EP4352808A1

EP4352808A1 - Method for detecting a hydrogen leak in a fuel cell system and fuel cell system for implementing such a method

Info

Publication number: EP4352808A1
Application number: EP22733400.0A
Authority: EP
Inventors: Gino Paganelli; Lionel JEANRICHARD-DIT-BRESSEL
Original assignee: Symbio SAS
Current assignee: Symbio SAS
Priority date: 2021-06-11
Filing date: 2022-06-10
Publication date: 2024-04-17
Also published as: FR3123985A1; FR3123985B1; WO2022258839A1

Abstract

This method for detecting a hydrogen leak applies to a fuel cell system (10) comprising a fuel cell (12); a hydrogen supply system (30) comprising a reservoir (32) and the supply circuit (34) connecting the reservoir to the anode compartment (16) of the fuel cell and comprising an ejector (36) of venturi type; a recirculation circuit (60) for recirculating unconsumed hydrogen between the anode compartment of the fuel cell and the venturi-type ejector (36), the recirculation being sustained by the venturi-effect ejector. The method comprises steps involving calculating the total flow rate of hydrogen consumed; calculating the flowrate of hydrogen admitted to the injector; determining the leak rate as the difference between the flowrate of hydrogen admitted and the total flowrate of hydrogen consumed; and detecting a potential leak of hydrogen by comparing the leak rate against at least a threshold value, such that the method detects all of the hydrogen leaks that occur in the system downstream of the ejector.

Description

TITLE: Hydrogen leak detection method in a fuel cell system and fuel cell system for the implementation of such a method

The present invention relates to a method for detecting hydrogen leaks in a fuel cell system and a fuel cell system making it possible to implement such a method.

A fuel cell system generally comprises a fuel cell allowing the production of electrical energy by an electrochemical reaction, as well as a hydrogen supply system and a purge system to allow the operation of this fuel cell.

A fuel cell generally comprises the series association of unitary elements which each consist essentially of an anode and a cathode separated by a polymer membrane allowing the passage of protons from the anode to the cathode. The anode supplied with hydrogen is the site of an oxidation half-reaction H ₂ (g) => 2 H ⁺ (aq) + 2 e ^~ . At the same time, the cathode supplied with oxidizer, for example pure oxygen or oxygen contained in air, is the site of a reduction half-reaction: 0 ₂ (g) + 4 H ⁺ (aq) + 4 e => 2 H ₂ 0 (I).

It is known from WO 2018/115630 A1 to add, to a fuel cell system, a system for recirculating hydrogen not consumed by the fuel cell. Such a recirculation system improves the performance of the fuel cell. It is also known from this document to use a Venturi-type ejector to allow the recirculation of unconsumed hydrogen from the outlet of the anode of the fuel cell to the inlet of the anode.

However, such a fuel cell system has the disadvantage of not making it possible to estimate the hydrogen leaks occurring in the fuel cell system. However, it is important to detect such leaks to ensure the safe operation of the fuel cell system and to maintain optimal performance, without avoidable loss of hydrogen.

It is known to measure the hydrogen flow supplied to a fuel cell using a flow meter installed on the hydrogen supply line of the fuel cell. However, the use of a flowmeter is restrictive, since a flowmeter is generally bulky and has a pressure drop that is harmful to the efficiency of the fuel cell. Furthermore, an accurate flowmeter is expensive. It is known from KR 10-1393581 a method for determining hydrogen leaks in a fuel cell system, during operation. This method is based on the difference between the hydrogen consumption calculated according to the current produced by the fuel cell and the hydrogen consumption estimated according to the pressure drop through a hydrogen supply valve.

However, such a method is not very precise, in particular for low intensity hydrogen leaks. In particular, the fact of estimating the quantity of hydrogen supplied to the fuel cell system according to the operating parameters of a supply valve requires, on the one hand, approximations and, on the other hand, equipment. specific ones.

It is these drawbacks that the invention more particularly intends to remedy by proposing an improved method for detecting hydrogen leaks in a fuel cell system.

To this end, the invention relates to a method for detecting hydrogen leaks in a fuel cell system, the fuel cell system comprising at least: a stack of electrochemical cells forming a fuel cell comprising an anode compartment and a cathode compartment separated by a proton exchange polymer membrane; a hydrogen supply system comprising a hydrogen tank and a supply circuit connecting the hydrogen tank to the inlet of the anode compartment of the fuel cell, the supply circuit comprising a Venturi-type ejector; a circuit for recirculating hydrogen not consumed by the fuel cell between the outlet of the anode compartment of the fuel cell and the Venturi-type ejector of the supply circuit, the recirculation of the hydrogen not consumed being driven by the Venturi effect ejector; and

- a purge system comprising a purge and drainage valve for the anode compartment.

According to the invention, the hydrogen leak detection method comprises at least the following steps: a) calculating the total flow rate of hydrogen consumed by the fuel cell system; b) calculate the hydrogen flow admitted by the hydrogen supply system at the inlet of the venturi ejector; c) determining the leak rate by calculating the difference between the flow rate of hydrogen admitted and the total flow rate of hydrogen consumed; and d) detecting a possible hydrogen leak in the fuel cell system by comparing the leak rate with at least one threshold value, such that the method detects all of the hydrogen leaks occurring in the system fuel cell stack downstream of the Venturi-type ejector.

Thanks to the invention, the method for detecting a hydrogen leak in a fuel cell system makes it possible to detect a hydrogen leak with great precision and independently of the location of this leak in the fuel cell system, including in the event of a low-intensity leak.

According to advantageous, but not compulsory, aspects of the invention, this method for detecting hydrogen leaks in a fuel cell system incorporates one or more of the following characteristics, taken in isolation or in any technically admissible combination:

- During step a), the calculation of the total flow of hydrogen consumed Q _H 2.out by the fuel cell system is carried out from the following sum: Q _H 2.out = QH2.out _sto + QH2.out _xo + Ç _tf2.o ui _puri in which Q _H 2.out _sto is the base flow rate of hydrogen consumed by the fuel cell by electrochemical reaction; Q _H2.0 ut pure g is the flow of hydrogen lost in the purges of the anode compartment of the fuel cell; and QH2.OUÎ _XO is the flow rate of hydrogen passing through the polymeric proton exchange membrane from the anode compartment to the cathode compartment.

- During step a), the flow of hydrogen lost Q _H 2.out purg in the purges of the anode compartment of the fuel cell is calculated according to the following equation: QH2.out _purg = x V _anode x P _anode.in where MW _h2 is the molar mass of dihydrogen, R is the universal ideal gas constant, T is the temperature within the anode compartment, V _anode is the volume of the anode compartment and Panode.in is the pressure gradient measured at the inlet of the anode compartment during a purge.

- The pressure gradient P _anode.in is obtained by a constant intake purge method consisting in postponing the restoration of the pressure lost in the anode compartment of the fuel cell during a purge by postponing the opening of a fuel system hydrogen supply valve.

- During step a), the flow rate of hydrogen consumed by permeation Q _H 2.out _xo through the polymer proton exchange membrane from the anode compartment to the cathode compartment is calculated according to the following equation:

QH2. ₀ ut _xo = ^MW ₂ ^ ^{h N} x Jxo x S _membrane in which MW _h2 is the molar mass of the dihydrogen, N is the number of electrochemical cells in the fuel cell, F is Faraday's constant, J _Xo is the cross over current density and S _membrane is the membrane area of an electrochemical cell.

- During step b), the flow rate of hydrogen admitted Q _H2 .i _n by the hydrogen supply system in the intake manifold of the Venturi type ejector is calculated according to whether the Flow occurring within the Venturi type ejector is a subsonic flow regime or a sonic flow regime.

- When the flow regime is subsonic, the hydrogen flow rate admitted Q _H 2.in _sub P ^ar I ^e hydrogen supply system in the intake manifold of the Venturi type ejector is calculated according to the following equation: Q _H 2.m _sub = 5 x P ₁ x A _c x in the ^ac l ^ue l ^e Pi ^is ' ^a pressure of the hydrogen admitted in the Venturi-type ejector inlet manifold, Ti is the temperature of the hydrogen admitted to the Venturi-type ejector inlet, P _å is the hydrogen pressure at the outlet of the Venturi-type ejector Venturi, d is the efficiency of the sonic neck of the Venturi type ejector, A _c is the smallest section of the sonic neck of the ejector, MW _h2 is the molar mass of dihydrogen, y is the adiabatic coefficient of dihydrogen and R is the universal ideal gas constant. In addition, when the flow regime is sonic, the hydrogen flow rate Q _H 2.in _S0Jl admitted by the hydrogen supply system into the intake manifold of the Venturi-type ejector is calculated with the following equation: Q _H 2.in _son dx P _x x A _c x r- ^XY x ( ) ^2C(U-1) in which Pi is the pressure of the hydrogen admitted into the inlet pipe of the ejector of Venturi type, Ti is the temperature of the hydrogen admitted into the intake manifold of the Venturi type ejector, d is the sonic throat efficiency of the Venturi type ejector, A _c is the smallest cross section of the sonic neck of the ejector, MW _h2 is the molar mass of dihydrogen, y is the adiabatic coefficient of dihydrogen and R is the universal ideal gas constant.

- The leak rate determined in step c) is filtered before performing the comparison with at least one threshold value.

- The leak rate determined in step c) is filtered using two different filters and the results obtained by these two filters are compared with two different detection thresholds: a first leak detection threshold compared to the leak rate filtered with a first-order low-pass filter with a time constant equal to a first value; and a second leak detection threshold compared to the leak flow rate filtered with a first order low pass filter of time constant equal to a second value. Moreover, the first leak detection threshold has a value higher than that of the second leak detection threshold and the first time constant value is lower than the second time constant value.

- The hydrogen leak detection method is carried out cyclically, in real time, and, preferably, steps c) and d) of the hydrogen leak detection method are carried out by a control computer of the fuel cell system at each ECU sampling period.

According to another aspect, the invention also relates to a fuel cell system for implementing the hydrogen leak detection method described above, this fuel cell system comprising at least: a stack of electrochemical cells forming a fuel cell comprising an anode compartment and a cathode compartment separated by a proton exchange polymer membrane; a hydrogen supply system comprising a hydrogen tank and a supply circuit connecting the hydrogen tank to the inlet of the anode compartment of the fuel cell, the supply circuit comprising a Venturi-type ejector; a circuit for recirculating hydrogen not consumed by the fuel cell between the outlet of the anode compartment of the fuel cell and the Venturi-type ejector of the supply circuit, the recirculation of the hydrogen not consumed being driven by the Venturi effect ejector;

- a purge system comprising a purge and drainage valve for the anode compartment; and a computer to implement steps a) to d) of the hydrogen leak detection method.

This fuel cell system induces the same advantages as those mentioned above regarding the leak detection method of the invention.

Advantageously, this fuel cell system further comprises a pressure sensor and a temperature sensor arranged upstream of the Venturi-type ejector and a pressure sensor arranged downstream of the Venturi-type ejector.

The invention will be better understood and other advantages thereof will appear more clearly in the light of the following description of an embodiment of a method for detecting hydrogen leaks in a fuel cell system. fuel and an embodiment of a fuel cell system, in accordance with the invention, given solely by way of example and made with reference to the appended drawings in which: [Fig. 1] Figure 1 is a fluid diagram of a fuel cell system according to the invention;

[Fig. 2] Figure 2 is a detail view of part of the fuel cell system of Figure 1;

[Fig. 3] FIG. 3 is a graph of the evolution over time of the pressure measured by a sensor of the fuel cell system of FIG. 1;

[Fig. 4] FIG. 4 is a diagram illustrating the steps of a hydrogen leak detection method in a fuel cell system, this method being in accordance with the invention; and

[Fig. 5] Figure 5 is an example of results obtained by the leak detection method shown in Figure 4 applied to a fuel cell system without hydrogen leakage.

A fuel cell system 10 is represented in FIGS. 1 to 2. This fuel cell system is, for example, intended to be integrated into a vehicle with an electric motor in order to produce electrical energy allowing the operation of the motor.

This fuel cell system 10 includes a fuel cell 12.

The fuel cell 12 comprises a stack of electrochemical cells 14, only one of which is represented in FIG. 1, for the sake of simplification.

In practice, the fuel cell 12 comprises an integer number N of electrochemical cells 14, N being preferably between 1 and several hundred, more preferably between 64 and 416.

Each electrochemical cell 14 includes an anode compartment 16, which forms the anode of the electrochemical cell, and a cathode compartment 18, which forms the cathode of the electrochemical cell. The anode compartment 16 and the cathode compartment 18 are separated by a proton exchange polymer membrane 20 .

During the operation of the fuel cell 12, the anode compartment 16 is supplied with combustible gas, generally pure dihydrogen, more commonly called hydrogen for the sake of simplification, and the cathode compartment 18 is supplied with oxidizing gas, generally dioxygen, plus commonly called oxygen for the sake of simplification, either pure or included in a mixture of gases, for example air.

In the remainder of the description, the terms “hydrogen” and “dihydrogen” are used interchangeably to designate dihydrogen.

Thus, the fuel cell 12 comprises as many anode compartments 16, cathode compartments 18 and membranes 20 as there are electrochemical cells 14, that is to say N. For simplification, all the anode compartments are assimilated to a single anode compartment 16, which forms the anode of the fuel cell 12, all the cathode compartments are assimilated to a single cathode compartment 18, which forms the cathode of the fuel cell 12 and all the membranes 20 are assimilated to a single membrane 20, which separates the anode and the cathode of the fuel cell 12.

An inlet 22 and an outlet 24 are defined at the level of the anode compartment 16.

To supply the anode compartment 16 with hydrogen, the fuel cell system 10 comprises a hydrogen supply system 30.

The hydrogen supply system 30 comprises a hydrogen tank 32 and a supply circuit 34 connecting the hydrogen tank 32 to the inlet 22 of the anode compartment 16. Thus, the inlet 22 is configured to let in hydrogen in the anode compartment.

An ejector 36, of the Venturi type, as well as a supply valve 38, arranged between the ejector 36 and the hydrogen tank 32, are installed on the supply circuit 34.

Preferably, one or more shut-off valves, not shown, are installed between the hydrogen tank 32 and the supply valve 38.

The supply valve 38 makes it possible to control the flow of hydrogen supplied by the hydrogen tank 32 to the fuel cell 12. Thus, the supply valve 38 is for example a proportional valve, i.e. say that it delivers a flow of hydrogen proportional to its opening.

To purge and to drain the anode compartment 16, the fuel cell system 10 includes a purge system 50. The purge system 50 includes a purge circuit 52 which connects the outlet 24 of the anode compartment 16 with the exterior of the system. fuel cell 10, for example the atmosphere, as well as a purge and drain valve 54 installed on the purge circuit 52 and allowing or prohibiting the purge depending on whether it is open or closed. Thus, the purge and drain valve 54 allows a so-called “discontinuous” purge of the anode compartment 16. Thus, the outlet 24 is configured to let gases and/or liquids out of the anode compartment 16.

In practice, the purge and drain valve 54 is, for example, a so-called “all or nothing” solenoid valve.

The purge system 50 makes it possible to purge the anode compartment 16 of liquid water and non-combustible gases, such as for example nitrogen or water vapour, which may accumulate there.

The fuel cell system 10 also includes a recirculation system 60. In practice, the recirculation system 60 is a circuit connecting the outlet 24 of the anode compartment 16 to the Venturi type ejector 36 of the supply circuit 34 which makes it possible to recirculate hydrogen.

In a manner known per se, the recirculation system 60 comprises hydrogen separation means, not shown, which make it possible to separate the hydrogen not consumed by the fuel cell 12 from the liquid water present at the outlet. 24 of the anode compartment 16, thus making it possible not to recirculate liquid water.

This recirculation of hydrogen is advantageous because it improves the performance of the fuel cell without increasing the consumption of hydrogen. In particular, this recirculation makes it possible to ensure a sufficient flow rate of hydrogen within the anode compartment 16 of the fuel cell 12 to avoid any accumulation of liquid water in the anode compartment and thus to avoid local shortages of hydrogen, allowing by to ensure optimal performance and durability of the fuel cell.

In addition, hydrogen recirculation is driven by the Venturi 36 type ejector, by Venturi effect.

As best seen in Figure 2, the ejector 36 comprises a hydrogen inlet pipe 62, which is in practice connected to the part of the supply circuit 34 located upstream of the ejector.

The ejector 36 also includes a sonic neck 64, into which is channeled all the hydrogen admitted by the hydrogen intake pipe 62. This sonic neck has the effect of increasing the speed of the hydrogen and reduce the pressure of the hydrogen, until a depression is created downstream of the sonic throat 64.

The ejector 36 also includes openings 66, arranged downstream of the sonic neck 64, which are connected to the downstream end of the recirculation system 60. Thus, the recirculation system 60 opens into the Venturi type ejector 36 at the level gills 66.

The vacuum created downstream of the sonic neck 64 by the Venturi effect causes suction through the vents 66, which makes it possible to suck in the hydrogen contained in the recirculation system and thus to circulate the hydrogen in this system.

The ejector 36 further comprises an outlet pipe 68, connected to the part of the supply circuit 34 located downstream of the ejector. The outlet pipe 68 of the ejector is therefore connected to the inlet 22 of the anode compartment 16 of the fuel cell 12, through a section of the supply circuit.

Thus, the hydrogen flow at the inlet of the ejector 36 corresponds to the hydrogen flow supplied by the hydrogen tank 32 and the hydrogen flow at the outlet of the ejector corresponds to the sum of the flow of hydrogen supplied by the hydrogen tank and the flow of hydrogen recirculated within the recirculation system 60.

The fuel cell system 10 also includes a control system 80.

The control system 80 includes, among other things, two actuators 82, 84, a temperature sensor 86, two pressure sensors 88 and 90 and a computer 92.

The actuators 82 and 84 are, for example, solenoids and make it possible to respectively actuate the supply valve 38 and the purge valve 54.

The actuator 82 makes it possible to open the supply valve 38 gradually and thus to adjust the flow of hydrogen supplied to the ejector 36 by the supply circuit 34 in proportion to the opening of the valve. food 38.

The actuator 84 makes it possible to open and close the purge and drain valve 54 and thus trigger and interrupt a purge of the anode compartment 16 of the fuel cell 12.

The temperature sensor 86 and the first pressure sensor 88 are installed on the supply circuit 34 upstream of the Venturi-type ejector 36. In practice, these two sensors are arranged as close as possible to the pipe 62 of admission of hydrogen from the ejector 36, therefore downstream of the supply valve 38, so as to respectively measure the temperature and the pressure of the gas at the inlet of the ejector.

We note “Ti” the temperature of the hydrogen admitted at the inlet of the ejector 36, measured by the temperature sensor 86.

We note “Pi” the pressure of the hydrogen admitted at the inlet of the ejector 36, measured by the pressure sensor 88.

The second pressure sensor 90 is installed on the supply circuit 34 downstream of the ejector 36.

We note “P2” the hydrogen pressure at the outlet of the ejector 36, measured by the pressure sensor 90.

In practice, the computer 92 can be a computer, an integrated circuit card equipped with a microprocessor, an automaton or else software executed on a server.

Preferably, the computer 92 is a control computer for the fuel cell 10.

Computer 92 controls actuators 82 and 84 and retrieves data from temperature sensor 86 and pressure sensors 88 and 90.

Advantageously, the computer 92 executes control software for the fuel cell system 10. In particular, the computer 92 makes it possible to adapt the flow rate of hydrogen supplied to the fuel cell 12 by adjusting the opening of the valve 38 according to operating parameters of the fuel cell, such as for example the hydrogen pressure P2 measured at the inlet of the anode compartment 16.

The computer 92 sends a command signal S38 to open the valve 38 to the actuator 82, the command being proportional to the hydrogen flow to be supplied to the fuel cell 12. This signal S3s to command the opening of the valve 38 is updated in real time during operation of fuel cell system 10.

The computer 92 also makes it possible to control the purges of the anode compartment 16 by controlling the opening of the purge and drain valve 54, by means of a signal S ₅ 4 sent to this valve, for example according to a predefined frequency or in fuel cell operating parameters function 12.

Advantageously, the hydrogen used in the fuel cell system 10 is very dry and very pure, that is to say that the gas stored in the hydrogen tank 32 is composed of at least 99.97% dihydrogen.

The presence of other gases in the hydrogen admitted into the inlet pipe 62 of the ejector 36, such as for example water vapour, is therefore neglected.

Consequently, the properties of pure dihydrogen are used in the calculations presented in the remainder of the description.

Moreover, in the calculations presented in the remainder of the description, the temperature Ti is expressed in kelvin (K) and the pressures Pi and P2 are expressed in pascals (P) or in bar.

We will now describe in more detail and with reference to FIG. 3 an example of a method for purging the anode compartment 16 by the purge system 50, as implemented in the fuel cell system 10.

This example of purge method is not limiting, and other methods of purging the anode compartment 16 can be implemented without departing from the scope of the invention.

Figure 3 shows the evolution of the pressure P^ in the anode compartment 16 during a discontinuous purge of the anode compartment 16 towards the outside of the fuel cell system 10.

The pressure P^ in the anode compartment 16 corresponds to the pressure P ₂ measured by the pressure sensor 90, since the supply circuit 34 downstream of the ejector 36 and the anode compartment 16 are directly connected by the inlet 22 of the anode compartment. Before the anode compartment 16 is purged, the pressure Pi ₆ in the anode compartment 16 is equal to a nominal pressure, denoted “P _nom ”.

In practice, to purge the anode compartment 16, the purge valve 54 is opened on command from the computer 92, for a predefined time interval, between times to and ti.

At the time of the opening of the purge valve 54, that is to say at the instant to, the signal S38 for controlling the opening of the supply valve 38 transmitted by the computer 92 is frozen, from so that, between to and ti, the opening of the supply valve 38 is maintained at a constant level. Thus, between to and ti, the hydrogen flow supplied by the supply circuit 34 to the ejector 36 is constant. This is referred to as “constant intake purge”.

As seen in Figure 3, the opening of the purge valve 54 leads to a decrease in pressure in the anode compartment 16. Indeed, since the opening of the supply valve 38 is fixed during the purge , the gas escaping to the outside through the purge valve 54 is not replaced.

The pressure in the anode compartment 16 then decreases to a minimum, denoted “Pmin”.

During the opening time of the purge valve 54, between to and ti, a pressure gradient is thus observed. From this pressure gradient, it is possible to calculate a purged hydrogen flow in grams per second (g/s) using the following equation:

[Equation 1] in which: - MW _h 2 is the molar mass of dihydrogen, in practice equal to 2.0159 grams per mole (g/mol),

- R is the universal ideal gas constant, in Joules per mole kelvin (Jxmol ¹ xK ¹ ), equal to approximately 8.314 Jxmol ¹ xK ¹ ,

- T is the temperature within the anode compartment 16, expressed in Kelvin, measured by a sensor not shown in the figures,

- ^ _a n _ode is ^the volume of the anode compartment 16, in liters (I), and

Panode.in is the pressure gradient P2 measured at the inlet of the anode compartment 16 by the pressure sensor 90 during a purge, in bars per second (bar/s). The presence of non-combustible gases, such as for example nitrogen or water vapour, in the anode compartment 16, is neglected for this calculation because, in practice, these gases have a negligible influence.

Advantageously, the calculation of the hydrogen purge flow is performed by the computer 92 in real time.

Thanks to this method of purging the anode compartment, it is possible to know in real time the flow rate of pure hydrogen Q _H2.0 ut g lost in the purges of the anode compartment 16.

At the end of the purge, that is to say at the instant ti corresponding to the closing of the purge valve 54, the computer 92 sends a modified opening control signal S ₃₄ to the actuator 82 of the supply valve 38, so as to restore the pressure lost in the anode compartment 16 during the purge.

As visible in FIG. 3, the opening control is thus adapted so as to increase the flow rate of hydrogen admitted from the minimum pressure P _m m until a pressure Pi ₆ equal to the nominal pressure P _n0 m is found in the anode compartment 16.

According to this purge method, the modification of the opening control of the supply valve 38 allowing the increase in the flow rate of hydrogen admitted, and thus the restoration of the nominal pressure P _n0 m, is deferred in time by relative to the opening of the purge and drain valve 54, which makes it possible to generate a pressure gradient and thus to calculate the flow rate of hydrogen purged, thanks to equation 1. This method is then referred to as “delayed purge”.

In practice, the opening time of the purge and drain valve 54, which corresponds to the purge time between time to and time ti, is between 0.1 second and 1 second. For example, the purge time is 0.2 seconds.

Thus, the phase difference between the opening of the purge and drain valve 54 and the admission of hydrogen compensating for the loss of hydrogen linked to the purge is negligible. The decrease in pressure between the nominal pressure P _n0 m and the minimum pressure P _m in therefore has no impact on the operation of the fuel cell 12.

An example of a hydrogen leak detection method in the fuel cell system 10 according to the invention will now be described in more detail and with reference to FIG.

Preferably, this hydrogen leak detection method is implemented automatically by the computer 92.

The hydrogen leak detection method is performed in real time. In addition, this method is carried out cyclically. The steps carried out in each cycle of the hydrogen leak detection method in accordance with the invention are now described. This method comprises a first step 110 of starting.

This method comprises a second step 120 of measurements, carried out after step 110 of starting. During this step, the temperature Ti is measured by the temperature sensor 86, the pressure Pi is measured by the first pressure sensor 88 and the pressure P ₂ is measured by the second pressure sensor 90, then the data from the sensors 86, 88 and 90 are transmitted to computer 92.

The method comprises a third step 130 of calculating the total flow of hydrogen consumed by the fuel cell system 10, subsequent to step 120 of measurements. This total flow of hydrogen consumed is denoted Q _H 2.out-

The total flow of hydrogen Q _H2.0 ut consumed by the fuel cell system 10 is calculated using the following equation:

[Equation 2] where:

Q _H 2.out _sto is a base flow of hydrogen consumed by the electrochemical reaction occurring in the electrochemical cells 14 of the fuel cell 12;

Q _H 2.out _xo is a flow rate of hydrogen consumed by permeation of the proton exchange polymer membrane 20; and is I ^e flow rate of purged hydrogen calculated using equation 1 introduced above, corresponding to the loss of hydrogen in the purges.

The flow rate of purged hydrogen Q _H 2.out puv g is zero when no purge of the fuel cell system 10 is in progress.

The base rate Ç _ii2.o ui _sto of hydrogen consumed by the fuel cell 12 is linked by a stoichiometric relationship to the electric current produced by the fuel cell. Thus this base flow is also called "stoichiometric consumption" and is calculated using the following equation:

[Equation 3] in which:

- MW _{h 2} is the molar mass of dihydrogen, in practice equal to 2.0159 grams per mole (g/mol);

N is the number of electrochemical cells 14 of the fuel cell 12; F is Faraday's constant, in practice equal to 96485 Coulombs per mole

(Coulombs/mol), also expressed in ampere-seconds per mole

(AxSxmol ^-1 ); and

- I is the current generated by the fuel cell 12, in amperes (A).

It is understood from equation 3 that the stoichiometric consumption Q _H 2.out _sto of the fuel cell 12 is directly proportional to the current generated by the fuel cell.

The flow rate _{C ii2.o} ui _xo of hydrogen consumed by permeation through the polymeric proton exchange membrane 20 corresponds to a flow rate of hydrogen transferred from the anode to the cathode during normal operation of the fuel cell 12 , that is to say at a flow rate of hydrogen passing through the membrane 20.

This transfer of hydrogen by permeation is also called the "cross over" mechanism and appears under the effect of the osmotic pressure appearing in the electrochemical cells 14 due to the strong difference in concentration of hydrogen on either side of the membrane 20. The rate of this hydrogen transfer is thus expressed as an equivalent crossover current.

The flow rate Ç _ii2.o ui _xo of hydrogen consumed by permeation is calculated using the following equation:

[Equation 4] in which, in addition to the variables mentioned with reference to equation 3:

- Jx _o is the crossover current density, in amperes per square centimeter

(A/cm ² ); and S m _e m _b r _a n _e is the area of the membrane 20 of an electrochemical cell 14, in square centimeters (cm ² ), which is known and constant.

It is understood from equation 4 that the flow rate Ç _{ii2 out¾o} of hydrogen consumed by permeation is directly proportional to the crossover current, which depends on the properties and characteristics of the polymeric proton exchange membrane 20 used.

Thanks to experimental measurements and theoretical calculations, the evolution of this crossover current density during the lifetime of the fuel cell 12 is known. The increase in crossover current over the life of the battery is relatively small and can reasonably be neglected. Indeed, the crossover current represents a very low value compared to the main current generated by the fuel cell. Typically, the crossover current has a value of about 2mA/cm2, so that the main current generated by the fuel cell has a value of approximately 1A/cm2. This results in a crossover flow, i.e. a flow of hydrogen consumed by permeation, approximately 500 times lower than the base flow, i.e. the flow of hydrogen consumed by the fuel cell 12.

During the third step 130 of calculating the total flow of hydrogen consumed by the fuel cell system 10 is therefore carried out by the computer 92 by applying equation 2 detailed above.

The hydrogen leak detection method comprises a fourth step 140 of calculating the flow rate of hydrogen admitted by the hydrogen supply system at the inlet of the Venturi type ejector, denoted “Q _H2 .in This fourth step 140 is carried out by computer 92 subsequently to step 120 of measurements and simultaneously to step 130.

More precisely, the flow rate of hydrogen admitted Q _h2 .m corresponds to the flow rate of hydrogen through the hydrogen inlet pipe 62 and the sonic neck 64 of the ejector 36 of the Venturi type.

The calculation of the allowed hydrogen flow rate Q _H2 .in depends on the flow regime within the Venturi-type ejector 36: the flow of hydrogen in the ejector can take two types of flow regime, depending on whether the ratio of the pressure P ₂ to the pressure Pi is lower or higher than a critical pressure ratio denoted V _cr .

The critical pressure ratio V _cr is calculated according to the following equation:

[Equation 5] in which y is the adiabatic coefficient of hydrogen at the temperature Ti measured by the temperature sensor 86, also called the Laplace coefficient.

Thus, when > V _cr , the flow regime of the admitted hydrogen is subsonic, i.e. the Mach number of the admitted hydrogen flow is less than 1. In this case, we denote Q _H2 .in _sub the hydrogen flow rate admitted, calculated using the following equation:

[Equation 6] in which: d is the efficiency of the sonic neck 64 so as to account for friction losses. Preferably, d is equal to 0.97. In practice d is adapted to each fuel cell system 10 thanks to an empirical calibration under controlled conditions;

- A _c is the smallest section of the sonic neck 64, expressed in square meters (m ² ); MW _{h 2} is the molar mass of dihydrogen;

- g is the adiabatic coefficient of dihydrogen at temperature Ti; and

- R is the universal ideal gas constant, in Joules per mole kelvin (Jxmol ¹ xK ¹ ), equal to approximately 8.314 Jxmol ¹ xK ¹ .

When - < V _cr , the flow regime of the admitted hydrogen is sonic, i.e. the Mach number of the flow of the admitted hydrogen is equal to 1. In this case of regime sonic flow, we denote Q _H 2.in _S0Jl the flow rate of hydrogen admitted, calculated using the following equation:

[Equation 7] with the variables mentioned in reference to equation 6.

Thanks to equations 6 and 7 presented above, it is possible to calculate during step 140 the hydrogen flow rate admitted Q _H2 .i _n as a function of the temperature Ti and the pressures Pi and P2, whatever the hydrogen flow regime.

The hydrogen leak detection method comprises a fifth step 150 of calculating the hydrogen leak rate in the fuel cell system 10, carried out by the computer 92 subsequently to steps 130 and 140. This leak rate of hydrogen is noted Q .ieak ^and is calculated using the following equation:

Qm.leak ⁼ QH2.in ^~ QH2.out

[Equation 8] in which:

- QH2.in equals Q _H 2.in _sub , calculated using equation 6 above, or Q _H 2.in _{son >} calculated using equation 7 above , depending on whether the hydrogen flow rate admitted into the Venturi-type ejector 36 is subsonic or sonic; and QH2.YES is calculated using equation 2 above.

Indeed, during normal operation of the fuel cell 12, the quantity of hydrogen consumed Q _H 2.out is equal to the quantity of hydrogen admitted Q _H 2.m · Thus, any difference between these two values is due to a hydrogen leak.

It should be noted that this equality is not verified during the transient phases of the operation of the fuel cell system 10, for example during the start-up of the system, when the anode compartment 16 and the circuits are filled with hydrogen. The hydrogen leak detection method therefore does not apply to these transient phases.

The hydrogen leak rate Q _H2 .i _e ak ^a i ^ns i obtained is expressed in grams per second (g/s).

Thus, it is possible to calculate the hydrogen leak rate in the fuel cell system 10, downstream of the Venturi-type ejector 36, with only the temperature Ti and the pressure Pi upstream of the measured data. ejector, the pressure P ₂ downstream of the ejector and the electric current generated by the fuel cell 12.

The method comprises a sixth step 160 of hydrogen leak detection in the fuel cell system 10. During this step, the hydrogen leak rate Q _H2 .i _e ak is compared with one or more threshold values to detect a possible hydrogen leak.

To perform this comparison, a first approach is to directly compare the hydrogen leak rate Qm.i _eak calculated during step 150 with a predefined leak detection threshold value, denoted Q _th r _es When the flow rate of calculated leak exceeds this threshold value Q _thres , then the computer 92 concludes that a leak is present. However, this approach has the drawback of being sensitive to measurement noise.

Thus, a second approach is preferred to perform this comparison. It consists of filtering the hydrogen leak rate Qm.i _eak calculated before performing the comparison with one or more leak detection threshold values.

Preferably, the filtering applied to the hydrogen leak rate Qm.i _eak is a first-order low-pass filter.

For example, the hydrogen leak rate Q _{H2 ieak} is filtered with a first-order low-pass filter with a time constant equal to ten seconds and the leak detection threshold Q _th r _es is set at 30 milligrams per second (mg/s).

The choice of filtering with a first-order low-pass filter with a time constant equal to ten seconds is advantageous because it makes it possible both to overcome measurement noise and the variability of measurements over time and to obtain a reactive detection method, that is to say capable of detecting a hydrogen leak very quickly after its appearance, and very precise.

It is observed that when the hydrogen flow admitted into the fuel cell system 10 is very low or zero, that is to say when the electric current produced by the fuel cell 12 is very low or zero, a Residual measurement error is unavoidable. To prevent this residual measurement error from leading to leak detection of hydrogen during step 160, which would then be a “false positive”, it is advantageous not to take account of small leaks, at a low flow rate of hydrogen admitted.

For this, the hydrogen leak rate Q _h2 ,i _e ak is filtered according to two different filters, then the results obtained by these two filterings are compared with two different detection thresholds, denoted Qthres.i and Qthres.2 in a variant of step 160.

For example, the two leak detection thresholds are defined as follows:

- the hydrogen leak rate Q _H2 .i _e ak ^is filtered with a first-order low-pass filter with a time constant equal to 1 second and the leak detection threshold Qthres.i is set at 60 milligrams per second (mg/s); and

- the hydrogen leak rate Qm.ieak ^is filtered with a first order low-pass filter with a time constant equal to 10 seconds and the leak detection threshold Qthres.2 is set at 30 milligrams per second (mg/ s).

In this example, to determine whether a hydrogen leak is taking place in the fuel cell system 10, the first detection threshold Qthres.i is always retained (A) and the second detection threshold Q _thres.2 , more selective , is retained only when the current produced by the fuel cell 12 is greater than or equal to 10 A.

Thus, the highest detection threshold is only considered when the hydrogen flow rate admitted is low, which makes it possible to avoid detecting a “false positive”, caused by the inaccuracy of the measurements at low flow rate.

Other values can be retained for the time constants of the two first-order low-pass filters, as well as for the associated detection thresholds Qthres.i and Qthres.2.

In general, the value of the time constant and the detection threshold of a first filter are respectively lower and higher than the value of the time constant and the detection threshold of a second filter.

It is also noted that the filtration of the hydrogen leak rate Q _H2,iea k overcomes the influence of the phase shift occurring during a purge between the opening of the purge and drain valve 54 and the admission of hydrogen compensating for the loss of hydrogen linked to the purge, since the duration of a purge is much less than the time constant of the filters used.

The hydrogen leak detection method comprises a seventh step 170 at the end of the cycle, carried out after the completion of the hydrogen leak detection step 160.

During the end of cycle step 170, if the computer 92 detects that the fuel cell system 10 is still in operation, then the method of detecting hydrogen leak begins a new cycle by again performing steps 110 to 170, beginning with start-up step 110.

Thus, throughout the operation of the fuel cell system 10, the hydrogen leak detection method is executed in real time.

In practice, this means that the data from the temperature 86 and pressure 88 and 90 sensors are transmitted in real time to the computer 92, as well as the measurement of the current produced by the fuel cell 10, and that a complete cycle of the hydrogen leak detection method is carried out at each sampling period of the computer 92.

The sampling period of computer 92 is for example between 1 millisecond and 10 milliseconds (ms).

More specifically, at each sampling period of the computer 92, the computer performs the calculations of steps 130, 140 and 150 and performs the hydrogen leak detection step 160.

There will now be described in more detail and with reference to FIG. 5 experimental results obtained by the hydrogen leak detection method applied to a fuel cell system 10 produced in accordance with the invention and not exhibiting any detectable leak. of hydrogen.

FIG. 5 compares: the calculation of the flow rate of hydrogen consumed Q _H2.0 ut by the fuel cell 12, carried out using equation 2 above, with the calculation of the flow rate of hydrogen Q _{H2 in} admitted to the entrance of the ejector 36 type

Venturi, made using equations 6 and 7 above.

This comparison is performed at numerous operating points of the fuel cell 12, that is to say for different values of electric current produced by the fuel cell. These different values of electric current produced are presented on the abscissa and the calculated hydrogen flow rates Q _H 2.out and Qm.in are presented on the ordinate.

To obtain these experimental results, the following relationships were applied to calculate the hydrogen flow rate Q _H2 inadmissible: [Equation 10]

To obtain these relationships, the following parameters were retained:

- A _c : smallest section of the sonic neck 64 equal to 1.43× ^10-6 m ² ;

- y: adiabatic coefficient of dihydrogen equal to 1.405; - d: equal to 0.97;

- MW _{h 2} : molar mass of dihydrogen, equal to 2.0159 g/mol; and

- R: universal ideal gas constant, equal to approximately 8.314 Jxmol ¹ xK ¹ .

It can be seen in FIG. 5 that over the entire operating range of the fuel cell 12, the differences between the calculation of the flow rate of hydrogen consumed Q _H2.0 ^u _t ^and the calculation of the flow rate of hydrogen Q _H2, not allowed are less than 10 milligrams per second (mg/s), except when the current produced by the fuel cell 12 is low, in practice less than 10 A, where the measurement inaccuracies lead to an error in the calculation of the flow rate of hydrogen admitted Q _H2,in of about 25 milligrams per second (mg/s).

Thanks to this comparison between the calculation of the hydrogen flow rate consumed and the calculation of the hydrogen flow rate admitted in the absence of a leak, the filtering thresholds used in step 160 in the case where two leak detection thresholds are used could be defined.

In addition, the small deviations noted in Figure 5 demonstrate the accuracy of the calculations used by the hydrogen leak detection method, in particular the accuracy of equations 1 to 8. Thus, the comparison represented in Figure 5 corresponds to a experimental confirmation on a fuel cell system 10 of the accuracy of the hydrogen leak detection method.

By “accurate”, it is meant here that the fuel cell system 10 and that the hydrogen leak detection method have the following advantages: good resolution, that is to say that the smallest variation of the calculation of the flow rate of hydrogen consumed Q _H 2.ou _t ^and of the calculation of the flow rate of hydrogen Q _H2.in allowed is low, in practice of the order of 1 milligrams per second (mg/s); and good accuracy, that is to say that the error between the flow rates of hydrogen consumed Qm.ou _t ^and of hydrogen Q _H2,in admitted calculated by the leak detection method and the hydrogen flow rates actual hydrogen consumed and admitted is low, in practice less than 2% FS (abbreviation for "full scale", i.e. the percentages are expressed as a percentage of the full scale) for a current electricity produced by the hydrogen fuel cell 12 greater than or equal to 10 A and less than 4% FS for an electric current produced less than 10 A. Thanks to the invention, it is therefore possible to propose a method making it possible to detect with precision, and in real time, hydrogen leaks occurring in a fuel cell system.

In addition, thanks to the invention, the physical principle of gas flow within the Venturi-type ejector 36 is used to enable the hydrogen leak detection method to be implemented.

In addition, the hydrogen leak detection method of the invention only requires the use of measuring tools that are simple to integrate into the fuel cell system 10 and inexpensive.

Indeed, the leak detection method of the invention only requires the use of a temperature sensor 86 and two pressure sensors 88 and 90. In addition, these sensors make it possible to perform other functions in the system. fuel cell 10.

For example, the second pressure sensor 90 is necessary to know the hydrogen pressure in the anode compartment 16 of the fuel cell 12, and thus regulate the admission of hydrogen into this anode compartment by the supply valve 38 .

Thus, it is simple and inexpensive to adapt a fuel cell system in order to be able to implement the leak detection method of the invention therein.

The leak detection method of the invention also has the advantage of being carried out during operation of the fuel cell system 10, which is more secure because it makes it possible to detect a leak triggered during the operation of the system. .

The leak detection method of the invention also has the advantage of not affecting the operation or the efficiency of the fuel cell system 10, which makes it possible not to increase the operating costs of the system.

The leak detection method of the invention also has the advantage of making it possible to detect all of the leaks produced in the fuel cell system 10 downstream of the Venturi-type ejector 36, independently of the location. in the system of these leaks, since the method is based on the comparison between the flow of hydrogen consumed and the flow of hydrogen admitted. In other words, the steps 130, 140, 150 and 160 are such that the method detects all of the hydrogen leaks occurring in the fuel cell system downstream of the Venturi-type ejector, i.e. that is, the hydrogen leak detection method is configured to detect all of the hydrogen leaks occurring in the fuel cell system downstream of the Venturi-type ejector. It should be noted that the recirculation of hydrogen in the recirculation circuit 60 does not affect the hydrogen leak detection method. Indeed, this recirculation makes it possible to increase the flow rate of hydrogen in the fuel cell 12, but does not modify either the quantity of hydrogen admitted, nor the quantity of hydrogen consumed. As a variant of the invention, not shown, the supply valve 38 is replaced by an injector providing a discontinuous hydrogen supply.

In a not shown variant of the invention, the hydrogen supply system 30 comprises a gas pressure reducer arranged between the hydrogen tank 32 and the supply valve 38. In a not shown variant of the invention, the purge valve 54 is replaced by a calibrated orifice, or by a proportional valve, in order to allow continuous purge of the anode compartment 16 of the fuel cell 12.

As a variant of the invention, not shown, the hydrogen leak rate Qm.ieak is filtered according to a number of filters different from two, for example according to three different filters. In this case, the result obtained by each of these filterings is compared with an associated detection threshold.

The embodiment and the variants contemplated above can be combined to generate new embodiments in the invention.

Claims

1. Method for detecting hydrogen leakage ( Qm.ieak ) in a fuel cell system (10), the fuel cell system comprising at least: a stack of electrochemical cells (14) forming a fuel cell ( 12) comprising an anode compartment (16) and a cathode compartment (18) separated by a proton exchange polymer membrane (20); a hydrogen supply system (30) comprising a hydrogen tank (32) and a supply circuit (34) connecting the hydrogen tank to an inlet (22) of the anode compartment (16) of the fuel, the supply circuit comprising an ejector (36) of the Venturi type; a recirculation circuit (60) of hydrogen not consumed by the fuel cell between an outlet (24) of the anode compartment of the fuel cell and the Venturi type ejector (36) of the supply circuit, the recirculation unconsumed hydrogen being driven by the ejector by Venturi effect; and - a purge system (50) comprising a valve (54) for purging and draining the anode compartment, characterized in that the hydrogen leak detection method comprises at least the following steps: a) calculating (130) the total flow rate of hydrogen consumed ( Qm.out ) by the fuel cell system (10); b) calculating (140) the flow rate of hydrogen admitted (Q _H2 ^.i _n ) by the hydrogen supply system ⁽ 30) in an inlet pipe (62) of the ejector (36) of the type Venturi; c) determining (150) the leak rate (Q _H 2.ieak) by calculating the difference between the flow rate of hydrogen admitted and the total flow rate of hydrogen consumed; and d) detecting (160) a possible hydrogen leak in the fuel cell system (10) by comparing the leak rate with at least one Threshold value ( Qthres Qthres.l Qthres.2) such that the method detects all of the hydrogen leaks ( Qm.ieak ) ^occurring in the fuel cell system (10) downstream of the Venturi-type ejector (36).

2. Hydrogen leak detection method (Q _H 2.i _eak ) according to claim 1, characterized in that during step a), the calculation (130) of the total flow hydrogen consumed Q _H 2.out P ^{ar the} fuel cell system (10) is made from the following sum (equation 2):

QHÏ.OUÎ — QH2.out _sto + QH2.out _xo Ί QH2.out _gurg where:

- Q _H 2.out _sto is ^the basic rate of hydrogen consumed by the fuel cell (12) by electrochemical reaction;

- QH2. _{where puv} g is ^the flow rate of hydrogen lost in the purges of the anode compartment (16) of the fuel cell; and

- QH2. _ORIXO is the flow rate of hydrogen passing through the proton exchange polymer membrane (20) from the anode compartment (16) to the cathode compartment (18).

3. Hydrogen leak detection method (Qm.ieak) according to claim 2, characterized in that during step a), the flow of hydrogen lost QH2.OUÎ puv g in the purges of the anode compartment (16) of the fuel cell (12) is calculated according to the following equation (equation 1): in which :

- MW _{h 2} is the molar mass of dihydrogen;

- R is the universal ideal gas constant;

- T is the temperature within the anode compartment (16);

V anode is the volume of the anode compartment (16); and

Panode.in is the pressure gradient (P2) measured at the inlet of the anode compartment (16) during a purge.

4. Hydrogen leak detection method (Q _H 2.ieak) according to claim 3, characterized in that the pressure gradient Panode.in is obtained by a method of purging in constant admission consisting in postponing the restoration of the pressure lost in the anode compartment (16) of the fuel cell (12) during a purge by delaying the opening of a hydrogen supply valve (38) of the supply system (30).

5. Hydrogen leak detection method (Q _H2 .ieak) according to any one of claims 2 to 4, characterized in that during step a), the flow of hydrogen consumed by permeation Q _H2.0 ut _xo through the proton exchange polymer membrane (20) from the anode compartment (16) to the cathode compartment (18) is calculated according to the following equation (equation 4): in which :

MW _h is the molar mass of dihydrogen;

N is the number of electrochemical cells (14) of the fuel cell (12); F is Faraday's constant; J _Xo is the crossover current density; and S _membrane is the surface of the membrane (20) of an electrochemical cell (14).

6. Hydrogen leak detection method {Q _H 2.ieak) according to any one of the preceding claims, characterized in that during step b), the flow rate of hydrogen admitted Q _H2 .i _n by the hydrogen supply system ⁽ 30) in the inlet pipe (62) of the Venturi type ejector (36) is calculated according to whether the flow regime occurring within the type ejector Venturi is a subsonic flow regime or a sonic flow regime.

7. Hydrogen leak detection method (Qm.ieak) according to claim 6, characterized in that:

• when the flow regime is subsonic, the flow rate of hydrogen Q _H 2.in _sub admitted by the hydrogen supply system (30) into the intake pipe (62) of the ejector (36) of Venturi type is calculated according to the following equation (equation 6): in which :

Pi is the pressure of the hydrogen admitted into the inlet pipe (62) of the Venturi-type ejector (36);

Ti is the temperature of the hydrogen admitted to the inlet of the Venturi-type ejector;

P is the hydrogen pressure at the outlet of the Venturi-type ejector; d is the efficiency of the sonic neck (64) of the Venturi-type ejector (36);

- A _c is the smallest section of the sonic neck (64) of the ejector (36); MW _h 2 is the molar mass of dihydrogen; y is the adiabatic coefficient of dihydrogen; and

- R is the universal ideal gas constant, and

• when the flow regime is sonic, the hydrogen flow QH2.in _S0Jl admitted by the hydrogen supply system (30) into the intake pipe (62) of the Venturi-type ejector (36) is calculated with the following equation (equation 7): in which :

Ti is the temperature of the hydrogen admitted into the inlet manifold of the Venturi-type ejector; d is the efficiency of the sonic neck (64) of the Venturi-type ejector (36);

- A _c is the smallest section of the sonic neck (64) of the ejector (36);

MW _{h 2} is the molar mass of dihydrogen;

- y is the adiabatic coefficient of dihydrogen; and

- R is the universal ideal gas constant.

8. Hydrogen leak detection method (Q _H 2.ieak) according to any one of the preceding claims, characterized in that the leak rate (Q _H 2.ieak) determined (150) in step c ) is filtered before performing the comparison with at least one threshold value ( Qthres Qthres.h Qthres.2)

9. Hydrogen leak detection method (Q _H 2.ieak) according to claim 8, characterized in that the leak rate (Q _H 2.ieak) determined (150) in step c) is filtered according to two different filters and in that the results obtained by these two filters are compared with two different detection thresholds (Qthres.i, Qthres.2) ^'

- a first leak detection threshold (Q _thres .1) compared to the leak flow rate filtered with a first-order low-pass filter with a time constant equal to a first value; and

- a second leak detection threshold ( Qthres.2 ) compared to the leak flow rate filtered with a first-order low-pass filter with a time constant equal to a second value, in that the first leak detection threshold has a value greater than that of the second leak detection threshold, and in that the first time constant value is less than the second time constant value.

10. Hydrogen leak detection method (Q _H 2.ieak) according to any one of the preceding claims, characterized in that the hydrogen leak detection method is carried out cyclically, in real time, and in that, preferably, steps c) and d) of the hydrogen leak detection method are carried out by a computer (92) for controlling the fuel cell system (10) at each sampling period of the computer .

11. Fuel cell system (10) for implementing the hydrogen leak detection method (Q _H 2.ieak) of one of claims 1 to 10 comprising at least: a stack of electrochemical cells (14) forming a fuel cell (12) comprising an anode compartment (16) and a cathode compartment (18) separated by a proton exchange polymer membrane (20); a hydrogen supply system (30) comprising a hydrogen tank (32) and a supply circuit (34) connecting the hydrogen tank to the inlet (22) of the anode compartment (16) of the cell fuel, the supply circuit comprising an ejector (36) of the Venturi type; a recirculation circuit (60) of hydrogen not consumed by the fuel cell between the outlet (24) of the anode compartment of the fuel cell and the Venturi type ejector (36) of the supply circuit, the recirculation unconsumed hydrogen being driven by the ejector by Venturi effect;

- a purge system (50) comprising a valve (54) for purging and draining the anode compartment; and a computer (92) configured to implement steps a) to d) of the hydrogen leak detection method.

12. Fuel cell system (10) according to claim 11, characterized in that the fuel cell system (10) further comprises a pressure sensor (88) and a temperature sensor (86) arranged upstream of the Venturi-type ejector (36) and a pressure sensor (90) arranged downstream of the Venturi-type ejector.