Method for operating a fuel cell

Abstract

The invention relates to a method for operating a fuel cell having a "ping-pong" architecture, in which two groups (10, 20) of electrochemical cells are alternately supplied. Advantageously, a purge phase is carried out in synchronization with a displacement of the non-reactive fluids accumulated in the form of a so-called "plug" (B) aggregate, when said plug (B) is, at least, partly located in the discharge lines (120, 220) of the fluidic circuit (4).

Description

[0001]"Method for operating a fuel cell" TECHNICAL FIELD OF THE INVENTION The present invention relates to the field of fuel cells, in particular proton exchange membrane fuel cells (or PEMFC for Proton Exchange Membrane Fuel Cell) comprising an alternating power supply, called "ping-pong". It can advantageously be implemented to optimize the performance of a PEFMC with ping-pong architecture. STATE OF THE ART A fuel cell is formed from a stack of "unit" electrochemical cells, each comprising an anode and a cathode electrically separated from each other by an electrolyte. In the case of a hydrogen cell, the fuel (hydrogen) is brought into contact with the anode, and the oxidant (oxygen) is brought into contact with the cathode. Oxidation and reduction reactions take place at the anode and cathode respectively, producing electricity, water and heat.The electrolyte can be in the form of a membrane that allows the protons resulting from the hydrogen oxidation reaction to pass through. This is the case for proton exchange membrane fuel cells (PEMFCs). The cell stack is only the site of the reaction: the reactants must be brought in, the products and non-reactive species must be removed, as must the heat produced. Generally separate anode and cathode fluid circuits are used to supply the cell stack with fuel and oxidant respectively, and to remove the products from the stack. The hydrogen supplying the anode compartment can be advantageously reinjected into the anode compartment, by a recirculation pump on the fluid circuit, so as to distribute the non-reactive species at the anode, typically the water produced by the reaction and the nitrogen from the cathode across the membrane.The fluid circuit can be more complex in the case of particular cell architectures. In the so-called "ping-pong" architecture disclosed by patent document FR2975227, the cell is divided into several groups of cells fluidically connected to each other by their respective outlets. In this known architecture illustrated in Figure 1, the supply at inlet 11, 21 of the different groups 10, 20 is done alternately during certain operating phases of the cell. A first group 10 is supplied at inlet 11 by the combustible fluid (hydrogen) while the supply of a second group 20 is cut off at inlet 21. The combustible fluid passes through the first group, which then operates nominally, and leaves moistened and slightly depleted at outlet 12 of the first group 10. This fluid then supplies the second group 20 via outlet 22 of the second group 20.This direct supply of the first group and in the opposite direction of the second group aims to mix pockets of highly depleted fluid formed within the cells of the first group, to dissipate them by mixing them. This prevents cells of the first group from operating for a prolonged period in the presence of a highly depleted stagnant fluid. The second group 20 is then supplied at inlet 21 with the combustible fluid while the supply of the first group 10 is cut off at inlet 11. The combustible fluid passes through the second group, which then operates nominally, and leaves moistened and slightly depleted at outlet 22 of the second group 20. This fluid then supplies the first group 10 via outlet 12 of the first group 10.This direct feeding of the second group and in the opposite direction of the first group also makes it possible to dissipate pockets of highly depleted fluid formed within the cells of the second group during the previous cycle, by mixing them. This prevents cells of the second group from operating for a prolonged period in the presence of a highly depleted stagnant fluid. The principle of this ping-pong architecture consists of alternately feeding the different groups of cells so as to alternate phases of nominal operation with phases of feeding in the opposite direction for each of the groups. After a few cycles of alternating feeding, the groups 10, 20 are fed simultaneously and a purge valve 40 common to the different groups is opened so as to evacuate the accumulated non-reactive products and gases, in particular water and nitrogen.In constant (imposed) current operation, a voltage drop is generally observed during the purging of these accumulated non-reactive products and gases. An object of the present invention is to provide a method of operating a fuel cell with ping-pong architecture, improving the performance of the fuel cell, and in particular limiting the voltage or power drop during purging. Other objects, features and advantages of the present invention will become apparent upon examination of the following description and the accompanying drawings. It is understood that other advantages may be incorporated.SUMMARY OF THE INVENTION To achieve this objective, according to one embodiment, a method for operating a fuel cell is provided, said fuel cell comprising a plurality of groups of electrochemical cells connected to each other by a fluid circuit, said fluid circuit comprising, for each group, a supply line at the inlet of said group, and a switching member on said supply line, and, a discharge line at the outlet of said group, said fluid circuit being configured so that the discharge lines communicate fluidically with each other and are connected to a common purge member. The operating method comprises at least: ^ An alternating supply cycle comprising at least: i.A first supply phase in which a first group among the plurality of groups is supplied directly with supply fluid, at its inlet, the switching member of said first group being on while the switching members of the other groups of the plurality of groups are blocked, said other groups being supplied indirectly, during this first supply phase, at their respective outlets, by the supply fluid from the outlet of the first group, ii.A second supply phase in which a second group different from the first group is supplied directly with supply fluid, at its inlet, the switching member of said second group being on while the switching members of the other groups of the plurality of groups are blocked, said other groups being supplied indirectly, during this second supply phase, at their respective outlets, by the supply fluid coming from the outlet of the second group, ^ A purge phase following the cycle of alternating supplies, said purge phase comprising an opening of the purge member so as to evacuate non-reactive fluids coming at least in part from the supply fluid.Advantageously, the purge phase is triggered in a synchronized manner with a movement of the non-reactive fluids accumulated in the form of an aggregate called a plug, when said plug is located at least partly in the discharge lines. Thus synchronized, the purge phase of a fuel cell with ping-pong architecture is triggered at the optimal moment with respect to the alternating supply cycle. A principle of the method according to the present invention consists in triggering the opening of the purge member in a controlled manner, by determining the optimal moment to trigger this opening, and preferably the duration of this opening. In particular, the triggering of the purge phase is done during a supply phase ^ of the alternating supply cycle and before the end of the supply phase ^.This approach is fundamentally different from the approach taught in patent document FR2975227, which simply purges the groups of the cell after a few power cycles, during a "dead end" phase where the two groups are first powered simultaneously in order to push back the non-reactive species near the outlet of the two respective groups, before opening the purge member. This dead end phase is carried out immediately at the end of a standard duration power phase of one of the groups. This causes a drop in the voltage delivered by the cell and a drop in performance during the purge phase. This leads to increasing the purge frequency, in order to limit the voltage drop. As a result, the loss of hydrogen in the purges, and therefore the hydrogen consumption by the system, are increased.On the contrary, the method according to the present invention is based on a better understanding of the reaction and fluidic mechanisms occurring within the cell. In the context of the development of the present invention, it was identified that the non-reactive products and gases accumulate in a localized manner in the fluidic circuit, in the form of an aggregate called a plug. This plug moves according to the injections of feed fluid alternately in each of the groups of the cell. This movement causes a limited mixing of the gas mixture, inducing only a spreading of the plug, without achieving a real dilution of the non-reactive species in the flow of reactive species. In the context of the development of the present invention, it was observed that the dead-end phase following the end of a feed phase of a group, as disclosed by document FR2975227, does not allow the plug to be moved outside the cell group.According to this known method, the plug cannot be located in the discharge lines. Indeed, it has been identified that the plug is pushed by the feed and pulled by the consumption of reagents. Since the latter does not exist in the discharge lines, the dead end phase stabilizes the position of the plug where it is at the end of the feed phase, without moving it in the discharge lines. However, it has been determined that the plug is not in the discharge lines at the end of the feed phase. It is therefore necessary to synchronize the triggering of the purge phase with the movement of the plug, when it is in the discharge lines. According to the invention, the triggering of the purge phase is carried out before the end of the feed phase which precedes the purge phase.By comparison, implementing the operating method according to the present invention makes it possible to significantly reduce the purge frequency. This can make it possible to reduce the quantity of hydrogen released during the purge phase by up to approximately 60%, while obtaining a voltage delivered by the cell that is substantially constant. It has also been observed that the plug tends to oscillate around an equilibrium position in the fluid circuit. For example, when the cell comprises two identical groups connected by symmetrical discharge lines, this equilibrium position is typically located at the midpoint between the outlets of the groups, equidistant from the outlets. According to a preferred possibility, the purge member is located at this equilibrium position. At this equilibrium position, the plug has a substantially constant speed of movement regardless of its direction of movement.Triggering the opening of the purge member is then facilitated, regardless of the power cycle that precedes this triggering. Due to the inertia of the plug's movement, the plug passes through its equilibrium position with a delay after opening one of the switching members. An object of the invention consists in particular in determining this delay so as to open the purge member at the moment when the plug passes through its equilibrium position, as close as possible to the purge member. This makes it possible to minimize the loss of unconsumed fuel. This makes it possible to reduce the opening time of the purge member. The stability of the performance of the fuel cell is improved. Furthermore, by triggering the opening of the purge member at the moment when the plug passes in the immediate vicinity, it has been observed that the voltage delivered by the cell, at constant current and constant hydrogen pressure, hardly varies.Proper synchronization of the opening of the purge member with the movement of the plug thus advantageously makes it possible to avoid a drop in the voltage delivered by the cell. With a higher overall voltage of the fuel cell during this purge phase and a lower rate of hydrogen released, the overall efficiency of the fuel cell and its stability are improved. Consequently and advantageously, the method for operating a fuel cell having a ping-pong architecture according to the present invention makes it possible in particular, by optimizing the purge time, to reduce the consumption of combustible fluid and to improve the electrical performance of this fuel cell.BRIEF DESCRIPTION OF THE FIGURES The aims, objects, as well as the characteristics and advantages of the invention will emerge more clearly from the detailed description of an embodiment thereof which is illustrated by the following accompanying drawings in which: Figure 1 represents a fuel cell having a ping-pong architecture according to the prior art. Figure 2 represents a simplified diagram of a fuel cell having a ping-pong architecture, according to an embodiment of the present invention. Figure 3 represents cycles of alternating supply of inputs to the groups of the cell, according to an embodiment of the present invention. Figure 4A represents the movement of the plug of non-reactive products and gases during a first phase of supply of the fuel cell, according to an embodiment of the present invention.Figure 4B represents the movement of the plug of non-reactive products and gases during a second phase of supply of the fuel cell, according to an embodiment of the present invention. Figure 5 represents a purge phase when the plug is at an equilibrium position near the purge member, according to an embodiment of the present invention. Figure 6 represents cycles of alternating supply of inputs to the groups of the cell, with different purge times, according to an embodiment of the present invention. Figure 7A represents the voltage delivered by the fuel cell during its operation, by carrying out periodic purges according to a first purge time, according to an embodiment of the present invention.Figure 7B represents the voltage delivered by the fuel cell during its operation, by performing periodic purges according to a second purge time, according to an embodiment of the present invention. Figure 7C represents the voltage delivered by the fuel cell during its operation, by performing periodic purges according to a third purge time, according to an embodiment of the present invention. Figure 7D represents the voltage delivered by the fuel cell during its operation, by performing periodic purges according to a fourth purge time, according to an embodiment of the present invention. Figure 8 represents a simulation of the nitrogen concentration at the purge member with respect to the alternating supply phases, according to an embodiment of the present invention. The drawings are given as examples and are not limiting of the invention.They constitute schematic representations of principle intended to facilitate the understanding of the invention and are not necessarily on the scale of practical applications. In particular, the different members and the different parts of the fluid circuit are illustrated by diagrams which are not representative of reality. DETAILED DESCRIPTION OF THE INVENTION Before beginning a detailed review of embodiments of the invention, optional characteristics which may possibly be used in association or alternatively are set out below: According to one example, the alternating supply cycle has a period T and the purge phase is triggered during a supply phase ^ and before the end of the supply phase ^, at the end of a delay ^ ^ ^ after the start of the supply phase ^ such that: 1 1 + exp(^^) ^ ^ ^ ≥ 0.8. ln ^ 2 with ^ = ^ / 2 the half-period, and. With ^ the active surface of the battery, ^ the current density and ℱ the Faraday constant, ^ the ideal gas constant, ^ the temperature of the battery, ^ the volume crossed by the fluid in the battery, ^ the pressure of the fluid in the battery and ^ ^^^the saturated vapor pressure at the cell temperature. The plug typically comprises nitrogen and water vapor in addition to hydrogen which is in reduced concentration. According to one example, the alternating feed cycle has a period T and the purge phase is triggered during a feed phase ^ and before the end of the feed phase ^, after a delay ^ ^ between 0.2. T / 2 ≤ ^^ ≤ 0.7. T / 2. According to one example, the alternating feed cycle has a period T and the purge phase is triggered during a feed phase ^ and before the end of the feed phase ^ , after a delay ^ ^ ^ = T / 4. In practice, liquid water is also present in the fluid circuit. This liquid water can affect the dynamics of the plug movement. A delay approximately equal to T / 4 makes it possible to take into account a delay in the movement of the plug due to liquid water, and / or to evacuate this liquid water.According to one example, the purge phase includes a "dead end" delay and the opening of the purge member is carried out after the dead end delay. According to an alternative example, the purge phase does not include a "dead end" delay and the opening of the purge member is carried out at the triggering of the purge phase. According to one example, the stack comprises only two groups, the outlets of these two groups are located at an equal distance from the purge member, and the first and second supply phases of the alternating supply cycle have the same duration T / 2. This "symmetrical" construction of the stack and the alternating supply cycle makes it possible to better control the oscillations of the plug in the fluid circuit. This facilitates the determination of the optimal moment for triggering the opening of the purge member.According to one example, the purge phase comprises, before or upon opening of the purge member, an inlet supply to all the groups, each switching member being conductive. According to one example, the opening of the purge member is carried out after the inlet supply to all the groups, after a so-called "dead-end" delay. This makes it possible to stabilize the plug in its equilibrium position. If the triggering of the purge phase is carried out at the right time, as determined by the invention, the dead-end phase freezes the movement of the plug between the groups, in the discharge lines. According to one example, the inlet supply to all the groups is carried out simultaneously with the opening of the purge member. According to one example, the purge phase does not comprise an inlet supply to all the groups before the opening of the purge member. The opening of the purge member thus directly follows a partial supply phase of a group.This makes it possible to shorten the purge phase. The performance of the cell is improved or more constant. According to one example, a constant electric current is imposed on the fuel cell. According to one example, the feed fluid injected during the alternating feed cycle has a constant pressure. According to one example, the opening of the purge member is triggered when at least 70%, and preferably at least 80%, of the length LB of the plug is located in the discharge lines, the length LB of the plug being measured in the direction in which the discharge line extends at the point in question. According to one example, the opening of the purge member is triggered when the plug is entirely located in the discharge lines. The discharge of the plug is optimized. The quantity of unconsumed feed fluid discharged is reduced.According to one example, the opening of the purge member is triggered when the plug completely occupies the discharge lines. The plug may possibly overflow into the active zones of the cell. According to one example, the opening of the purge member is maintained so that at least a portion of the plug is evacuated from the fluid circuit through the purge member, preferably at least 70%, preferably entirely. The opening time of the purge member is typically determined as a function of the volume of the plug or the quantity of non-reactive products and gases to be evacuated. The opening time of the purge member is minimized when the plug is at the purge member or as close as possible to the purge member. The quantity of unconsumed feed fluid discharged is reduced. The voltage drop is limited or eliminated. The performance of the cell is optimized.According to one example, the fuel cell comprises: ^ a first group of electrochemical cells having a first inlet and a first outlet, ^ a second group of electrochemical cells having a second inlet and a second outlet, ^ a fluid circuit intended to supply said first and second groups with a fluid, and to evacuate said fluid from said first and second groups. The fluid circuit typically comprises a so-called upstream part comprising: ^ a main supply conduit configured to be connected to a reservoir storing the fluid at a pressure P1, called high pressure, ^ a first supply line connected to the main supply conduit and connected to the first inlet of the first group, ^ a second supply line connected to the main supply conduit and connected to the second inlet of the second group.The fluid circuit typically comprises a so-called downstream part comprising: ^ a first discharge line connected to the first outlet of the first group, ^ a second discharge line connected to the second outlet of the second group, ^ said first and second discharge lines being connected to each other, so as to allow a supply of fluid to the first group via the first outlet, by the fluid having passed through the second group, and conversely so as to allow a supply of fluid to the second group via the second outlet, by the fluid having passed through the first group, ^ a main discharge conduit connected to the first and second discharge lines, and configured to be connected to an exhaust.The cell typically further comprises: ^ at least one expansion member located in the upstream part of the fluid circuit, configured to reduce the pressure of the fluid coming from the reservoir, from high pressure to low pressure, ^ a first switching member on the first supply line, configured to allow or block a flow of the fluid towards the first inlet, ^ a second switching member on the second supply line, configured to allow or block a flow of the fluid towards the second inlet, ^ a purge member on the main discharge conduit, configured to allow or block a flow of the fluid towards the exhaust.According to one example, the at least one expansion member comprises a first expansion member configured to decrease the pressure of the fluid from the reservoir, from high pressure to medium pressure, and a second expansion member configured to decrease the pressure of the fluid from medium pressure to low pressure. Unless incompatibility, technical characteristics described in detail for a given embodiment may be combined with the technical characteristics described in the context of other embodiments described by way of example and non-limiting. In particular, elements described or illustrated for certain embodiments of the cell and its operating method may be combined so as to form another embodiment which is not necessarily illustrated or described. Such an embodiment is obviously not excluded from the invention.A fuel cell according to the present invention comprises at least two groups of electrochemical cells. Those skilled in the art will have no difficulty in implementing an embodiment comprising more than two groups of electrochemical cells. In the context of the present invention, the term "fuel cell system with ping-pong architecture" means a system comprising at least two groups of electrochemical cells distributed in one or more stacks. Each group is thus made up of a series of electrochemical cells electrically and fluidically interconnected. The cells are typically in the form of a membrane-electrode assembly commonly referred to as MEA. The cell here comprises at least two stacked cells, and preferably at least ten.The ping-pong architecture fuel cell system typically has different operating phases, in particular, a first phase during which a first group is supplied directly at the inlet of the first group, the second group being supplied only at the outlet of the second group by the gas leaving the first group, a second phase during which the second group is supplied directly at the inlet of the second group, the first group being supplied only at the outlet of the first group by the gas leaving the second group, and a purge phase. The purge phase optionally comprises a third phase of simultaneous supply of the first and second groups at the inlet of the first and second groups, and an opening of the purge member. The first and second phases are carried out alternately and are typically of equal duration, with changes in regimes close to the cell. The purge phase follows the first and second phases.One principle of the invention is to trigger the purge phase after a partial feed phase, i.e. a shortened feed phase, so that the plug is located in the right place to open the purge member. Thus, the feed phase that precedes the purge phase has a duration that is shorter than the duration of each first and second phase of an alternating feed cycle. Determining this shorter duration is an object of the present invention. The terms "high pressure", "medium pressure" and "low pressure" are perfectly clear to those skilled in the art. A high pressure is strictly greater than a medium pressure. A medium pressure is strictly greater than a low pressure. A high pressure is generally greater than 50 bars, or even greater than 100 bars. A high pressure can reach up to 700 bars, or even more depending on the conditions.A medium pressure is generally between 5 and 40 bars, typically between 5 and 20 bars. A low pressure is generally less than 4 bars. In the context of the present invention, an inlet or an outlet of a group of the cell, typically each have a structural aspect and a functional aspect. Thus structurally, the inlet and the outlet correspond to first and second passage orifices for the combustible fluid. Functionally, the inlet and the outlet respectively designate the admission and the discharge of the combustible fluid. To the extent that the direction of circulation of the fluid within the group is alternately reversed, the inlet and the outlet can be functionally interchanged. Thus, the first passage orifice can form the inlet or the outlet functionally, and conversely, the second passage orifice can form the outlet or the inlet functionally.To facilitate understanding of the circulation of the fluid in the cells of the battery, we retain only the terms inlet / outlet and their corresponding references on the accompanying drawings, respectively X1 (X= 1…2) for the inlet and X2 (X= 1…2) for the outlet, independently of their functional assignment. In the context of the present invention, the outlets of the groups are in fluid communication with each other, via the discharge lines specific to each of the groups. Each discharge line extends between the outlet of the group considered and a common main discharge conduit. Each discharge line may borrow sections of other lines, for example other discharge lines. Thus, sections of a given discharge line may be common with sections of other discharge lines. Tapping points or branches may be physically present along the discharge line(s).Devices may also be present along these discharge lines, for example regulating devices. All the discharge lines may form an outlet manifold of the cell connected to the main discharge conduit of the cell. The purge device is typically located on the main discharge conduit of the cell, preferably near an intersection of the discharge lines. The operation of the cell according to the invention is based on a succession of supply and discharge phases that differ from each other. Unless explicitly stated, the terms "succession" or "successive" do not necessarily imply, even if this is generally preferred, that the phases follow one another immediately, intermediate phases or steps being able to separate them. In the attached figures, a direction of circulation of the fluid in the fluid circuit is indicated by an arrow.In the attached figures, for the sake of clarity, a single cell comprising two groups is illustrated. It is understood that the description of this cell extends to all variants in number of cells, groups, etc. In the examples which follow, the fuel cell is described and illustrated for a ping-pong architecture comprising two groups of substantially identical cells. It is nevertheless perfectly conceivable to implement more than two groups and / or groups of different dimensions, without departing from the general principle of synchronized purging explained below. The groups can also be distributed in different cells fluidically connected to each other. The original idea implemented in the context of the development of the present invention consists in particular in triggering the opening of the purging member at the optimal moment, when the plug formed by the non-reactive products and gases passes in the immediate vicinity of the purging member.The existence and movement of a plug have indeed been observed through numerical simulations in the context of the ping-pong power supply, while this operation could suggest a homogenization linked to mixing. Triggering the purge phase at the optimal moment makes it possible to limit the release of combustible gas. This makes it possible to reduce the opening time of the purge member. This makes it possible to stabilize the voltage delivered by the battery. The performance of the battery is thus generally improved, particularly in terms of consumption and power supplied. A "plug" is understood to mean a volume of gas comprising a significant fraction of non-reactive species and / or reaction products accumulated and located or localizable in the fluid circuit. Such a plug does not mean that the fluid circuit is completely obstructed. The plug may, but is not necessarily, form a partial obstruction of the fluid circuit.The plug is typically similar to a pocket of gas not participating in the reaction. It is mainly composed of nitrogen and water vapor, it can also contain other non-reactive species present in the air or resulting from parasitic reactions such as Argon or carbon dioxide. This pocket of gas becomes a hindrance to the reaction and constitutes the plug when the proportion of dihydrogen in the gas mixture falls below 70% (excluding humidity). The formation of the plug and its movements can be demonstrated by numerical simulation. In the following, the operating method is implemented and described for the anode compartment of a hydrogen fuel cell, i.e. for the fluid circuit supplying the cells on the anode side with hydrogen (fuel).The method can also be implemented in the cathode compartment supplied with oxidant, i.e. for the fluid circuit supplying the cells on the cathode side, typically when the oxidant consists mainly of oxygen (oxygen content for example greater than 50% in the dry gas). As mentioned above, Figure 1 illustrates a fuel cell 1 having a typical ping-pong architecture. It is connected to a high-pressure hydrogen tank 2 and comprises a fluid circuit composed of a part 3 upstream of the groups 10, 20 of electrochemical cells, and a part 4 downstream of the groups 10, 20 of electrochemical cells. The groups 10, 20 can be nested in a single stack, with alternating cells of the first group and the second group, as described in patent document FR2975227.The upstream part 3 typically comprises a shut-off valve 30 at the outlet of the tank 2, a main supply conduit 300 on which a first pressure reducer 31 and a second pressure reducer 32 are successively mounted. The first pressure reducer 31 makes it possible to reduce the pressure of the fluid coming from the tank 2 from a so-called high pressure of the order of a few hundred bars to a so-called medium pressure of the order of a few tens of bars. The second pressure reducer 32 makes it possible to reduce the pressure of the fluid coming from the first pressure reducer 31 from the medium pressure to a so-called low pressure of the order of a few bars. The upstream part 3 of the fluid circuit also comprises a first supply line 100 connecting the main supply conduit 300 to the inlet 11 of the first group 10 of electrochemical cells, and a second supply line 200 connecting the main supply conduit 300 to the inlet 21 of the second group 20 of electrochemical cells.The first and second supply lines 100, 200 each respectively comprise at least one switching member, typically a first switching valve 101v and a second switching valve 201v ensuring either the passage or the blocking of the fluid to each of the groups 10, 20. Other supply configurations, for example with injectors, are also possible. The operating method described here focuses more particularly on the purging phase of the cell, regardless of the supply configuration allowing alternating supply of the different groups of the cell. The downstream part 4 of the fluid circuit typically comprises a first discharge line 120 connecting the outlet 12 of the first group 10 of electrochemical cells to the main discharge conduit 400, and a second discharge line 220 connecting the outlet 22 of the second group 20 of electrochemical cells to the main discharge conduit 400.The first and second discharge lines 120, 220 typically form an outlet manifold of the cell 1. The main discharge duct 400 is connected to an exhaust 41, and is provided with a purge valve 40. Figure 2 illustrates in a simplified manner certain elements of the fuel cell illustrated in Figure 1. In particular, the groups 10, 20 of electrochemical cells and the discharge lines 120, 220 of the fluid circuit are schematically aligned here along a single axis x. This makes it easier to visualize a position of the products and non-reactive gases accumulated along the x axis, as a function of the alternating supply phases. The point M here corresponds to the intersection of the discharge lines 120, 220 and the main discharge duct 400. The point M is here located at an equal distance from the outlets 12, 22.The cell here has a construction symmetry which facilitates the understanding of the mechanisms of displacement of the "plug" of non-reactive products and gases in the groups 10, 20 and in the discharge lines 120, 220. Even if this construction symmetry is preferred in practice, other cell dimensions are possible, for example with discharge lines of different lengths connected to groups of different sizes. Those skilled in the art will be able to adjust this dimensioning according to their needs. Figure 3 illustrates cycles of alternating supply of the two groups 10, 20. A supply cycle here has a period T and comprises a first phase where only one of the groups is supplied for a duration between t0 and t1, for example the first group 10, followed by a second phase where only the other of the groups is supplied for a duration between t1 and t2, for example the second group 20.According to one possibility, in particular when the groups 10, 20 are dimensioned in the same way, the duration of the first phase is equal to the duration of the second phase t1 – t0 = t2 – t1 = T / 2. This simplifies the dynamics of movement of the plug within the groups. The movement of the plug typically exhibits periodic or pseudoperiodic oscillations in this case. Other alternating feeding cycles are nevertheless possible. Figures 4A, 4B schematically illustrate an alternating feeding cycle of period T with a first feeding phase of duration T / 2 and a second feeding phase of duration T / 2. The period T is typically between 0.1 s and 10 s. According to one example, the period T, i.e. the duration of an alternating feeding cycle, is of the order of 1 s. According to one example, the half-period T / 2, that is to say the duration of a feeding phase, is between approximately 0.3 s and 0.5 s.Figure 4A schematically illustrates the first feeding phase of the alternating feeding cycle, at time t1, i.e. at the end of the first feeding phase, just before switching to the second phase. The switching valve 101v is open, the switching valve 201v is closed and the purge valve 40 is closed. The feed fluid, typically hydrogen in the anode compartment, is injected at the inlet 11 of the group 10. It is partly consumed while passing through the group 10, then circulates along the discharge conduits 120, 220 to enter the group 20 via the outlet 22 of the group 20. It is partly or totally consumed within the group 20. During the reaction of the hydrogen from the anode compartment with the oxygen from the cathode compartment, reaction products, typically water, possibly carbon dioxide, are generated within the groups 10, 20.The majority of reactions take place in the cathode compartment, but reaction products are typically transported by diffusion or permeation to the anode compartment through the proton exchange membrane. Other non-reactive species, typically nitrogen, can also pass from the cathode compartment to the anode compartment by permeation through the proton exchange membrane. These reaction products and non-reactive species typically form a plug B, in which the hydrogen content is reduced, which here moves "along the x-axis" due to the injection of hydrogen and the consumption of hydrogen. Plug B is in fact both pushed by the injection of hydrogen in group 10 towards the end of group 20 which is not supplied, and pulled or sucked towards this end by vacuum, due to the disappearance of the reactive gases present in this zone.This phenomenon of suction of the plug towards the second group, linked to the consumption of reactants not compensated by an injection of reactant into the second group, intervenes significantly in its displacement. Conversely, the supply of reactant to the first group compensates this consumption of reactant on the side of the plug corresponding to the injection. The kinetics of displacement of plug B is therefore linked to the supply of new reactant gases at the inlet of the supplied group and to the disappearance of reactant gases in the zones where the reactions take place. This kinetics of displacement is therefore directly proportional to the current imposed on the fuel cell, which controls the reaction kinetics. At time t1, at the end of the first supply phase, plug B is located on the side of inlet 21 of group 20 which is not supplied, at a position P2 as illustrated in Figure 4A.Figure 4B schematically illustrates the second feeding phase of the alternating feeding cycle, at time t2, that is to say at the end of the second feeding phase and just before a new switchover. The switching valve 201v is open, the switching valve 101v is closed and the purge valve 40 is closed. The hydrogen is injected at the inlet 21 of the group 20. It is partly consumed while passing through the group 20, then circulates along the discharge conduits 220, 120 to enter the group 10 via the outlet 12 of the group 10. It is partly or totally consumed within the group 10. As before, the plug B moves under the influence of the injection and consumption of hydrogen. At time t2, at the end of the second supply phase, the plug B is located on the side of the input 11 of the group 10 which is not supplied, at a position P1 as illustrated in figure 4B.During the alternating feed cycles, the plug therefore oscillates between two positions P1, P2 located on the side of the inlets 11, 21 of the groups 10, 20. This oscillation occurs around an equilibrium position corresponding here substantially to the median position M of the midpoint. It is this particular oscillation dynamic which promotes the grouping of the products and non-reactive species into a moving agglomerate which is called here the plug. Indeed, a non-reactive gas particle will be sucked / displaced towards the side opposite the injection all the faster the further it is from it because there is a large reactive surface between the particle and this end. Similarly, a non-reactive particle located close to the end without injection will undergo a weak suction / displacement towards it, tending towards a zero speed at the level of the inlet which is not fed.This kinetic gradient favors the concentration of non-reactive species to form the plug, and the homogeneous movement of the latter around point M, rather than a mixture linked to the mixing of these species with reactive species. The speed of movement of plug B decreases in a substantially linear manner between the inlet of the group supplied with hydrogen and the inlet of the group which is not supplied. The maximum speed of movement of plug B is obtained when it is closest to the open supply valve, at the beginning of a supply phase, for example at t0, t1, t2 etc. The closer plug B approaches the non-supplied end (closed supply valve), the more its speed of movement decreases since the flow of hydrogen which carries away the non-reactive gases by suction is reduced. At the midpoint M, the speed of plug B is of the order of the arithmetic mean between the maximum speed and the minimum speed of the plug.The oscillation frequency of plug B follows the frequency 1 / T of the alternating feed cycles. The more the frequency 1 / T of the alternating feed cycles increases, the more the amplitude A of the oscillations of plug B decreases, and vice versa. The volume of plug B and / or the length LB of plug B increase with each alternating feed phase. It is thus necessary to evacuate plug B after a few alternating feed cycles, before the volume of plug B reaches 20% or 30% of the total volume of the anode compartment, for example after about ten alternating feed cycles. Figure 5 schematically illustrates a purge phase according to one embodiment of the operating method. The purge valve 40 is open. The switching valves 101v, 201v are open here. This promotes the evacuation of plug B by pushing symmetrically on each side at the inlets 11, 21.The opening of the purge valve 40 is triggered when the plug B is substantially at the middle position M, closest to the purge valve 40. Such an optimized triggering advantageously makes it possible to minimize the opening time, since the plug is directly in the right place, i.e. at the discharge. The plug B then crosses the smallest active surface section before reaching the exhaust. This also makes it possible to minimize the loss of unreacted hydrogen, since the latter is essentially located in the groups 10, 20 and is not trapped between the plug B and the purge valve 40. The opening time of the purge valve 40 is typically between 0.1 s and 10 s.The movement of the cap to a substantially middle position is carried out by applying a partial feeding phase, i.e. by following a feeding phase of one group with a feeding phase of the other group of substantially shorter duration. The optimal trigger delay – for which the cap B is at the middle position M – with respect to the last feeding phase can be determined from a model of the movement of the cap such as that described above. In particular, it appears that the cap B passes through the middle position M in the middle of each feeding phase, at approximately T / 4 after the start of a feeding phase. The precision on the trigger delay is preferably of the order of 10 ms. Figure 6 illustrates four delays d1, d25, d50, d75 corresponding respectively to a trigger at 1%, 25%, 50%, 75% of the duration of a power phase, i.e. at times t = 0.01.T / 2, t = 0.25.T / 2, t = 0.50.T / 2, t = 0.75.T / 2. Figures 7A to 7D illustrate the cell voltage V as a function of the operating phases C1, D1, E1, C2, E2, C3 of the battery, for the different triggering delays d1, d25, d50, d75. Phases C1, C2, C3 each correspond to operation of the battery during several alternating power supply cycles, typically around ten cycles in a row. Phase D1 corresponds to an optional phase called "dead-end" during which all the groups of the battery are powered simultaneously and the purge valve is closed. The plug during this phase is forced to stabilize at its position in the fluid circuit. In the case where the blockage is not already located in the evacuation ducts, it can be noted that the dead-end phase does not allow the blockage to be brought outside the cell group, and therefore in particular into the evacuation lines.Indeed, the plug is pushed by the supply and pulled by the consumption. However, the latter does not exist in the discharge line. This optional phase is therefore preferably carried out at the end of the triggering delay. In this case, the purge valve remains closed and all the switching valves are open. Thus, when the triggering delay corresponds to the optimal position of the plug in the fluid circuit, near the purge valve, the dead-end phase can produce a synergistic effect with the synchronized triggering to improve and stabilize the positioning of the plug as close as possible to the purge valve. Phase D1 can therefore be associated with the purge phase E1, typically by being inserted between the triggering delay and the actual opening of the purge valve. Phases E1, E2 correspond to purge phases during which the purge valve is open.When the purge valve is opened, the switching valves can be opened, simultaneously or previously. Phase E2 corresponds to a purge phase without prior dead-end. The purge phase is thus advantageously shortened. The performance of the battery is more constant. The average voltage delivered is higher than in the case of a purge phase with dead-end. According to one possibility, the operating method according to the invention provides only purge phases without dead-end. Several trigger times for the purge phases are presented in Figures 7A to 7D to illustrate the importance of choosing a correct start time for the purge step, the subject of the invention. Figure 7A illustrates a triggering of D1 + E1 (purge phase with dead end) and E2 (purge phase without dead end) at 1% of a supply phase.In this case, the voltage V drops at the beginning of the purge phase, typically when the plug is not yet brought close to the discharge. The hydrogen supply of the group of cells containing the plug is disadvantaged compared to the other group where the pressure drop is lower, which induces this drop in the voltage V delivered by the cells. Figure 7B illustrates a triggering of D1 + E1 (purge phase with dead end) and E2 (purge phase without dead end) at 25% of a supply phase. In this case, the drop in voltage V, still at the beginning of the purge phase, is more moderate than in the previous case, the plug still not being mainly positioned in the discharge ducts. Figure 7C illustrates a triggering of D1 + E1 (purge phase with dead end) and E2 (purge phase without dead end) at 50% of a supply phase.In this case, the drop in voltage V is minimized or even eliminated, especially when the trigger is not preceded by a dead-end. The plug is in an ideal position as close as possible to the purge valve and the hydrogen participating in the reaction is not evacuated during the purge. It continues to participate in the reaction. This contributes to maintaining the voltage V delivered by the cells. Figure 7D illustrates a trigger of D1 + E1 (purge phase with dead end) at 75% of a feeding phase. In this case, the drop in voltage V remains more moderate than in the case of triggers at 1% and 25%. This may be the consequence of a slight delay in the dynamics of the plug displacement, for example due to species with greater displacement inertia, such as liquid water for example.Furthermore, since the latter is retained in the discharge line when the plug passes, the hindrance caused by the latter for the circulation of the reagent in the group containing it is reduced. Typically, during a purge before T / 4, the plug is associated with liquid water. This causes a greater pressure drop. The supply of the two groups is unbalanced. The effect of the plug on the voltage is accentuated. During a purge after T / 4, the liquid water is less present with the plug, the supply imbalance is less strong and the effect of the plug on the voltage is reduced. According to a preferred possibility in practice, for a two-group assembly, the opening of the purge valve is triggered directly in the middle of a supply phase, without dead-end, after a delay corresponding to approximately 50% of the duration of the supply phase, i.e. at approximately T / 4.A certain tolerance with respect to T / 4 remains acceptable for triggering the opening of the purge valve. This triggering can, for example, be carried out after a delay of between 30% and 70% of the duration of the feeding phase. Other models can be implemented to determine the triggering delay of the purge phase, with or without dead end, after the start of a feeding cycle, in particular to take into account other parameters and / or variable conditions, for example concerning the sizing of the groups, the durations of the different feeding phases, a variation in the electrical intensity and / or hydrogen pressure, the increase in the size of the plug, etc. Simulations can be implemented on the basis of these models to refine the determination of the triggering delay if necessary. Charts can be established for a wider exploitation of different ping-pong architecture cells.An example model is described below. Regardless of the number of cell groups, during a given feeding phase, there are always two sets of groups: ^ The set ^. ^ in which ^ ^ group(s) receive the feed fluid directly at the inlet, ^ The assembly ^ ^ in which ^ ^ group(s) receive the feed fluid having passed through the assembly ^ ^, indirectly. In accordance with the principle of the "ping-pong" architecture, a collector connected to the purge member is in fluidic connection with all the groups of cells, therefore with the two sets, whatever their number of groups. The ping-pong process is divided into successive supply phases of these two sets placed in fluidic series. The flow in the stack is here modeled by a unidirectional flow along a dimensionless scale graduated between -1 and +1, the position 0 corresponding to the optimal position with respect to the purge member. The fluid entry is done alternately in ^ ^ = −1 and in ^ ^ = +1. The first set ^ ^ of cell groups is traversed in the interval [-10[, and the second set ^ ^ is traversed in the interval ]01]. We define ^, ^ ^and ^ as being respectively a molar flux, the total volume flow rate and the velocity of the gas mixture whatever its nature along its position x between -1 and 1. Their sign is positive if ^ ^ = −1, and negative if ^ ^ = +1. For each element ^^ of the flow, the material balance on hydrogen can be written: ^ ^^ ( ^ + ^^ ) = ^ ^^ ( ^ ) + ^^^ ( ^ ) Where ^ ^^ is a source term corresponding to hydrogen consumption. By integrating [Math 3], we obtain the molar flow rate at the abscissa ^: With The source term is: With ^ ( ^ ) the active surface of the cell where the reaction takes place, ^ ( ^ ) the current density and ℱ the Faraday constant. If the two sets ^ ^ and ^ ^ have a different number of cells, ^ ( ^ )will take a different value depending on whether ^ is positive or negative. This formulation therefore allows the problem to be generalized to any number of groups of cells. Thus, for ^ ^ = 1 with ^ and ^ constant we obtain: For ^ ^ = −1 with ^ and ^ constant we obtain: The total gas flow ^ ^^^ ( ^ ) contains water vapor: The total volume flow rate is given by: The stack has a volume accessible to the gas ^ ( ^ ) , with a possible distinction between negative and positive abscissas, as for ^ ( ^ ) . The flow length in a group set ^ ^ or ^ ^ being 1, the passage section perpendicular to the flow is ^(^). We deduce the speed as a function of the abscissa: In order to determine the best time for purging, from the velocity equation, we can calculate the position of a molecule in the flow using the equation: This differential equation can be solved numerically in the general case using standard simulation tools (e.g., Matlab Simulink software). Simulation allows us to predict the instant ^ for which ^ ( ^ ) = 0, corresponds to a passage of the molecule between the two sets ^ ^ and ^ ^of cell groups. If the molecule is harmful to the operation of the cell, this is when the purge should be opened. To control the cell, it will therefore be possible to implement a purge management algorithm integrating either this model directly or a table of simulation results in the operating conditions achievable by the cell. More complete models can be developed, also taking into account the passage of gas in non-active zones (in particular the purge collector), or the flow of liquid water. The latter case is more complex, but also more precise, because the purge is used to evacuate not only inert gases, but also liquid water, the presence of which harms the distribution of the feed fluid also called reactant (fuel or oxidant).Without making the model more complex, it is possible to adapt the version presented here so that it takes into account, on the one hand, additional volumes by increasing, for example, the value of ^(^), and on the other hand, a flow of liquid water by considering that it circulates with the gas at a speed lower than the latter, making it appear later in the collector. According to an example of implementation of the model, we consider that the two sets ^. ^ and ^ ^ each comprise a group of cells. They are identical, and are crossed by the same electric current. We also make the following approximations: ^ ( ^ ) = ^ ^ ( ^ ) = ^ ^ ^^^ ( ^ ) = ^ ^^^ (^) ^(^) = ^ ^ ( ^ ) = ^ ^(^) = ^ = ^^, with ^ the thickness of the frontal surface flow volume ^. It comes: ^ ^(^) is a piecewise constant function, which alternates between -1 and +1. The position as a function of time ^(^) must first be defined for a molecule starting from ^ ^ . We designate by ^ ^ the instant of hydrogen injection at the start of a supply phase of index ^. We can write for ^ greater than ^ ^ and excluding change of hydrogen entry point the following relationship: With This is a first-order linear differential equation whose solution is of the form: A being negative, we have for ^ tending towards infinity: Moreover, at ^ = ^ ^ : Let ^ = 2^ ^ (^ ^ ) In the end, it comes: When alternating between the two hydrogen inputs, the molecule does not have time to travel through the entirety of the two groups of cells (it converges to this point at infinity). The starting point for the ^ + 1 phase is noted ^^^ ^ ∗ ^^^ − ^ ^^^ ^ , with ^ ^ ^ ^^ the time of opening of the hydrogen inlet for the phase This valve should have opened for the molecule to arrive at this location from the entrance. We can then write that the starting point for phase ^ + 1 is the same as the stopping point for phase ^: ^^^ ^ ^ ^^ − ^ ^ ∗ ^ ^ ^ = ^^^ ^ ^ ^^ − ^ ^ ∗ ^ ^ 2^ ^ ^^ ^ ^ ^ ^ ^ ^^^ ^^^^ ^ ^ ^^ − ^ ^ ∗ ^ ^ ^^ − ^ ^ ^^ ^ ^ ^ ^ ^ = 2^ ^ ^^ ^ ^ ^ ^^^ ^^^^ ^ ^ ^^ − ^ ^ ∗ ^^ − ^ ^ ^^ ^ ^ ^ We note that ^ ^ ^^^ ^ ^ = −^ ^ ^^ ^ ^ ^^ ^, from which: The theoretical purge time is obtained for ^ = 0. The purge time of the supply phase ^ is noted ^ ^ ^ . It verifies: Moreover, in steady state, the difference between ^^ ∗ ^ and ^ ^ is constant and equal to ^. We also have the relation ^ ∗ ∗ ^ ^^ = ^ ^ + ^, with ^ = ^ / 2 the half-period of alternation between the power supply phases. We can write: ^ ^ ∗ 1 ^ ^^ − ^ ^^^ = ln ^1 − exp ^A^^ ^ ^ ^^ − ^ ∗ ^ ^^ + ^^^^ ^ ∗ ^ ∗ ^ By setting ^ = ^ ^^^ − ^ ^^^ = ^ ^ − ^ ^ , we obtain: exp(^^) = 1 − exp^^(^ + ^)^ = 1 − exp(^^) exp(^^) 1 1 ^ − = 1 exp ( ^^ ) exp ( ^^ ) 1 1 ^ ^ = ln ^ 1 + exp(^^) Finally, the purge trigger delay ^ ^ ^ ^ ^ = ^^ − ^ ^ is written: Numerical application: ^ = 210 ^ Pa ^ = 353 K ^ ^^^ = 510 ^Pa ^ = 1 A / cm² ^ = 350 cm² ^ = 0.3 mm (this value typically takes into account non-active volumes in the flow) ^ = 0.4 s (period T = 0.8 s) ^ = −3.38 s -1 The trigger delay obtained for these numerical values ​​is: ^ ^ ^ = 0.14 s, which corresponds to 34.3% of the half-period ^. Figure 8 illustrates the simulated evolution of the nitrogen concentration in the collector at the purge device (curve N) as a function of the feed phase changes. This model under Matlab Simulink makes it possible to visualize the spread of the plug, mainly formed of nitrogen. In this particular case, the moment when the nitrogen concentration is maximum in front of the purge device (local maximum) occurs well before half of the half-period, as in the previous analytical calculation. The optimum synchronized triggering of the opening around 50% of the half-period observed in practice is probably attributable to liquid water. We will therefore ensure that: This equation represents a purge trigger interval according to the previous analytical calculation. In practice, there is a margin of uncertainty on the local concentrations of the species, their behavior in the configuration considered, and the spread of the plug that we are trying to evacuate. The previous inequality characterizing the purge trigger time must therefore be understood in a context of + / - 20% uncertainty, and the purge trigger time consequently becomes constrained by: ^ ^ ^^^^^ ( ^^ ^ ^ ^^^^^ ^^ ^ ≥ (1 − 20%) ln ) ( )^ ^ ^ or ^^ ≥ 0.8. ^ ln ^ Through the examples described above, it is clear that the operating method according to the invention makes it possible to improve the consumption and performance of a fuel cell with ping-pong architecture. This method can be advantageously implemented for automotive or heavy transport applications (land, sea, air, etc.). The invention is not limited to the embodiments described above. Several cells, with groups of nested cells, can also be envisaged within the scope of the present invention, by implementing the principle of synchronized triggering of the purge phase.

Claims

CLAIMS 1. Method for operating a fuel cell (1), said fuel cell (1) comprising a plurality of groups (10, 20) of electrochemical cells connected together by a fluid circuit (3, 4), said fluid circuit (3, 4) comprising, for each group, a supply line (100, 200) at the inlet of said group, and a switching member (101v, 201v) on said supply line, and, a discharge line (120, 220) at the outlet of said group, said fluid circuit being configured so that the discharge lines (120, 220) communicate fluidically with each other and are connected to a common purge member (40), Said method comprising at least: ^ A cycle (C1, C2) of alternating supplies comprising at least: o A first supply phase in which a first group (10) among the plurality of groups is supplied directly with fluid power supply, at its input (11),the switching member (101v) of said first group (10) being on while the switching members (201v) of the other groups (20) of the plurality of groups are blocked, said other groups (20) being supplied indirectly, during this first supply phase, at their respective outlets (22), by the supply fluid coming from the outlet (12) of the first group (10), o A second supply phase in which a second group (20) different from the first group (10) is supplied directly with supply fluid, at its inlet (21), the switching member (201v) of said second group (20) being on while the switching members (101v) of the other groups (10) of the plurality of groups are blocked, said other groups (10) being supplied indirectly, during this second supply phase, at their respective outlets (12),by the supply fluid from the outlet (22) of the second group (20), ^ A purge phase (E1, E2) following the cycle of alternating supplies, said purge phase comprising an opening of the member of, purge (40) so as to evacuate non-reactive fluids originating at least in part from the supply fluid, said method being characterized in that the purge phase is triggered in a synchronized manner with a movement of the non-reactive fluids accumulated in the form of an aggregate called a plug (B), when said plug (B) is located at least in part in the evacuation lines (120, 220), the alternating supply cycle having a period T and the purge phase being triggered during a supply phase ^ and before the end of the supply phase ^, after a delay ^ ^ ^ such that: ^ ^ ^ ≥ 0.

8. with ^ = ^ / 2 the half-period, and ^^^^ ^ = − 2ℱ^ ( ^ − ^ ^^^ )With ^ the active surface of the battery, ^ the current density and ℱ the Faraday constant, ^ the ideal gas constant, ^ the temperature of the battery, ^ the volume crossed by the fluid in the battery, ^ the pressure of the fluid in the battery and ^ ^^^the saturated vapor pressure at the cell temperature.

2. Method according to the preceding claim in which the alternating feed cycle has a period T and the purge phase is triggered during a feed phase ^ and before the end of the feed phase ^, after a delay ^ ^ between 0.

2. T / 2 ≤ ^^ ≤ 0.

7. T / 2.

3. Method according to any one of the preceding claims in which the alternating feed cycle has a period T and the purge phase is triggered during a feed phase ^ and before the end of the feed phase ^, after a delay ^ ^ ^ = T / 4.

4. Method according to any one of the preceding claims in which the stack comprises only two groups (10, 20), the outlets (12, 22) of these two groups (10, 20) are located at equal distance from the purge member (40), and the first and second supply phases of the alternating supply cycle have the same duration T / 2. 5.Method according to any one of the preceding claims in which the purging phase comprises, before or during the opening of the purging member (40), an input supply (11, 21) of all the groups (10, 20), each switching member (101v, 201v) being conductive.

6. Method according to the preceding claim in which the inlet supply of all the groups is carried out simultaneously with the opening of the purge member (40).

7. Method according to claim 5 in which the opening of the purge member (40) is carried out after the inlet supply of all the groups, after a so-called "dead-end" delay.

8. Method according to any one of claims 1 to 4 in which the purge phase does not include inlet supply of all the groups before the opening of the purge member (40).

9. Method according to any one of the preceding claims in which a constant electric current is imposed on the fuel cell.

10. Method according to any one of the preceding claims in which the supply fluid injected during the alternating supply cycle has a constant pressure. 11.Method according to any one of the preceding claims in which the opening of the purge member (40) is triggered when at least 70%, and preferably at least 80%, of the length LB of the plug (B) is located in the discharge lines (120, 220), the length LB of the plug (B) being measured in the direction in which the discharge line extends at the point in question.

12. Method according to any one of the preceding claims in which the opening of the purge member (40) is triggered when the plug (B) is entirely located in the discharge lines (120, 220).

13. Method according to any one of the preceding claims in which the opening of the purge member (40) is triggered when the plug (B) entirely occupies the discharge lines (120, 220).

Images