogen) reacts chemically with the oxygen of air. The reaction occurs electrochemically and the energy is released as a combination of low-voltage DC electrical energy and heat. Types of fuel cells dier principally by the type of electrolyte they utilize Fig. 1. The type of electrolyte, which is a substance that conducts ions, determines the operating temperature, which varies widely between types [17, 18, 19, 20, 21]. Proton Exchange Membrane (or solid polymer) Fuel Cells (PEMFCs) are presently the most promising type of FC for automotive use and have been used in the majority of prototypes built to date.
enarbmeM sreyal noitcaeR

Figure 1: Dierent layers of an elementary cell. Source [17]

As the gases are supplied in excess to ensure a good operating of the cell, the non-consumed gases have to leave the FC carrying with them the produced water [22]. 3. Battery The battery has the characteristics of high energy density and relatively low power density (but still times higher than the power density of a FC). The in3

retaw ropav dna riA

setalp ralopiB

C a ht o d e reyal noisuffiD

setalp ralopiB

A n o d e

retaw ropav dna negordyH

ternal resistance is the major factor for the limited discharging and charging current capability. The internal equivalent series resistance has dierent values under charging and discharging operating conditions. The charging and discharging eciency are nonlinear functions of current and SOC. The battery can be modelled as an equivalent circuit such as a voltage source and an internal resistor. Because FC and battery have advantages and disadvantages of their own, it should be benecial to have hybrid energy-power sources, in which FC system supplies the base energy while battery supplies peak power for fast acceleration and captures the braking energy for regeneration. 3.1. Experimental modelling and identication of the battery This part presents a model of a Nickel-based battery [24]. The objective is to obtain battery model and parameters in function of the temperature and the state of charge. The test are performed on a nickel cadmium Saft battery (Ccell = 60Ah, Vcell = 1.3V ). The internal battery resistance is function of the temperature T which represents the electrolyte temperature (and not the ambient temperature). Fig. 2 shows the experimental values of the internal battery resistance versus the temperature and a quadratic interpolation given by equation (2). The experimental tests were done under the discharge of the battery at a constant current. It was noticed that the battery resistance, for the Ni-type battery, is quite constant in function of the SOC. rB is experimentally obtained by using the following equation: rB = e B VB iB (1) (2)
5

rB = a1 T 2 + b1 T + c1 where T is the temperature in C, a1 = 3, 89.107, b1 = 3, 64.10 2, 35.103

2.5 2.4 2.3 2.2 2.1 rB () 2 1.9 1.8 1.7 1.6 1.5 x 10
3

, c1 =

Experimental rB Quadratic interpolation of r

B

10

15

20

25 30 Temperature C

35

40

45

50

Figure 2: Behaviour of the internal battery resistance as a function of the temperature

Fig. 3 presents the experimental battery emf versus the SOC. The linear interpolation of eB is given by equation 3. eB is measured at the cell terminals with iB = 0. eB = a2 SOC + b2 (3) where a2 = 2, 5.103, b2 = 1, 25
1.45

1.4

1.35 eB

1.3

Experimental eB Linear interpolation of eB

1.25

10

20

30

40 SOC

50

60

70

80

Figure 3: Evolution of the battery emf eB versus the SOC

The battery SOC is obtained using the following equation as given in [24]: SOC = SOCInit 100 CN iB dt (4)

with SOCInit is the initial battery SOC, CN is the rated capacity of the battery. In the simulation section, the battery is composed of 50 cells connected in serial and in parallel to obtain an emf around 12 V. 4. Port Controlled Hamiltonian system PCH systems were introduced by [25] and has since grown to become a large eld of interest in the research of electrical, mechanical and electro-mechanical systems. Some of the advantages of expressing systems in PCH form is the fact that they cover a large set of physical systems and capture important structural properties. Consider the nonlinear system given by: x = f (x) + g(x)u y = h(x) (5)

where x I n is the state vector, y I m is the output vector, f (x), g(x) and R R h(x) are locally Lipschitz functions and u I m is the control input. R A PCH form of the system (5) is given by: x = [J(x) R(x)] H(x) + g(x)u y = g (x) H(x) 5 (6)

where J(x) is an n n skew symmetric matrix, R(x) is a n n positive semidenite symmetric matrix and H is the gradient vector of the energy function H(x) of the system (5). H= H H H (x) (x) (x) x1 x2 xn
T

(7)

g (x) is a full-rank left annihilator of g(x), that is g (x)g(x) = 0. PCH systems, with H(x) non-negative, are passive1 systems. A recent and very interesting approach to solve these problems is the IDA-PBC method, which is a general way of stabilizing a large class of physical systems, see [26, 27]. IDA-PBC approach consists on identifying the natural energy function of the system called H(x), then rewrite the nonlinear system (5) versus the gradient of the energy function. 5. Hybrid DC Source using Fuel Cell as Main Source 5.1. Structure of the hybrid source As shown in Fig. 4, the studied system comprises a DC link supplied by a FC and a non reversible DC-DC Boost converter which maintains the DC voltage VDC to its reference and desired value Vd [28], and a battery storage device which is connected to the DC link through a current reversible DC-DC converter. The function of FC is to supply the mean power to the load, whereas the storage device is used as a power source: it supplies peak loads required during acceleration and braking. 5.2. Dynamic modelling The proposed hybrid structure is given by Fig. 5. It is composed by a FC as a main source, a FC DC-DC Boost converter, a DC Bus, a battery, a bidirectional current battery DC-DC converter and a RL load. The load corresponds to a DC motor. The breaking phases are not simulated because our experimental setup has a passive (RL) load. Hence, the simulated breaking phases cannot be validated experimentally. Consequently, the DC motor emf can be neglected. The load inductance models the dynamic changes in the load. Consider in the sequel: VF C , iF C , eB , iB , Vs as the FC voltage and current, battery emf and current, and the boost output voltage, respectively. rB , Cs , CDC , LDC and LB are the internal battery resistance, the boost capacitance, the DC bus capacitance and inductance and the inductance of the DC buck-boost converter, respectively.
1 A passive system is one where the stored energy cannot exceed the energy supplied to it from its environment, the dierence being dissipated.

LDC Air, O2 H2

DC Bus CDC

PEM FC
VFC
B dii r e c t

Load

FC B

VDC iB

PBC
Quantity of H2

iB

Energy supervision under source limitations using Fuzzy Logic

Figure 4: Hybrid sources structure controlled by PBC and optimized by FL

5.2.1. PEMFC model (Static model) [17] The characteristic FC voltage as a function of the FC current magnitude is presented in Fig. 6. Here, only a static model of the FC stack is considered, as the dynamic of the current (and thus voltage) remains relatively low versus the battery current. The obtained curve is composed of three main regions corresponding to three preeminent phenomena: the electrochemical activation phenomena (region 1), a linear part (region 2) where the voltage drop is due to electronic and ionic internal resistances and the last region where the diusion kinetics of gases through the electrodes becomes the limiting factor (region 3). This last zone is characterized by a brutal voltage fall. ist + in io ist + in iLim

Vst = E A.log

Rm (ist + in ) + B.log 1

Vst = f (ist ), where E is the reversible voltage of the FC, ist is the delivered current, io is the exchange current, A is the slope of the Tafel line, iLim is the limiting current, B is the constant in the mass transfer, in is the internal current and Rm is the membrane and contact equivalent resistance. The FC parameters are: Fuel Cell parameters P [W ] E[V ] io [A] Rm [m] A[V ] B[V ] iLim [A] in [A] 500 27.1 6.54 45 1.35 1.19 100 0.23

yretta B

Vd

SOC

(8)

iFC VFC

LFC

iDC
CS

LDC

DC Bus

iL
LL

TFC

VS TB

CDC

VDC
Load

RL

VL

PEMFC

iB
Battery

LB

rB eB

VB

TB

Figure 5: Hybrid sources electrical model

V
E E0

FC

region1

region2

region3

FC

Figure 6: Static FC characteristic.

5.2.2. PEMFC + FC Boost converter model dVs dt diF C dt 1 [(1 uF C ) iF C iDC ] Cs 1 [ (1 uF C ) VS + VF C ] LF C

= =

(9) (10)

If uF C = 1 TF C is closed. 5.2.3. DC Bus model dVDC dt 1 [iDC iL + (1 uB ) iB ] CDC 8

(11)

diDC dt

1 LDC

[VS VDC ]

(12)

If uB = 1 TB is closed, and TB (open). 5.2.4. Battery + Battery DC-DC converter model VB diB dt 5.2.5. RL Load model diL dt VL 1 [RL iL + VL ] LL VDC = = eB rB iB 1 [ (1 uB ) VDC + VB ] LB (13) (14)

= =

(15) (16)

the overall model of the hybrid system can be written in a state space model by choosing the following state space vector: x = = x1 , VS , x2 , x 3 , x 4 , x5 , x6 iF C , VDC ,
T T

iDC , iB , iL

The control is a vector function of the FC and Battery control laws (duty ratios of the boost and the buck-boost converters) = or u= The 6
th

F C , B

(1 uF C ), (1 uB ) uF C , uB
T

(17) (18)

order overall state space model is then : x1 x2 x3 x4 x5 x6 y = = = = = = = 1 [F C x2 x4 ] Cs 1 [F C x1 + VF C ] LF C 1 [x4 x6 + B x5 ] CDC 1 [x1 x3 ] LDC 1 [eB rB x5 B x3 ] LB 1 [RL x6 + x3 ] LL x3 9

(19)

with VF C = VF C (x2 ) given by (8). In the sequel, VF C will be considered as a measured disturbance, and from physical consideration, it comes that VF C [0, Vd [. 5.2.6. Equilibrium: The equilibrium vector is: x = = x1 , Vd , x2 , x3 , x4 , Vd , x5 ,
Vd RL ,

x6 B , i

T Vd RL T

2 Vd RL VF C ,

(20)

Where Vd is the desired DC Bus voltage. An implicit purpose of the proposed structure Fig. 5 is to use the battery as a power source supplying the transient peak power. The equilibrium value of the battery current will be given by the supervision task by the FL control and is discussed in section 8, hence x5 = B . i The following equations obtained at the equilibrium will be used in the closed loop dynamic and the stability proof. F C x2 x4 F C x1 + VF C x4 x6 + B x5 x1 x3 eB rB x5 B x3 RL x6 + x3 = = = = = = 0 0 0 0 0 0
T

(21)

= or u= uF C ,

F C , uB
T

B =

VF C Vd ,

eB Vd

(22) (23)

VF C Vd ,

eB Vd

The natural energy function of the system is H= 1 T x Qx 2 (24)

where Q = diag{Cs ; LF C ; CDC ; LDC ; LB ; LL } is a diagonal matrix. 6. Problem Formulation The purpose is the control of the load voltage and consequently the DC Bus voltage by the mean of the control of the two DC-DC converters. The load is unknown and the load resistor is subject of changes. The second aim is to maintain a constant mean energy delivered by the FC, without a signicant power peak, as the transient power is supplied by the battery. A third purpose consists on recovering energy through the charge of the battery. Finally, energy supervision under source limitations should be incorporated in the control task. 10

7. Port-Controlled Hamiltonian representation of the system In the sequel, a closed loop PCH representation is given. The desired closed loop energy function is: Hd = 1 T x Q x 2 (25)

where x = x x is the new state space dening the error between the state x and its equilibrium value x. The state space equation can be written with the new error variables (where (21) has been used): x1 x2 x3 x4 x5 x6 y 1 [F C x2 x4 + F C x2 x4 ] Cs 1 [F C x1 + VF C F C x1 ] LF C 1 [4 x6 + B x5 + x4 x6 + B x5 ] x CDC 1 [1 x3 ] x LDC 1 [eB rB x5 B x3 rB x5 B x3 ] LB 1 [RL x6 + x3 ] LL x3

= = = = = = =

(26)

The representation (26) in function of the gradient of the desired energy (25) is given by (27). x = [J (F C , B ) R] with Hd = Cs x1 , LF C x2 , CDC x3 , Ai (, ) = x LDC x4 , LB x5 , LL x6
T

Hd + Ai (, ) x

(27)

x2 Cs [F C F C ] 1 x [F C F C ] LF C x5 [B B ] CDC 3 x LB

(28)

[B B ] 0

J (F C , B ) = J T (F C , B ) is a skew symmetric matrix dening the interconnection between the state space and R = RT 0 is symmetric positive semi denite matrix dening the Damping of the system.

11

J (F C , B ) R = 0
F C Cs LF C F C Cs LF C

0 0 0

1 Cs LDC

0 0 0 0 0

0 0 0
1 CDC LDC B CDC LB 1 CDC LL

1 Cs LDC

0 0
B CCD LB

0 0
1 CDC LL

1 CDC LDC

0 0 0

0 0

rB L2 B

0 0

(29)

RL L2 L

The following control laws are proposed: uF C = uF C uB = uB r5 x where r is a positive design parameter. Proposition 1. The origin of the closed loop PCH system (27), with the control laws (30) and (23) with the radially unbounded energy function (25), is globally asymptotically stable. Proof. The closed loop dynamic of the PCH system (27) with the laws (30) and (23) with the radially unbounded energy function (25) is: x = [J (F C , B ) R ] Hd + Q
+rV L where R = diag{0; 0; 0; 0; rBL2 d ; R2 } = R L
B L

(30)

(31)

0 and Hd = A Hd

Q=

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0

x r CCD5LB

0 0 0 0 0

0 0 0 0 0 0

(32)

The derivative of the desired energy function (25) along the trajectory of (31) is: T T Hd = Hd x = Hd (R A) Hd (33) According to the theorem stated by author in [27], (33) is non positive if [R A] r L is non negative. The eigenvalues of [R A] are {0; 0; 0; 0; LB ; R2 } and are non 2 LL B negative. Consequently, Hd =
T T Hd x = Hd (R A) Hd 0

(34)

12

The asymptotic stability proof is deduced from the derivative analysis of Hd x and the invariance principle of the LaSalle theorem [29] with Hd () = Hd () = 0 x 2 . 8. Fuzzy logic for the source limitations In order to increase the FC lifespan and to minimize the use of H2 , one can decide to use the battery as the primary (or unique) source, if this latter is fully charged and to solicit the battery in the peak transient. Hence, the battery supplies the transient peak power to the load: If the battery is discharged, it can not provide this power, in this case two possibilities are available: 1. If the Hydrogen tank is full, the FC can be used to recharge the battery. Hence, B should be negative. i 2. If the Hydrogen tank is in a critical level, the battery is not used, neither charged and nor discharged. Hence, B = 0 i If the battery in not discharged. Hence, if the Hydrogen tank decreases below 50%, the battery should start helping the FC by supplying a small quantity of energy (small positive value for B ). If the Hydrogen tank i starts dangerously to decline, the battery should increases its percentage of supplied energy (big positive value for B ). i B values are in the admissible range depending on the used battery characi teristics. The available H2 in the fuel tank is measured by the mean of an hydrogen tank sensor. As mentioned before, the control relies on linguistic rules, coming from a human expertise, thus fuzzy logic control is here chosen. Fig. 7 illustrates the chosen membership functions for the two inputs, the available H2 and the SOC of the battery. Z, Half and Full stand for Zero (the source is completely empty), 50% available and the source is full, respectively. The amount of the available energy is given in per unit between 0 (empty) and 1 (full). Three membership functions are enough to describe the state of the inputs (the two linguistic variables are the remaining H2 and SOC). For each input, three linguistic values are considered (Zero, Half and Full), the FL allows to identify the percentage of belonging to each set. Hence, for each functioning point, inputs range in the set [0, 1]. Fig. 8 shows the chosen membership functions for the battery current reference where NB, N, Z, P, PB stand for Negative Big, Negative, Zero, Positive and Positive Big, respectively. The chosen battery has a maximum current of 5A.
2 it

x can be easily founded that Hd () = 0 (3 = x5 = x6 = 0) x

using(21)

x4 = x1 = x2 = 0

13

H2 1 Z Half Full 1

SOC Z Half Full

0 0 1

0 0 1

Figure 7: Membership functions for the H2 and SOC FL inputs.

Consequently, the battery reference current is in the set [5, 5]A where -5A corresponds to a deep discharge of the battery and if Hydrogen is available a rapid recharge of the battery is needed (negative big value of the battery current). If B is in the range N, the battery is recharged with a small negative current. If i B is Zero, the battery is not used. When B is positive, the battery is providing i i power to the load to help the FC. If B is Positive Big, this corresponds to a i fast discharge of the battery to help the FC in a transient peak power or when the Hydrogen tank dangerously decreasing to zero. Authors have chosen ve membership functions to obtain the following scenario: 1. 2. 3. 4. 5. NB: Fast battery recharge; N: Slow battery recharge; Z: The battery is not used (neither discharged, nor recharged); P: Slow battery discharge; PB: Fast battery discharge.

iB 1 NB N Z P PB

0 -5

Figure 8: Membership functions for the B output of the FL. i

Table 1 presents the dierent rules to obtain the battery current reference according to the values of H2 and SOC. One can noticed that when H2 and SOC are full, this case corresponds to the startup of the vehicle when the available hydrogen and the battery are charged. In this case, B is positive allowing the i battery to help FC with the startup phase. 14

Table 1: Rules for the Fuzzy Logic supervision

PP H2

PP SOC Z PP P P Z Z Half N Full NB

Half P Z N

Full PB P P

9. Simulations Two dierent situations are tested and simulated in order to exhibit the benet of the proposed supervision under dierent source limitations. 1. situation 1: The battery is half charged (SOC = 50%) and the hydrogen tank is full 2. situation 2: The battery is completely charged and the hydrogen tank is half lled (H2 = 50%) 9.1. Situation 1:
60 24

V & V (V)

40 22

DC

VFC(V)
0 1 2 3 t(s) 4 5 6

20 0 20

20 18 16

6 15 4

3 t(s)

i (A)

0 2

iFC(A)
0 1 2 3 t(s) 4 5 6

10

3 t(s)

Figure 9: (a) DC Bus voltage and its reference. (b) Load current.

Figure 10: (a) FC voltage. (b) FC current.

11.7
FC

0.8 0.6 0.4

V (V)

11.5 11.4

11.6

3 t(s)

0.75 11.3

10

3 t(s)

0.7 0.65

3 t(s)

i (A)

15

RL()
10 0

3 t(s)

3 t(s)

Figure 11: (a) Battery voltage. (b) Battery cur- Figure 12: (a) FC Boost control. (b) Battery DCDC converter control. (c) Load resistance change. rent.

15

250 P 200
B

Available H (%)

1 0.8 0.6 0 0.8 0.6 0.4 1 2 3 t(s) 4 5 6

PFC PL

150

Power (W)

100

Battery discharge
50

SOC (%)

& i (A)

3 t(s)

10 0 10

50

Battery recharge

100

Bref

3 t(s)

3 t(s)

Figure 13: FC power, battery power and load Figure 14: H2 level, SOC of the battery, battery power. current and its reference.

In Fig. 9, the DC Bus voltage Vd is tracking its variable reference VDC . The load current iL is also subject to load variation. Fig. 10 presents the FC voltage and current VF C &iF C . A smooth behaviour of the FC voltage and current is observed. As in situation 1 the battery is not completely charged, the FC current is bigger in order to recharge the battery. Hence, it can be seen in Fig. 11 the battery current is negative at the starting corresponding to the recharging phase of the battery. Once the battery recharged and the hydrogen fuel tank decreases, the battery starts to supply the load. Fig. 12 presents the FC Boost control signal uF C , the battery converter control signal uB and the changes in the load resistance RL . Fig. 13 illustrates the benet of this supervision technique where it can be seen that when the battery is not charged, the FC (under the condition that the hydrogen fuel tank is full) recharges the battery in addition to feed the load. When the battery is recharged and the hydrogen level decreases, the battery is helping to supply the load. Fig. 14 illustrates the available amounts of H2 (top) and SOC (middle) and the calculated B by the FL and the real iB (bottom). It can be seen that i according to the remaining H2 and the remaining battery SOC, the membership functions calculate the optimal battery reference B in order to manage the i source limitations. The PBC has then the charge to track the battery reference, that is done quite well (no overshoot and no steady state error are observed). 9.2. Situation 2: Fig. 15 shows the response of the system to the DC Bus voltage reference changes Vd &VDC and the load current iL . Good tracking performances are obtained. Fig. 16 presents the FC voltage and current VF C &iF C . A smooth behaviour of the FC voltage and current is observed regarding the changes in the DC Bus voltage reference. The battery supplies the transient power. In this situation the hydrogen fuel tank is not full, then the FL supervision decides to use the battery (which was initially charged). Consequently, the FC

16

60 24

V & V (V)

40 22

DC

VFC(V)
0 1 2 3 t(s) 4 5 6

20 0 20

20 18 16

6 15 4

3 t(s)

i (A)

0 2

iFC(A)
0 1 2 3 t(s) 4 5 6

10

3 t(s)

Figure 15: (a) DC Bus voltage and its reference. (b) Load current.
12.5

Figure 16: (a) FC voltage. (b) FC current.

0.8
FC

V (V)

12

0.6 0.4

11.5 0.75 11

3 t(s)

10

3 t(s)

0.7 0.65

3 t(s)

i (A)

15

RL()
10 0

3 t(s)

3 t(s)

Figure 17: (a) Battery voltage. (b) Battery cur- Figure 18: (a) FC Boost control. (b) Battery DCDC converter control. (c) Load resistance change. rent.

current is low (comparing to the situation 1) and the battery current is always positive (discharging) as it can be seen in Fig. 17. Fig. 18 presents the FC Boost controller uF C , the battery bidirectional converter controller uB and the changes in the load resistance RL . uF C and uB are in the set [0, 1]. It can be seen from Fig. 19 that the battery is supplying a large amount of energy to compensate the lack of hydrogen. Fig. 20 illustrates the available amounts of H2 and SOC and in the bottom the calculated B by the FL and i the real iB . It can be seen from this last Fig. that according to the remaining H2 fuel level and the remaining battery SOC, the dened membership functions calculate the optimal battery reference B in order to manage the source limi itations. The PBC has then the charge to track the battery reference, that is done quite well (no overshoot and no steady state error are observed). 10. Experimental results Fig. 21, 22 and 23 are an illustration of the experimental test benches built by authors by using basic commercial elements. The used components are: Ballard Nexa 1.2 kW FC (Fig. 22) connected to the DC bus via the DC-DC Boost (designed and built by authors, see 23), battery connected to the DC Bus

17

200 P
B

Available H (%)

PFC 150 PL

0.45 0.4 0.35 0 1 2 3 t(s) 4 5 6

Power (W)

SOC (%)

100

0.9 0.8 0.7 0 1 2 3 t(s) 4 5 6

50

& i (A) i

10 0 10

50

Bref

3 t(s)

3 t(s)

Figure 19: FC power, battery power and load Figure 20: H2 level, SOC of the battery, battery power. current and its reference.

through a current bidirectional DC-DC buck-boost converter (Fig. 21), the load is constituted by a variable resistor in parallel with the DC bus capacitance.

Figure 21: Battery, load and DC-DC buck-boost setup.

Figure 22: Ballard Nexa FC.

Figure 23: Global experimental setup.

The FC boost and the battery DC-DC current bidirectional converter are controlled by means of a real time dSpace DS1104 board. The considered battery delivers a voltage of 12V. The desired DC bus reference Vd is online changed using Matlab-ControlDesk, the operating scenario consists on the control of the FC boost and the battery DC-DC buck-boost to meet, at the DC bus, the desired voltage Vd . In order to exhibit, during the few seconds of the experimentation duration, the ability of dealing with the tank (H2 and electricity) limitation, scroll input buttons are implemented in the ControlDesk to emulate a variation of the hydrogen level tank and the battery SOC3 . Fig. 24 presents the experimental results. The experimental test consists rst in the regulation problem of the DC bus voltage VDC toward its reference Vd and the battery current iB toward its reference iBref which is dened using the FL technique. A change in the DC bus
3 It is evident that during the short considered experimentation duration, the level of the hydrogen fuel tank or the battery SOC can not be aected.

18

voltage reference is applied directly from the ControlDesk and consists on a step change from 42V, to 41V and then to 43V as it can be seen in the Fig. 24(a). One can see that the tracking of these two variables towards their references are very good until 19 seconds. At approximatively 19 seconds, the hydrogen level starts decreasing, this variation of the hydrogen fuel level is emulated by adjusting the scroll input button. When the hydrogen level is becoming critical, around 20% at 25 seconds, the FL rules compute the new battery reference which is approximatively multiplied by two as it can be seen from Fig. 24 (d) and (j). The battery current tracking is very good without overshoot and no steady state error. Thus, the battery provides a large amount of energy to overcome the lack of hydrogen and the FC supplied energy decreased as it shown in the power ows Fig. 25. At approximatively 32 seconds, the battery SOC scroll input button is decreased (Fig. 24(h)) emulating a decreasing in the state of charge of the battery due to its contribution in feeding the load during the previous seconds. Fig. 24(j) shows that FL rules computes the new battery current reference. If this battery SOC decreasing happened when the hydrogen fuel tank is full, the new battery current reference can be negative allowing the recharge of the battery using the FC. But as the hydrogen fuel is also around 40%, the new battery current reference slightly decreased managing then the overall sources (hydrogen fuel and battery SOC). One can see from Fig. 25 that the amount of the supplied battery energy decreases and the amount of the supplied FC energy increased with the same ratio in order that the FC plus the battery energies match at each time the requested load energy. One may notice also that at approximatively 8 seconds, the transient power peak due to the step change in the DC bus reference is provided by the battery allowing a smooth behaviour on the FC energy shape. 11. Conclusion A dynamic modelling of a hybrid source system composed of a FC and a battery sources is presented.PCH structure of the overall system is given. The problem of the DC bus voltage control is solved using simple linear controllers based on the IDA-PBC approach. These controllers need the measurement of two variables only (the FC voltage VF C and the battery current iB ). The proposed structure allows supplying and absorbing the power peaks by using battery which also allows recovering energy. Global Stability proof is given insuring the convergence of the state space vectors towards their desired equilibrium values. The other contribution of this paper consists on the supervision of the energy ow under source limitations using the fuzzy logic technique. This supervision allows to calculate in real time the reference of the battery current allowing to decide from which source the load will be supplied (or with which usage ratio of each of the sources). The energy supervision needs the measurement of the hydrogen fuel tank level and the state of charge of the battery is calculated upon the measurement of the battery current and temperature. The FL energy supervision aims to calculate the optimal equilibrium value of the battery current (according to the chosen rules). Hence, the incorporation of the 19

44 (a) 42 40 12.2 12.1

Vdc(V) V (V)
d

1.5 iL(V) (b) 1 0.5 3 (d) 2 1 40 0 0.8 (f) 0.6 0.4 1 Remaining H2 (h) 0.9 0.8 4 (j) uB uFC 2 0 0 10 20 30 iFC(V) 40 iB(V)

10

20

30

40

10

20

30

40

(c)

V (V)
B

12 39.2 (e) 39

10

20

30

VFC(V) 38.8 1 (g) 0.5 0 1 (i) 0 1 0 10 20 30 40 0 10 20 30 40

SOC 0 10 20 iB(A) iBref(A) 30 40

10

20

30

40

10

20 t(s)

30

40

10

20 t(s)

30

40

Figure 24: Experimental results. (a) DC Bus voltage and its reference, (b) load current, (c) battery voltage, (d) battery current, (e) FC voltage, (f) FC current, (g) remaining hydrogen level in percentage, (h) battery SOC in percentage, (i) battery DC-DC converter and the FC Boost converter controls, (j) battery current and it reference

FL technique didnt aect the stability of the whole system, since the battery current reference is always a bounded value.An encouraging simulation results has been obtained exhibiting also the robustness of the proposed controllers towards load resistor variations. A low power experimental setup has been built by authors.The designed system is online controlled using Matlab dSpace and the hydrogen fuel tank level and the battery SOC have been online emulated using ControlDesk. The realized test scenario validates the control objectives.It is experimentally shown that the proposed controllers react to the sources limitation and manage the energy ows between the two sources in order to match at each time the requested load power demand. In addition, the same simple controllers allows the recharge of the battery if the SOC is critical and the hydrogen fuel level is full. These encouraging experimental results are validating the overall theoretical contribution.

20

60 50 FC Power Battery Power Load Power

Powers(W)

40 30 20 10 0

10

20

30

40

Figure 25: Experimental battery, FC and load powers.

t(s)

References
[1] Becherif M., Passivity-based control of hybrid sources: Fuel cell and Battery, 11th IFAC Symposium on Control in Transportation Systems (CTS06), Netherlands, 2006. [2] Hilairet M., Bethoux O., Azib T., and Talj R., Interconnection and damping assignment passivity-based control of a fuel cell system, IEEE International Symposium on Industrial Electronics (ISIE), pp. 219 - 224, Italy, 2010. [3] Kolavennu P., Telotte J.C., and Palanki S., Analysis of battery backup and switching controller for a fuel-cell powered automobile, International Journal of Hydrogen Energy, vol. 34(1), pp. 380-387, January 2009. [4] Kim Y.B., and Kang S.J., Time delay control for fuel cells with bidirectional DC/DC converter and battery, International Journal of Hydrogen Energy, vol. 35(16), pp. 8792-8803, August 2010. o [5] Andreasen S.J., Ashworth L., Remn I.N.M, and Kr S.K., Directly connected series coupled HTPEM fuel cell stacks to a Li-ion battery DC bus for a fuel cell electrical vehicle, International Journal of Hydrogen Energy, vol. 33(23), pp. 7137-7145, December 2008. [6] Gilbert E.G., Vehicle cruise: Improved fuel economy by periodic control, Automatica, pp. 159-166, vol. 12(2), 1976. [7] Halpin, S.M.; Ashcraft, S.R.: Design considerations for single-phase uninterruptible power supply using double-layer capacitors as the energy storage element, IEEE-IAS, San Diego, vol.4, pp.23962403, 1996. [8] Kim M.J., and Peng H., Power management and design optimization of fuel cell/battery hybrid vehicles, Journal of Power Sources, pp. 819-832, vol. 165(2), 2007.

21

[9] Becherif M., Ayad M.Y., Henni A., Aboubou A., Wack M., and Laghrouche S., Passivity-Based Control applied to DC Hybrid Power Source using Fuel Cell and Supercapacitors, Energy Conversion and Management, vol. 51(7), pp 1468-1475, July 2010. [10] Ayad M.Y., Becherif M., Henni A., and Wack M., Vehicle Hybridization with Fuel Cell, Supercapacitors and batteries by Sliding Mode control, Elsevier REJ, Renewable Energy Journal, 2010. [11] Jeong K.S., and Soo Oh S., Fuel economy and life-cycle cost analysis of a fuel cell hybrid vehicle, Journal of Power Sources, pp. 58-65, vol. 105(1), 2002. [12] Arce A., del Real A.J., and Bordons C., MPC for battery/fuel cell hybrid vehicles including fuel cell dynamics and battery performance improvement, Journal of Process Control, vol. 19(8), pp. 1289-1304, September 2009. [13] Fernandez L.M., Garcia P., Garcia C.A., and Jurado F., Hybrid electric system based on fuel cell and battery and integrating a single dc/dc converter for a tramway, Energy Conversion and Management, vol. 52(5), pp. 2183-2192, May 2011. [14] Van Mierlo J., Maggetto G., and Lataire Ph., Which energy source for road transport in the future? A comparison of battery, hybrid and fuel cell vehicles, Energy Conversion and Management, vol. 47(17), pp. 2748-2760, October 2006. [15] Hankache W., Caux S., Hissel D., and Fadel M., Genetic Algorithm Fuzzy Logic Energy Management Strategy for Fuel Cell Hybrid Vehicle, IFAC09 Power Plant & Power System, Tampere, Finland, 2009. [16] Tekin M., Hissel D., Pra M.C. and Kaumann J.M., Energy management strate egy for embedded fuel cell system using fuzzy logic, IEEE Transactions on Industrial Electronics, vol. 54, n 1, pp. 595-603, 2007. [17] Larminie J., and Dicks A., Fuel Cell Systems Explained, 2nd edition, Wiley, 2000. [18] Krein P.T., Elements of Power Electronics, Oxford University Press, USA, 2001. [19] Garcia-Gabin W., Dorado D., and Bordons C., Real-time implementation of a sliding mode controller for air supply on a PEM fuel cell, Journal of Process Control, vol. 20(3), pp. 325-336, March 2010. [20] Zenith Z., and Skogestad S., Control of fuel cell power output, Journal of Process Control, vol. 17(4), pp. 333-347, April 2007. [21] Suppes G.J., Plug-in hybrid with fuel cell battery charger, International Journal of Hydrogen Energy, vol. 30(2), pp. 113-121, February 2005. [22] Zenith F., and Krewer U., Modelling, dynamics and control of a portable DMFC system, Journal of Process Control, vol. 20(5), pp. 630-642, June 2010. [23] Friede W., Modlisation et caractrisation dune pile ` combustible du type PEM, e e a PhD Thesis, GREEN-INPL Nancy, August 2003.

22

[24] Boulon L., Hissel D., and Pra M.C., Multi-Physics Modelling and Energy Mane agement of a Battery Supercondensator Electric Vehicle Taking Into Account the Operating Temperature Conditions. ELECTRIMACS, Quebec, Canada, 2008. [25] Van der Schaft A.J., and Maschke B., On the hamiltonian formulation of nonholonomic mechanical systems, Reports on Mathematical Physics, pp. 225233, vol. 34(2), 1994. [26] Ortega R., Van der Schaft A.J., Maschke B., and Escobar G., Interconnection and damping assignment passivitybased control of portcontrolled hamiltonian systems, Automatica, pp. 585596, vol. 38(4), 2002. [27] Becherif M., and Mendes E., Stability and robustness of Disturbed-Port Controlled Hamiltonian system with Dissipation, 16th IFAC World Congress, Prague, 2005. [28] Mangold M., Bck A., and Hanke-Rauschenbach R., Passivity based control of u a distributed PEM fuel cell model, Journal of Process Control, vol. 20(3), pp. 292-313, March 2010. [29] Lasalle J.P., Stability theory for ordinary dierential equations, Journal of Differential Equations, vol. 4, no. 1, pp. 57-65, 1968.

23