Fuel cell system and method for operating a fuel cell system

Abstract

The invention relates to a method (100) for operating a fuel cell system (200), said method (100) involving: - determining (101) a hydrogen mass flow supplied to the fuel cell system (200); - adjusting (103) the fuel cell system (200) depending on the determined hydrogen mass flow, wherein the hydrogen mass flow supplied is determined continuously during operation of the fuel cell system (200) on the basis of a hydrogen concentration value, determined by a machine learning model, in a fuel mass flow supplied to the fuel cell system (200).

Description

[0001] Description and methods for operating a tems

[0002] The presented invention relates to a method for operating a fuel cell system, a fuel cell system and a program product, according to the attached claims.

[0003] State of the art

[0004] In polymer electrolyte membrane fuel cells (PEM fuel cells), which are powered by hydrogen and air, a proton-conducting membrane separates the gases of two galvanic half-cells from each other.

[0005] In this system, one anode side is filled with hydrogen or fuel, and one cathode side is filled with air. Due to concentration differences and a membrane that is not completely gas-tight, gases pass between the anode and cathode sides.

[0006] In particular, the accumulation of nitrogen on the anode side from the air on the cathode side causes the hydrogen concentration on the anode side to continuously decrease during operation. Besides reducing the voltage and thus the efficiency of the respective fuel cells, this also leads to the risk of hydrogen depletion, which can cause irreversible damage to the fuel cells.

[0007] Hydrogen concentration sensors are generally not used in mobile PEM fuel cell systems. The flow rate in a hydrogen metering valve for injecting fuel into an anode subsystem can be calculated using a nozzle equation, assuming a value for the open cross-section and a fuel density. However, the exact valve position of electromagnetically actuated valve plungers, as used in hydrogen metering valves, is subject to hysteresis, making the position estimation quite inaccurate. Furthermore, fluctuating fuel qualities can also lead to uncertainties in the density calculation.

[0008] Disclosure of the invention

[0009] Within the scope of the presented invention, a method for operating a fuel cell system, a fuel cell system, and a software product are introduced. Further features and details of the invention will become apparent from the respective dependent claims, the description, and the drawings. Features and details described in connection with the fuel cell system according to the invention naturally also apply in connection with the method and the software product according to the invention, and vice versa, so that the disclosure of the individual aspects of the invention is always, or can always be, mutually referenced.

[0010] The presented invention serves in particular to provide a possibility for the robust operation of a fuel cell system.

[0011] Thus, according to a first aspect of the presented invention, a method for operating a fuel cell system is presented.

[0012] The presented method comprises determining the hydrogen mass flow rate supplied to the fuel cell system and adjusting the fuel cell system based on this determined hydrogen mass flow rate. The determination of the supplied hydrogen mass flow rate is performed continuously during operation of the fuel cell system based on a hydrogen concentration value in the fuel mass flow supplied to the fuel cell system, determined by a machine learner. In the context of the presented invention, a machine learner is understood to be an algorithm that has been trained to make decisions. For example, a machine learner or learning algorithm can be an artificial neural network or a support vector machine.

[0013] In the context of the presented invention, a fuel mass flow is understood to be an anode gas mixture or a fluid flow flowing in an anode subsystem of a respective fuel cell system, i.e., a total anode mass flow.

[0014] The presented invention is based on the principle that, by means of the machine learner provided according to the invention, a hydrogen concentration supplied to the fuel cell system in a fuel supplied to the fuel cell system or in a fluid flowing on an anode side, i.e., in an anode subsystem of the fuel cell system, is continuously determined, in particular over an entire operating range, during operation of the fuel cell system.

[0015] Based on the determined hydrogen concentration, the fuel cell system can be adjusted particularly precisely to the actual state of a fuel cell stack of the fuel cell system, so that an electrochemical reaction taking place in the fuel cell stack can be controlled, i.e., steered or regulated with particular precision.

[0016] By continuously, i.e., on an ongoing basis, determining the hydrogen concentration, a transition of nitrogen from the cathode subsystem to the anode subsystem is detected, among other things, or logically represented by the machine learner.

[0017] To logically model the processes in the fuel cell system, the machine learner can be trained, for example, using a fundamental truth. This fundamental truth can be provided, for instance, by a training fuel cell system with a hydrogen concentration sensor in the anode subsystem, so that the machine learner adapts its internal logic or mathematical model during training until the respective input values ​​result in output values ​​that correspond to the fundamental truth.

[0018] Accordingly, the presented invention enables a data-based calculation of the total density and mixture viscosity of an anode gas mixture, so that a total anode mass flow rate or fuel mass flow rate can be determined via the anode subsystem using, for example, the throttling equation, by taking into account, for example, a pressure difference between an anode inlet and an anode outlet, an electrical current consumed by a cathode blower and / or a pressure difference between a point before a jet pump and a point after the jet pump.

[0019] Based on the determined hydrogen concentration or fuel mass flow rate, the flow rate of a hydrogen metering valve can be validated, or a diagnosis of the hydrogen metering valve can be performed. Alternatively or additionally, the determined hydrogen concentration or fuel mass flow rate can be used to quantify permeation losses, i.e., to determine the age of the fuel cell system membrane. Accordingly, the presented method enables optimized efficiency and consumption determination, and a correspondingly more robust operation of the fuel cell system, since, for example, fluctuating fuel quality or increasing permeation losses can be taken into account when selecting a fuel cell system purging strategy.

[0020] It may be possible to infer the composition of the fuel mass flow based on the hydrogen concentration determined by the machine learner.

[0021] Contamination at a filling station or during hydrogen production can lead to impurities in the fuel supplied to a fuel cell system. This distorts the fuel balance and, in the worst case, can result in fuel depletion and irreversible damage to the fuel cells. To detect these impurities, the hydrogen concentration profile in the exhaust system during the first anode flush after refueling can be analyzed.

[0022] During an anode flush, the anode subsystem is purged with fresh hydrogen while the cathode subsystem remains closed. This means the air system operates in bypass mode to dilute the purge gas discharged through the purge valve. The purge valve in the anode subsystem is opened continuously or with a high duty cycle, allowing a large volume of gas to flow through the anode subsystem and raising the hydrogen concentration to a sufficiently high level as quickly as possible. Since no electrical current is yet being drawn, there is no consumption in the fuel cell system; that is, the supplied gas flows unchanged through the fuel cell system after initial dilution and exits the anode subsystem via the purge valve. Depending on the actual hydrogen concentration in the supplied fuel, the speed of sound in the narrowest cross-section of the purge valve changes, as described above.Accordingly, the composition of the fuel can be determined as follows:

[0023] 1. Initiating an anode flush, ensuring that the differential pressure between a point before the flushing valve and a point after the flushing valve is > 0.528 and that a sufficiently high exhaust gas mass flow is ensured by operating the air system in bypass mode.

[0024] 2. Open the purge valve, wait until the variance of a hydrogen concentration curve in the exhaust tract falls below a predetermined threshold.

[0025] 3. Evaluating a maximum of the trend relative to a given reference.

[0026] If the maximum value is lower than the reference value, this indicates an increased inert component or impurity level in the fuel, and therefore the fuel mass flow rate supplied to the fuel cell system must be adjusted accordingly during operation. Furthermore, the machine learner may be trained to assign a hydrogen concentration value in the fuel mass flow to at least the following values: pressure difference between an anode inlet and an anode outlet of the fuel cell system, power consumption of a fan in the fuel cell system, and / or pressure difference between a point upstream of a jet pump in the fuel cell system and a point downstream of the jet pump.

[0027] While an increased transfer of gases through membranes (crossover) is primarily noticeable at the purge valve and thus downstream of the anode subsystem, a drift of the hydrogen metering valve mainly affects the operation of a jet pump and a pressure drop in the anode subsystem. Therefore, the machine learner can be provided with input values ​​such as the differential pressure between an anode inlet and an anode outlet, as well as a pressure increase caused by the jet pump.

[0028] To calculate the pressure loss, the following dependence on the viscosity of the gas mixture according to equation (1) can be used: jö'Ap M~ ] (1)

[0029] Where Ap is the pressure drop between the anode inlet and the anode outlet, p is the density of a gas mixture flowing in the anode subsystem, and p is the dynamic viscosity of the gas mixture. If the hydrogen mass flow rate supplied to the anode subsystem is coupled to, for example, a generated electric current or a cathode pressure via feedforward control, this value remains essentially constant over the lifetime of a fuel cell system. However, a change in the supplied hydrogen mass flow rate, for example, due to drift in a characteristic curve, leads to changes in the mixture density and viscosity upstream of the anode inlet or in the jet pump, which can alter both pressure drops.

[0030] It can also be provided that, based on the determined hydrogen concentration, a hydrogen mass flow rate discharged through a purge valve of the fuel cell system is calculated. Using the nozzle equation, a hydrogen mass flow rate discharged through a purge valve, i.e., a purge / drain valve, can be determined for a known hydrogen concentration in the fuel mass flow.

[0031] It may also be provided that the hydrogen mass flow discharged through the purge valve of the fuel cell system is determined using a control space that includes the following parameters: hydrogen mass flow supplied to the fuel cell system, hydrogen concentration in the fuel mass flow supplied to the fuel cell system, hydrogen mass flow consumed by the fuel cell system, hydrogen mass flow discharged through permeation losses, mass flow of liquid water introduced through permeation, and mass flow of nitrogen introduced through permeation.It is provided that the hydrogen mass flow rate discharged through permeation losses, the liquid water mass flow rate introduced through permeation, and the nitrogen mass flow rate introduced through permeation are determined by means of a further machine learner, which is trained using a training fuel cell system, wherein the training fuel cell system comprises a fuel cell stack, an exhaust line, a fuel line with an anode recirculation circuit, at least one valve line connected to the anode recirculation circuit, and a mass spectrometer, and wherein the machine learner is provided with a hydrogen concentration and / or nitrogen concentration determined by the mass spectrometer in an anode subsystem of the training fuel cell system as a basic truth during training.

[0032] In particular, it may be further provided that the additional machine learner is trained with different compositions of fuel mass flow supplied to the fuel cell system and the same exhaust gas mass flows, whereby a purge valve of the training fuel cell system is always flowed through with a critical mass flow during the training.

[0033] Training of the second machine learner can, for example, be performed using a gas composition in the anode subsystem of the training fuel cell system, determined by a mass spectrometer of the training system, as a baseline. This allows the second machine learner to create a data-driven density and viscosity model. During training, possible operating scenarios of the application fuel cell system, such as drift of the hydrogen dosing valve or increased nitrogen transfer through the membrane, can be simulated, enabling precise conclusions to be drawn about the respective effects by evaluating the results.

[0034] Furthermore, a hybrid model for hydrogen balancing and for calculating the total anode mass flow can be formed to enable stoichiometry-controlled operation of the fuel cell system.

[0035] The amount of hydrogen currently consumed is directly available when operating a fuel cell system, as it is directly coupled to the electrical current currently supplied by the fuel cell system via Faraday's equation. In contrast, the permeation flows of hydrogen, nitrogen, and gaseous water cannot be directly calculated and are not constant, particularly across the operating range or the aging state of the fuel cell system. However, these can be continuously determined by a machine learner and used to adjust the fuel cell system.

[0036] For example, by knowing the hydrogen mass flow rate discharged through the purge valve of the fuel cell system, as well as the valve's opening duration and / or opening frequency, a purge strategy can be optimized. The purge strategy can then be adjusted based on the determined hydrogen mass flow rate discharged through the fuel cell system's purge valve, such that the determined hydrogen mass flow rate corresponds to a predefined value. Accordingly, the purge strategy can be chosen relative to a value that closely reflects the actual operating state of the fuel cell system, thus eliminating the need for a buffer or a conservative over-design of the purge strategy. This saves fuel and maximizes the fuel cell system's lifespan.

[0037] Increased crossover flows of nitrogen and, if applicable, gaseous water or hydrogen from the anode subsystem to the cathode subsystem alter the gas composition at the purge valve. This is accompanied by a change in the speed of sound, particularly if the hydrogen content changes significantly or a two-phase mixture formed by liquid water is passed through the purge valve, causing a significant change in the critical mass flow rate, i.e., the maximum mass flow rate that can pass through the purge valve. Due to the altered flow characteristics at the purge valve, provided critical conditions (i.e., p1 / p0 > 0.528) are met, the mass flow rates passing through the purge valve will change considerably. Therefore, it is expected that the rising edges of a peak in the hydrogen concentration curve in the exhaust tract of the fuel cell system will differ depending on the gas composition at the start of a purge process.Accordingly, by training a further machine learner with different anode gas compositions and the same exhaust gas mass flow rates, a gas composition at the purge valve can be inferred. During training, it must be ensured that the purge valve is critically flowed through in each case.

[0038] To determine the anode gas composition downstream of a fuel cell system during operation, a comparison of the hydrogen concentration peak in the exhaust tract of the fuel cell system with a predefined reference can be performed once it has been ensured that the purge valve is operating at critical flow. A deviation between the determined profile and the reference can lead to various measures. For example, the anode stoichiometry can be increased to raise the hydrogen partial pressure at the anode outlet. Alternatively, the overall pressure level in the anode subsystem can also be increased to raise the hydrogen partial pressure.

[0039] Accordingly, it may be provided that during the operation of the fuel cell system, a value of a pressure drop generated by a purging process is compared with a predetermined threshold value, and if the pressure drop is greater than the threshold value, a hydrogen concentration profile measured in the exhaust gas tract is compared with a reference profile, and if the measured profile differs from the reference profile, an anode stoichiometry is increased or an anode pressure is increased.

[0040] It may also be possible to compare the determined hydrogen concentration with a flow rate of a hydrogen metering valve of the fuel cell system in order to determine the state of the hydrogen metering valve.

[0041] By comparing the determined hydrogen concentration with the flow rate of the hydrogen metering valve, a diagnosis of the hydrogen metering valve can be made, so that, for example, a fault can be detected or correct function can be validated.

[0042] It may also be provided that, based on a pressure loss in the anode subsystem and a pressure increase caused by a jet pump of the fuel cell system, a conclusion can be drawn about the state of the hydrogen metering valve, taking into account the composition of the fuel supplied to the fuel cell system.

[0043] According to a second aspect, the presented invention relates to a fuel cell system for converting energy.

[0044] The presented fuel cell system comprises a fuel cell stack, a hydrogen metering valve for supplying fuel to an anode subsystem of the fuel cell stack, a first pressure sensor in an anode recirculation circuit upstream of an inlet to the fuel cell stack, a second pressure sensor in the anode recirculation circuit downstream of the fuel cell stack, a hydrogen concentration sensor in the exhaust tract of the fuel cell system, and a computing unit, wherein the computing unit is configured to carry out a possible embodiment of the presented method.

[0045] According to a third aspect, the presented invention relates to a program product, wherein the program product comprises program code means which, when executed on a computing unit, configure the computing unit to execute a possible embodiment of the presented method.

[0046] Advantages described in detail for the method of operating a fuel cell system according to the first aspect of the invention apply equally to the fuel cell system for converting energy according to the second aspect of the invention and to the program product according to the third aspect of the invention, and vice versa.

[0047] Further advantages, features, and details of the invention will become apparent from the following description, in which exemplary embodiments of the invention are described in detail with reference to the drawings. The features mentioned in the claims and in the description can each be essential to the invention individually or in any combination.

[0048] They each show schematically:

[0049] Figure 1 shows a possible embodiment of the presented method, and

[0050] Figure 2 shows a representation of a possible configuration of the presented fuel cell system.

[0051] Fig. 1 shows a method (100) for operating a fuel cell system.

[0052] Method 100 comprises a determination step 101, in which a hydrogen mass flow rate supplied to the fuel cell system is determined, and a setting step 103, in which the fuel cell system is set as a function of the determined hydrogen mass flow rate. The determination of the supplied hydrogen mass flow rate is carried out continuously during operation of the fuel cell system based on a hydrogen concentration value in a fuel mass flow supplied to the fuel cell system, determined by a machine learner. Figure 2 shows a fuel cell system 200 for energy conversion.

[0053] The fuel cell system 200 comprises a fuel cell stack 201, a hydrogen metering valve 203 for supplying hydrogen-containing fuel to an anode subsystem 205 of the fuel cell system 200, a first pressure sensor 207 in an anode recirculation circuit 209 upstream of an inlet to the fuel cell stack 201, a second pressure sensor 211 in the anode recirculation circuit 209 downstream of the fuel cell stack 201, a hydrogen concentration sensor 213 in the exhaust tract of the fuel cell system 200, and a computing unit 215, wherein the computing unit 215 is configured to perform the method 100 according to Fig. 1.

[0054] To determine the hydrogen mass flow rate metered into the anode subsystem 205 via the hydrogen metering valve 203, the machine learner is continuously and synchronously evaluated by, for example, first values ​​determined by the first pressure sensor 207 and second values ​​determined by the second pressure sensor 211. This means that the machine learner receives input values, particularly in combination with other values ​​such as an electrical current supplied by the fuel cell stack 201, an electrical current drawn by a recirculation blower 217, a differential pressure across a jet pump 219, and / or values ​​determined by a hydrogen concentration sensor 221 in the exhaust tract 223. In this way, a delta of hydrogen concentration during flow through the anode subsystem 205 can be determined across the entire operating range of the fuel cell system 200.

[0055] The determined hydrogen mass flow through the anode subsystem 205 can be used, for example, to set an activation interval or activation duration of a purge valve or purge / drain valve 225, so that the purge valve or purge / drain valve 225 is activated according to an actual state in the anode subsystem 205 and no conservative design buffer needs to be maintained during activation.

Claims

Claims 1. Method (100) for operating a fuel cell system (200), wherein the method (100) comprises: Determine (101) a hydrogen mass flow supplied to the fuel cell system (200), Adjustment (103) of the fuel cell system (200) depending on the determined hydrogen mass flow rate, wherein the determination of the supplied hydrogen mass flow rate is carried out continuously during operation of the fuel cell system (200) based on a value of a hydrogen concentration in a fuel mass flow rate supplied to the fuel cell system (200) determined by a machine learner.

2. Method (100) according to claim 1 , characterized in that the composition of the fuel mass flow is inferred from the hydrogen concentration determined by the machine learner.

3. Method (100) according to claim 1 or 2, characterized in that the machine learner is trained to assign at least to values ​​from the following list of values ​​a value of a hydrogen concentration in the fuel mass flow: Pressure difference between an anode inlet and an anode outlet of the fuel cell system (200), Power consumption of a blower of the fuel cell system (200), pressure difference between a point before a jet pump (219) of the fuel cell system (200) and a point after the jet pump (219).

4. Method (100) according to one of the preceding claims, characterized in that a hydrogen mass flow rate discharged through a purge valve (225) of the fuel cell system (200) is determined on the basis of the determined hydrogen concentration.

5. Method (100) according to claim 4, characterized in that the hydrogen mass flow rate discharged through the purge valve (225) of the fuel cell system (200) is determined using a control volume comprising the following parameters: fuel mass flow rate supplied to the fuel cell system (200), hydrogen concentration in the fuel mass flow rate supplied to the fuel cell system (200), hydrogen mass flow rate consumed by the fuel cell system (200), hydrogen mass flow rate discharged through permeation losses, mass flow rate of liquid water introduced through permeation, mass flow rate of nitrogen introduced through permeation, wherein the hydrogen mass flow rate discharged through permeation losses, the mass flow rate of liquid water introduced through permeation, and the mass flow rate of nitrogen introduced through permeation are determined by means of a further machine learner that is trained using a training fuel cell system.wherein the training fuel cell system comprises a fuel cell stack, an exhaust line, a fuel line with an anode recirculation circuit, at least one valve line connected to the anode recirculation circuit, and a mass spectrometer, and wherein the machine learner is provided, during training, with a hydrogen concentration and / or nitrogen concentration determined by the mass spectrometer in an anode subsystem of the training fuel cell system as a basic truth.

6. Method (100) according to claim 5, characterized in that that the further machine learner is trained with different compositions of fuel mass flow supplied to the fuel cell system (200) and the same exhaust gas mass flows, wherein a purge valve of the training fuel cell system is always supplied with a critical mass flow during the training.

7. Method (100) according to one of the preceding claims, characterized in that during operation of the fuel cell system (200) a value of a pressure drop generated by a purging process is compared with a predetermined threshold value and, in the event that the pressure drop is greater than the threshold value, a hydrogen concentration profile measured in the exhaust gas tract is compared with a reference profile and, in the event that the measured profile differs from the reference profile, an anode stoichiometry is increased or an anode pressure is increased.

8. Method (100) according to one of the preceding claims, characterized in that the determined hydrogen concentration is compared with a flow rate of a hydrogen metering valve of the fuel cell system (200) in order to determine a state of the hydrogen metering valve.

9. Method (100) according to one of the preceding claims, characterized in that a state of the hydrogen metering valve is inferred from a pressure loss in the anode subsystem (205) and a pressure increase caused by a jet pump of the fuel cell system (200), taking into account a composition of the fuel supplied to the fuel cell system (200).

10. Fuel cell system (200) for converting energy, wherein the fuel cell system (200) comprises: a fuel cell stack (201), a hydrogen metering valve (203) for supplying fuel to an anode subsystem (205) of the fuel cell system (200), a first pressure sensor (207) in an anode recirculation circuit (209) upstream of an inlet to the fuel cell stack (201), - a second pressure sensor (211) in the anode recirculation circuit (209) after the fuel cell stack (201), a hydrogen concentration sensor (213) in the exhaust tract of the fuel cell system (200), a computing unit (215), wherein the computing unit (215) is configured to perform a method (100) according to any one of claims 1 to 9.

11. Program product, wherein the program product comprises program code means which, when executed on a computing unit, configure the computing unit to execute a method (100) according to any one of claims 1 to 9.

Images