Method and system for diagnosing and estimating health state of fuel cell based on polarization characteristics

Abstract

The invention provides a method and a system for diagnosing and estimating the state of health of a fuel cell based on polarization characteristics, and relates to the technical field of fuel cells, and the method comprises the following steps: constructing a proton exchange membrane fuel cell model based on a multi-condition multi-physical field, extracting health state characteristics of the to-be-diagnosed battery from the proton exchange membrane fuel cell model under a preset typical working condition; and inputting the health state characteristics into a trained fuel cell health state diagnosis model to obtain the health state of the to-be-diagnosed battery evaluated by the fuel cell health state diagnosis model. The method not only can accurately diagnose the health state of the fuel cell, but also is low in implementation difficulty and cost and relatively high in interpretability, so that the dependence of high-precision diagnosis of the health state of the fuel cell on expert experience is reduced.

Description

Technical Field

[0001] The present invention relates to the field of fuel cell technology, and in particular to a method and system for diagnosing and estimating the health status of a fuel cell based on polarization characteristics. Background Art

[0002] Due to the advantages of high energy efficiency and zero emission of proton exchange membrane fuel cells (PEMFC), PEMFC systems have become an indispensable part of the future green energy structure. However, the PEMFC system is complex, in which the stack is coupled by multiple physical fields of gas, water, electricity, and heat. External auxiliary equipment such as air compressors, flow controllers, and humidifiers are difficult to coordinate and control. The abnormal health status is numerous and the sources are complex, making it difficult to accurately locate and evaluate it. In order to enhance the operational reliability and safety of the PEMFC system, low-cost, high-performance fuel cell health status diagnostic technology is necessary.

[0003] At present, the methods for fuel cell health status diagnosis are mainly divided into two types: physical model-based and data-driven. The method based on the mechanism model constructs a numerical analysis model that includes multiple mechanisms such as electrochemistry, heat transfer and porous medium two-phase flow, and relies on the residual between the predicted value and the measured value to detect the health status. However, the mechanism model usually only considers the physical fields of gas, water, electricity and heat, and can only realize the health status diagnosis of local physical fields. In addition, the strong coupling effect between them will also significantly affect the reliability of diagnosis, making the diagnosis result dependent on expert experience. The data-driven diagnosis method needs to first extract the health status characteristics of the fuel cell, which are generally derived from the global time series voltage, pressure and temperature. In addition, frequency domain features such as electrochemical impedance spectroscopy (EIS) and distribution relaxation time (DRT) are also used for health status feature extraction. After that, the health status features are classified with the help of machine learning algorithms such as least squares support vector machine, XGBoost, or deep learning models such as self-attention and bidirectional gated recurrent unit. However, data-driven diagnostic methods are heavily dependent on data quality, and complex prediction algorithms are difficult to deploy on hardware devices. Therefore, when using such methods to implement health status diagnosis, there is the problem of how to extract high-information-dimensional health status features.

[0004] Therefore, how to provide a fuel cell health status diagnosis method with high accuracy, low implementation cost and no reliance on expert experience is an important issue that needs to be urgently addressed in the industry. Summary of the invention

[0005] In view of this, the present invention proposes a fuel cell health status diagnosis and estimation method and system based on polarization characteristics to solve the problem that the diagnosis accuracy of the current fuel cell health status diagnosis method depends on expert experience and has high implementation cost.

[0006] The technical solution of the present invention is achieved in this way:

[0007] According to a first aspect, an embodiment of the present invention provides a method for diagnosing and estimating a fuel cell health status based on polarization characteristics, the method comprising:

[0008] A proton exchange membrane fuel cell model is constructed based on a multi-physical field with multiple conditions, and the health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions;

[0009] Inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model;

[0010] The fuel cell health status diagnosis model is trained by the following steps:

[0011] Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery;

[0012] Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum;

[0013] A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of a sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature of the sample is used as a fault state feature;

[0014] Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature;

[0015] The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

[0016] In combination with the first aspect, in a first implementation of the first aspect, the feature fusion of the sample high-frequency impedance feature and the sample polarization resistance feature to obtain the sample fusion feature specifically includes:

[0017] Sort the obtained sample polarization resistance characteristics and the sample high-frequency impedance characteristics in ascending order according to the magnitude of the frequency;

[0018] The sorted sample polarization resistance characteristics and the sample high-frequency impedance characteristics are sequentially connected to obtain a sample fusion characteristic.

[0019] In combination with the first aspect, in a second implementation of the first aspect, the sample polarization health status characteristic includes a sample voltage of the fuel cell, and the calculation formula of the sample voltage is:

[0020] E=U0-η act -η ohm -η con

[0021] Where, E represents the sample voltage; η act represents the sample activation loss; η ohm represents the sample ohmic loss; η con represents the sample concentration loss; U0 represents the open circuit voltage.

[0022] In combination with the second embodiment of the first aspect, in the third embodiment of the first aspect, adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum specifically includes:

[0023] Adding a preset sinusoidal excitation voltage to the sample voltage to obtain a sinusoidal excitation current;

[0024] Obtaining an electrochemical impedance spectrum of the sample according to a sinusoidal excitation voltage and a sinusoidal excitation current;

[0025] According to the preset adaptive weight factor, the high-frequency impedance characteristics of the sample corresponding to each frequency band are extracted from the electrochemical impedance spectrum of the sample.

[0026] In combination with the third embodiment of the first aspect, in the fourth embodiment of the first aspect, the adaptive weight factor is obtained by minimizing the deviation of a preset number of the sample high-frequency impedance characteristics and a preset number of sample polarization resistance characteristics corresponding to the sample high-frequency impedance characteristics.

[0027] In combination with the first aspect, in a fifth embodiment of the first aspect, the radial basis function is used as a discrete basis, and the relaxation time distribution of the sample electrochemical impedance spectrum is extracted based on the radial basis function, and the sample polarization resistance feature characterized by the relaxation time distribution is extracted, and the polarization resistance feature is used as a fault state feature, specifically including:

[0028] Fitting the sample battery to a parallel array of several resistors and an infinite number of resistors and capacitors to obtain a non-negative distribution function describing the time relaxation characteristics of the electrochemical system of the sample battery, and converting the non-negative distribution function to obtain a logarithmic term for adapting the frequency of the electrochemical impedance spectrum;

[0029] Using radial basis function as a discrete basis, and discretizing the logarithmic term based on the radial basis function to obtain a relaxation time distribution extraction model;

[0030] Fitting the sample electrochemical impedance spectrum using the relaxation time distribution extraction model to extract the relaxation time distribution of the sample electrochemical impedance spectrum;

[0031] Each internal resistance polarization process of the sample battery is characterized as a corresponding relaxation time distribution peak, the sample polarization resistance characteristics characterized by the relaxation time distribution are extracted, and the area of ​​each relaxation time distribution peak is used as a corresponding fault state feature.

[0032] In combination with the first aspect, in a sixth implementation of the first aspect, the using of a least squares support vector machine to classify the sample fusion features and train the fuel cell health status diagnosis model specifically includes:

[0033] A least squares support vector machine is constructed using a Gaussian function, and an initial fuel cell health status diagnosis model is constructed based on the least squares support vector machine;

[0034] The sample fusion features are classified using a fuel cell health status diagnosis model, and the parameters of the fuel cell health status diagnosis model are adjusted using a Lagrangian function.

[0035] In combination with the sixth implementation of the first aspect, in the seventh implementation of the first aspect, the sample fusion features are classified by using the fuel cell health status diagnosis model, and the parameters of the fuel cell health status diagnosis model are adjusted by using the Lagrangian function, specifically including:

[0036] A least squares support vector machine is constructed based on a preset Gaussian function to obtain an initial fuel cell health status diagnosis model;

[0037] Using an initial fuel cell health status diagnosis model to classify the sample fusion features, and obtain several groups of health status;

[0038] The Lagrangian function of the initial fuel cell health status diagnostic model is constructed, and the bias and weight of the initial fuel cell health status diagnostic model are adjusted using the Lagrangian function until the prediction error of the health status is less than a preset value, thereby obtaining a trained fuel cell health status diagnostic model.

[0039] In combination with the sixth implementation of the first aspect, in an eighth implementation of the first aspect, the expression of the fuel cell health status diagnosis model is:

[0040]

[0041] Among them, y(x) represents the assessed health status; K(x l ,x) represents the kernel function, σ represents the kernel parameter; b represents the weight of the fuel cell health status diagnosis model; N represents the total number of health status types; a i represents the i-th Lagrangian coefficient.

[0042] According to a second aspect, an embodiment of the present invention further provides a fuel cell health status diagnosis and estimation system based on polarization characteristics, the system comprising:

[0043] A feature extraction module is used to construct a proton exchange membrane fuel cell model based on a multi-physical field with multiple conditions, and to extract health status features of the battery to be diagnosed from the proton exchange membrane fuel cell model under preset typical working conditions;

[0044] A health diagnosis module, used for inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model;

[0045] The fuel cell health status diagnosis model is trained by the following steps:

[0046] Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery;

[0047] Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum;

[0048] A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of the sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature is used as a fault state feature;

[0049] Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature;

[0050] The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

[0051] According to the third aspect, an embodiment of the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics as described in any one of the above are implemented.

[0052] According to a fourth aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for diagnosing and estimating the health status of a fuel cell based on polarization characteristics.

[0053] According to a fifth aspect, an embodiment of the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics as described in any one of the above items.

[0054] Compared with the prior art, the fuel cell health status diagnosis and estimation method and system of the present invention based on polarization characteristics have the following beneficial effects:

[0055] By simulating a variety of typical working conditions, and constructing a proton exchange membrane fuel cell model based on a multi-physical field of gas, water, electricity, heat, aging and other conditions, the health status characteristics of the battery to be diagnosed under these typical working conditions are extracted. After that, the health status characteristics are input as input data into the trained fuel cell health status diagnosis model to obtain the health status output by the fuel cell health status diagnosis model. In addition, since the fuel cell health status diagnosis model is also constructed based on a multi-physical field of multiple conditions during the training process, the proton exchange membrane fuel cell model is obtained to extract the sample polarization health status characteristics of the sample battery under typical working conditions. At the same time, the sample polarization health status characteristics use the sample activation loss to characterize the hydrogen leakage loss, the sample ohmic loss to characterize the membrane resistance increase loss, and the sample concentration loss to characterize the electrochemical active surface area loss, so as to represent the degradation mechanism of the sample battery, and the radial basis function is used as the discrete basis to extract the DRT of the sample electrochemical impedance spectrum. DRT can identify polarization processes with different time constants in complex electrochemical systems. The time constants and impedance sizes of different electrochemical processes can be accurately separated and quantified according to the position and area of ​​the peak in the DRT, so as to directly obtain the time scale distribution of the system under study. The fuel cell health status diagnosis model will also extract the polarization resistance characteristics represented by the DRT, use the polarization resistance characteristics as fault state characteristics, and accurately classify the health status characteristics through the least squares support vector machine. This can not only accurately diagnose the health status of the fuel cell, but also has low implementation difficulty and cost and strong interpretability, thereby reducing the dependence of high-precision fuel cell health status diagnosis on expert experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a flow chart of the method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics of the present invention;

[0058] Figure 2 A schematic diagram of a flow chart of health status assessment of a fuel cell water content estimation method based on a physical information neural network according to the present invention;

[0059] Figure 3 It is a structural schematic diagram of a PEMFC model constructed in the fuel cell health status diagnosis and estimation method based on polarization characteristics of the present invention;

[0060] Figure 4It is a schematic diagram of the process of training a fuel cell health status diagnosis model in the fuel cell health status diagnosis and estimation method based on polarization characteristics of the present invention;

[0061] Figure 5 The structure diagram of the fuel cell health status diagnosis and estimation system based on polarization characteristics provided by the present invention is shown;

[0062] Figure 6 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0063] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Due to the advantages of high energy efficiency and zero emissions, PEMFC systems have become an indispensable part of the future green energy structure. However, the PEMFC system is complex, in which the stack is coupled by multiple physical fields of gas, water, electricity, and heat. External auxiliary equipment such as air compressors, flow controllers, and humidifiers are difficult to coordinate and control. The abnormal health status is numerous and the sources are complex, making it difficult to accurately locate and evaluate it. In order to enhance the operational reliability and safety of the PEMFC system, low-cost, high-performance fuel cell health status diagnostic technology is necessary.

[0065] At present, the methods for fuel cell health status diagnosis are mainly divided into two types: physical model-based and data-driven. The method based on the mechanism model constructs a numerical analysis model that includes multiple mechanisms such as electrochemistry, heat transfer and porous medium two-phase flow, and relies on the residual between the predicted value and the measured value to detect the health status. However, the mechanism model usually only considers the physical fields of gas, water, electricity and heat, and can only realize the health status diagnosis of local physical fields. In addition, the strong coupling effect between them will also significantly affect the reliability of diagnosis, making the diagnosis result dependent on expert experience. The data-driven diagnosis method needs to first extract the health status characteristics of the fuel cell, which are generally derived from the global time series voltage, pressure and temperature. In addition, frequency domain features such as EIS and DRT are also used for health status feature extraction. After that, the health status features are classified with the help of machine learning algorithms such as least squares support vector machine, XGBoost, or deep learning models such as self-attention and bidirectional gated recurrent unit. However, the data-driven diagnosis method is heavily dependent on data quality, and complex prediction algorithms are difficult to deploy on hardware devices. Therefore, there is a problem of how to extract high-information-dimensional health status features when implementing health status diagnosis using such methods.

[0066] In summary, how to provide a fuel cell health status diagnosis method with high accuracy, low implementation cost and no reliance on expert experience is an important issue that needs to be solved urgently in the industry.

[0067] The fuel cell health status diagnosis and estimation method based on polarization characteristics provided in this specification can be applied to electronic devices with data processing capabilities. The electronic devices may include notebooks, desktop computers, smart phones, smart wearable devices (virtual reality glasses, smart watches, etc.), tablet computers, etc. Of course, the fuel cell health status diagnosis and estimation method based on polarization characteristics provided in this specification can also be applied to applications running in the above-mentioned electronic devices. For example, the fuel cell health status diagnosis and estimation method based on polarization characteristics can be applied to browsers with data processing capabilities, and can also be applied to processing software with data processing capabilities.

[0068] See also Figure 1 and Figure 2 , Figure 1 A flowchart of a method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to an embodiment of the present invention is shown. Figure 2 The figure is a flow chart of the health status assessment of the fuel cell water content estimation method based on the physical information neural network of the present invention. The method may include the following steps:

[0069] S101. A proton exchange membrane fuel cell model is obtained based on a multi-physical field with multiple conditions, and health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions.

[0070] See also Figure 3 As shown, a PEMFC model is constructed based on the coupling of multiple physical fields such as gas, water, electricity, heat, and aging. The PEMFC model serves as the source of the health status characteristics of the fuel cell, and a variety of typical operating conditions are designed in the PEMFC model to extract the health status characteristics of the battery to be diagnosed from the PEMFC model.

[0071] In this embodiment, typical operating conditions include but are not limited to normal, membrane dry, flooded and oxygen starved operating conditions. The health status characteristics may include gas velocity, gas pressure, gas mass fraction, liquid water saturation, modal water, ion potential, electron potential, stack temperature, voltage and other characteristics of the battery to be diagnosed.

[0072] S102: Input the health status characteristics into the trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model.

[0073] The health status feature is the input data of the trained fuel cell health status. The health status is the output data obtained after processing the input data of the trained fuel cell health status. The output data is used to evaluate the health status of the battery to be diagnosed.

[0074] See also Figure 4 , where the training process of the fuel cell health status diagnosis model is:

[0075] S103. Extracting a sample polarization health status characteristic of a sample battery from a proton exchange membrane fuel cell model under a preset typical operating condition.

[0076] Similarly, the sample polarization health status characteristics of the sample battery are extracted from the PEMFC model. The sample polarization health status characteristics correspond to the sample gas velocity, sample gas pressure, sample gas mass fraction, sample liquid water saturation, sample modal water, sample ion potential, sample electron potential, sample stack temperature, sample voltage and other characteristics of the sample battery.

[0077] More specifically, in the PEMFC model, based on the conservation of mass energy, the gas velocity u g It can be expressed as:

[0078]

[0079] In formula (1), ε represents the porosity of the reaction gas; s represents the liquid water saturation; ρ g Represents the density of the reaction gas; S m It can be understood that the gas velocity of the battery to be diagnosed and the sample gas velocity of the sample battery can be obtained by using this formula.

[0080] Based on the law of conservation of momentum energy, the gas pressure P g It can be expressed as:

[0081]

[0082] In formula (2), μ g Indicates the viscosity of the reaction gas; S u It is understandable that the formula can be used to obtain the gas pressure of the battery to be diagnosed and the sample gas pressure of the sample battery.

[0083] Based on the conservation of components, the gas mass fraction Y i It can be expressed as:

[0084]

[0085] In formula (3), S i represents the component source term; It is understood that the gas mass fraction of the battery to be diagnosed and the sample gas mass fraction of the sample battery can be obtained by using this formula.

[0086] Liquid water saturation s can be expressed as:

[0087]

[0088] In formula (4), ρ1 represents the density of liquid water; μ1 represents the viscosity of liquid water; κ1 represents the relative permeability; κ c1 represents the permeability of the catalyst layer; S1 represents the liquid water source term. It can be understood that the formula can be used to obtain the liquid water saturation of the battery to be diagnosed and the sample liquid water saturation of the sample battery.

[0089] Based on the conservation of membrane dissolved water concentration, the modal water λ can be expressed as:

[0090]

[0091] In formula (5), n d represents the proton transfer coefficient; ε cl represents the volume fraction of ion membrane in electrolyte; ρ mem represents the density of the ion membrane in the electrolyte; EW represents the equivalent mass of the ion membrane in the electrolyte; J ion represents current density; F represents Faraday constant; D mw represents the membrane water diffusion coefficient; S d It is understandable that the modal water of the battery to be diagnosed and the sample modal water of the sample battery can be obtained by using this formula.

[0092] Ionic potential It can be expressed as:

[0093]

[0094] In formula (6), C dl represents the double layer capacitance; η represents the double layer activation loss parameter; Represents the equivalent conductivity of ions; S ion It is understood that the formula can be used to obtain the ion potential of the battery to be diagnosed and the sample ion potential of the sample battery.

[0095] Electron potential It can be expressed as:

[0096]

[0097] In formula (7), Represents the equivalent conductivity of electrons; Se It is understood that the electron potential of the battery to be diagnosed and the sample electron potential of the sample battery can be obtained by using this formula.

[0098] Based on the law of energy conservation, the stack temperature T can be expressed as:

[0099]

[0100] In formula (8), C ρ,1 Represents the specific heat capacity of liquid water; C ρ,g represents the specific heat capacity of gas; Represents equivalent heat capacity.

[0101] Considering the sample activation loss η act 、Sample ohmic loss η ohm And the sample concentration loss η con For these three aging parameters, the sample voltage E of the fuel cell can be expressed as:

[0102] E=U0-η act -η ohm -η con (9)

[0103] In formula (9), U0 represents the open circuit voltage. It can be understood that the three aging parameters of activation loss, ohmic loss and concentration loss can be used when calculating the voltage of the battery to be diagnosed.

[0104] Since the anode activation loss η α Greater than 0 and cathode activation loss η c Less than 0, the fuel cell has the following approximate solution:

[0105]

[0106] In formula (10), R represents the gas constant and T represents the stack temperature; represents the anode transfer coefficient on the hydrogen side; represents the cathode transfer coefficient on the oxygen side; i a,ref represents the anode reference exchange current density; i c,ref represents the cathode reference exchange current density; i represents the stack current; i loss Indicates hydrogen leakage loss.

[0107] Since the conductivity of the membrane electrode is much smaller than that of the gas diffusion layer, flow channel, etc., the ohmic loss η ohm It can also be expressed as:

[0108]

[0109] In formula (11), σion Indicates the conductivity of the sample; L MEA represents the sample length of the membrane electrode; S MEA represents the sample cross-sectional area of ​​the membrane electrode.

[0110] Based on Fick's law and Nernst equation, the concentration loss η caused by the rapid consumption of reactants at the electrode con It can be expressed as:

[0111]

[0112] In formula (12), n represents the number of electrons involved in the reaction; i lim represents the sample limiting current density when the consumption rate is equal to the diffusion rate.

[0113] In this embodiment, three aging parameters are introduced: sample activation loss, sample ohmic loss, and sample concentration loss. The sample activation loss represents hydrogen leakage loss, the sample ohmic loss represents membrane resistance increase loss, and the sample concentration loss represents electrochemical active surface area loss. The degradation mechanism of the fuel cell is characterized by these three aging parameters. It can be understood that when calculating the health status of the battery to be diagnosed, the activation loss represents hydrogen leakage loss, the ohmic loss represents membrane resistance increase loss, and the concentration loss represents electrochemical active surface area loss.

[0114] Thus, based on Fick's law, the hydrogen leakage flux is It can be expressed as:

[0115]

[0116] In formula (13), represents the hydrogen concentration in the catalyst layer; represents the membrane hydrogen leakage coefficient, which increases with aging, and Θ1 represents the first activation factor, Indicates membrane hydrogen leakage coefficient; δ m It is understood that the formula can be used to obtain the hydrogen leakage flux of the battery to be diagnosed and the sample hydrogen leakage flux of the sample battery.

[0117] Hydrogen leakage current I loss It can be expressed as:

[0118]

[0119] It can be understood that the formula can be used to obtain the hydrogen leakage current of the battery to be diagnosed and the sample hydrogen leakage current of the sample battery.

[0120] Aging relationship with time, conductivity κ ion,dIt can be expressed as:

[0121] κ ion,d =θ2κ ion (15)

[0122] In formula (15), κ ion,d represents conductivity; κ ion represents the initial conductivity of the electrolyte; Θ2 represents the first activation factor. It can be understood that the conductivity of the battery to be diagnosed and the sample conductivity of the sample battery can be obtained using this formula.

[0123] The aging relationship of the cathode reference exchange current density that characterizes the corrosion pt can be expressed as:

[0124] i c,ref,d =θ3·10·L pt,c ·ECSA c ·i c,ref (16)

[0125] In formula (16), i c,ref,d represents the cathode reference exchange current density considering the aging mechanism; L pt,c Indicates cathode platinum loading; ECSA c represents the cathode electrochemical active area; Θ3 represents the third activation factor. It can be understood that the formula can be used to obtain the reference exchange current density of the battery to be diagnosed and the sample reference exchange current density of the sample battery.

[0126] The three activation factors can be expressed according to the relationship between the current cycle conditions:

[0127]

[0128] In formula (17), c represents the number of current cycles; a required for calculating the activation factor is ij It represents the empirical coefficient of the jth order of the ith activation factor.

[0129] S104, adding a preset sinusoidal excitation voltage to the sample polarization health status characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum.

[0130] Specifically, step S104 includes:

[0131] S1041. Add a preset weak sinusoidal excitation voltage V0sin(ωt) to the sample voltage of the fuel cell in formula (9), and the corresponding sinusoidal excitation current I0sin(ωt+φ) can be obtained.

[0132] S1042. Based on the sinusoidal excitation voltage and the sinusoidal excitation current, a sample electrochemical impedance spectrum of the fuel cell can be obtained. The specific calculation formula can be expressed as:

[0133]

[0134] In formula (18), R(ω) represents the electrochemical impedance spectroscopy, R(ω)=[Z' exp ,Z” exp i],Z' exp Represents the real value; Z" exp It is understandable that the formula can be used to obtain the electrochemical impedance spectrum of the battery to be diagnosed and the sample electrochemical impedance spectrum of the sample battery.

[0135] S1043, extracting the high-frequency impedance characteristics of the sample corresponding to each frequency band from the sample electrochemical impedance spectrum according to the preset adaptive weight factor. The specific calculation formula can be expressed as:

[0136]

[0137] In formula (19), X HFR represents the high frequency impedance characteristic, i.e., the high frequency ohmic impedance (High Frequency Resistance, HFR) of the fuel cell; R(ω0) represents the high frequency impedance value corresponding to the angular velocity ω0; ξ represents the preset adaptive weight factor. In this way, the sample high frequency impedance characteristics corresponding to each frequency can be obtained. It can be understood that the formula can be used to obtain the high frequency impedance characteristics of the battery to be diagnosed and the sample high frequency impedance characteristics of the sample battery.

[0138] The adaptive weight factor ξ is obtained by minimizing the deviation of a preset number of sample high-frequency impedance features and a preset number of sample polarization resistance features corresponding to these sample high-frequency impedance features. The corresponding calculation formula can be expressed as:

[0139]

[0140] In formula (20), r i represents the sample polarization resistance feature corresponding to the i-th sample high-frequency impedance feature; I represents the number of sample high-frequency impedance features / sample polarization resistance features participating in the deviation calculation. It can be understood that the number of sample data specifically selected for the deviation calculation can be configured by the user, that is, a preset number of sample data is selected to calculate the specific value of the adaptive weight factor ξ.

[0141] S105, using radial basis function as a discrete basis, and extracting DRT of the electrochemical impedance spectrum of the sample based on the radial basis function, extracting the polarization resistance feature of the sample represented by the DRT, and using the polarization resistance feature as a fault state feature.

[0142] More specifically, step S105 includes:

[0143] S1051. Fit the sample battery as a parallel array of several resistors and an infinite number of resistors and capacitors to obtain a non-negative distribution function that describes the time relaxation characteristics of the electrochemical system of the sample battery, and obtain a logarithmic term for adapting the frequency of the electrochemical impedance spectrum based on the non-negative distribution function. In this way, a non-negative distribution function and a logarithmic term converted based on the non-negative distribution function are obtained.

[0144] The fuel cell health diagnosis model can be interpreted as a main circuit resistor and an infinite parallel array of resistors and capacitors (RC), namely:

[0145]

[0146] In formula (21), Z DRT (f) indicates DRT sequence; R ∞ represents the main circuit resistance; g(τ) represents the non-negative distribution function of the time relaxation characteristics of the electrochemical system; f represents the frequency sequence; γ(*) represents the relaxation time distribution function, γ(lnτ) = γg(τ), this conversion is used to adapt to the logarithmic representation of the EIS frequency; τ represents the relaxation time.

[0147] DRT is an efficient, model-free EIS data analysis method that can identify polarization processes with different time constants in complex electrochemical systems by analyzing their connections with electrochemical systems and their multiple related processes. The time constants and impedance magnitudes of different electrochemical processes can be accurately separated and quantified based on the position and area of ​​the peaks in DRT, thereby directly obtaining the time scale distribution of the system under study.

[0148] S1052: Using radial basis function as a discrete basis, and discretizing the logarithmic term based on the radial basis function to obtain a DRT extraction model.

[0149] The fitting of formula (21) constitutes an ill-posed problem, and to solve this problem, it is necessary to first discretize the γ(lnτ) term using a finite discrete function, and then apply the regularized regression method to determine the discrete parameters. In this embodiment, a radial basis function (RBF) is used as the discrete basis, so the γ(lnτ) term introduced by the RBF can be expressed as:

[0150]

[0151] In formula (22), μ represents the shape parameter; φ μ represents the Gaussian function corresponding to the shape parameter μ; τ m Indicates the center time scale; X m represents the coefficient of RBF, x m =g(τ m ), M represents the number of characteristic times, and φ(·) represents the RBF function.

[0152] S1053. Based on the DRT extraction model characterized by the discretized γ(lnτ) term, the sample electrochemical impedance spectrum R(ω) is fitted. In this way, the impedance spectrum error objective function can be minimized, so that the DRT of the sample electrochemical impedance spectrum can be extracted. The impedance spectrum error objective function S(x) is:

[0153]

[0154] In formula (23), A' represents the real matrix coefficient; A" represents the imaginary matrix coefficient; Ω' represents the real matrix weight; Ω" represents the imaginary matrix weight; represents the regularization parameter; P(x) represents the regularization term; R ∞1 represents the high frequency impedance; x represents the DRT approximate parameter vector associated with the discretization basis.

[0155] S1054, characterize each internal resistance polarization process of the sample battery as a corresponding relaxation time distribution peak, extract the sample polarization resistance characteristics characterized by the relaxation time distribution, and use the area of ​​each relaxation time distribution peak as a corresponding fault state feature.

[0156] According to the characteristics of the internal resistance polarization process of the fuel cell, each polarization process is characterized as a separate DRT peak. Assuming that the frequency sequence f is arranged from small to large, there are κ identifiable polarization processes in the fuel cell and each polarization process corresponds to a DRT peak in a different frequency band, then the DRT sequence in formula (21) can be expressed as:

[0157] Z DRT (f)=[Z1(f1),Z2(f2),…,Z κ (f κ )] (twenty four)

[0158] In formula (24), f i represents the frequency corresponding to the ith identifiable polarization process, Z i represents f i Impedance sequence corresponding to the frequency band.

[0159] The DRT peak area S can be expressed as:

[0160] S=[S1(f1),S2(f2),…,S κ (f κ )] (25)

[0161] In formula (25), S i represents the peak area corresponding to the i-th identifiable polarization process.

[0162] See also Figure 4 , each identifiable polarization process appears as a semicircle in the frequency-impedance coordinate system. Assuming f i Satisfy f i =[f i1 ,f i2 ,…,f iκ ], f ij represents the jth sampling point of the i-th identifiable polarization process. The peak area is calculated by integrating all sampling points of each identifiable polarization process. Then S i (f i ) can be expressed as:

[0163]

[0164] In formula (26), f i1 represents the initial frequency of the i-th peak frequency band; Indicates the end frequency of the i-th peak frequency band.

[0165] Then the peak area S = [S1(f1), S2(f2), ..., S κ (f κ )] is regarded as the κ fault characteristics of the current fuel cell, so we can get Among them, X pol represents the polarization impedance characteristics of the sample, X i represents the polarization impedance characteristic of the i-th sample, and L represents the total number of sample data.

[0166] S106, performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature.

[0167] Specifically, the obtained sample polarization resistance features and sample high-frequency impedance features are sorted in ascending order according to the frequency, and then the sorted sample polarization resistance features and sample high-frequency impedance features are connected in sequence to obtain the sample fusion feature, that is, X fault =[X pol ,X HFR ], where X fault Represents sample fusion features.

[0168] S107. Classify the sample fusion features using a least squares support vector machine, and train a fuel cell health status diagnosis model.

[0169] Specifically, step S107 includes:

[0170] S1071. Use Gaussian function to construct a least squares support vector machine, and construct an initial fuel cell health status diagnosis model based on the least squares support vector machine.

[0171] For N groups of different health states Y=[y1,y2,…,y N ] sample fusion feature X = [X fault_1 ,X fault_2 ,…,X fault_N ], X fault_i Represents sample y i The corresponding fusion feature, in this embodiment, uses the Gaussian function Construct a fuel cell health status diagnosis model based on least squares support vector machine, namely:

[0172]

[0173] In formula (27), b represents the bias of the model; θ represents the weight of the model.

[0174] It should be noted that N represents the total number of health status types, and the health status may include normal, type 1 fault, type 2 fault, type K fault, and so on.

[0175] S1072. Classify the sample fusion features using the fuel cell health status diagnosis model, and adjust the parameters of the fuel cell health status diagnosis model using the Lagrangian function.

[0176] Construct the Lagrangian function according to the requirements of the least squares support vector machine optimization problem:

[0177]

[0178] In formula (28), e i represents the i-th prediction error; a i represents the i-th Lagrangian coefficient.

[0179] By taking partial derivatives of θ, b, e, and α, we can obtain the fuel cell health status diagnosis model under the optimal solution conditions, namely:

[0180]

[0181] In formula (29), y(x) represents the assessed health status; K(x l ,x) represents the kernel function, σ represents the kernel parameter.

[0182] That is, a least squares support vector machine is constructed based on a preset Gaussian function to obtain an initial fuel cell health status diagnostic model, and then the initial fuel cell health status diagnostic model is used to classify the sample fusion features to obtain several groups of health status, and then the Lagrangian function of the initial fuel cell health status diagnostic model is constructed, and the Lagrangian function is used to adjust the bias and weight of the initial fuel cell health status diagnostic model until the prediction error of the health status is less than the preset value, thereby obtaining a trained fuel cell health status diagnostic model.

[0183] The fuel cell health status diagnosis and estimation method based on polarization characteristics of the present invention simulates a variety of typical working conditions, and obtains a proton exchange membrane fuel cell model through multi-physical field construction based on multiple conditions such as gas, water, electricity, heat, aging, etc. to extract the health status characteristics of the battery to be diagnosed under these typical working conditions. After that, the health status characteristics are input into the trained fuel cell health status diagnosis model as input data to obtain the health status output by the fuel cell health status diagnosis model. In addition, since the fuel cell health status diagnosis model is also based on the multi-physical field construction of multiple conditions during the training process, the proton exchange membrane fuel cell model is obtained to extract the sample polarization health status characteristics of the sample battery under typical working conditions. At the same time, the sample polarization health status characteristics use the sample activation loss to characterize the hydrogen leakage loss, the sample ohmic loss to characterize the membrane resistance increase loss, and the sample concentration loss to characterize the electrochemical active surface area loss, so as to represent the degradation mechanism of the sample battery, and use the radial basis function as the discrete basis to extract the DRT of the sample electrochemical impedance spectrum. The DRT can identify polarization processes with different time constants in complex electrochemical systems. The time constants and impedance sizes of different electrochemical processes can be accurately separated and quantified according to the position and area of ​​the peak in the DRT, so as to directly obtain the time scale distribution of the system under study. The fuel cell health status diagnosis model will also extract the polarization resistance characteristics represented by the DRT, use the polarization resistance characteristics as fault state characteristics, and accurately classify the health status characteristics through the least squares support vector machine. This can not only accurately diagnose the health status of the fuel cell, but also has low implementation difficulty and cost and strong interpretability, thereby reducing the dependence of high-precision fuel cell health status diagnosis on expert experience.

[0184] The following describes a system provided by an embodiment of the present invention. The system described below and the method described above can be referenced to each other.

[0185] See also Figure 5 , Figure 5 A schematic diagram of the structure of a fuel cell health status diagnosis and estimation system based on polarization characteristics according to an embodiment of the present invention is shown. The system may include:

[0186] The feature extraction module 10 is used to construct a proton exchange membrane fuel cell model based on a multi-physical field with multiple conditions, and to extract the health status characteristics of the battery to be diagnosed from the proton exchange membrane fuel cell model under preset typical working conditions.

[0187] In this embodiment, a PEMFC model is constructed based on the coupling of multiple physical fields such as gas, water, electricity, heat, and aging. The PEMFC model serves as the source of the health status characteristics of the fuel cell, and a variety of typical operating conditions are designed for the PEMFC model to extract the health status characteristics of the battery to be diagnosed from the PEMFC model.

[0188] In this embodiment, typical operating conditions include but are not limited to normal, membrane dry, flooded and oxygen starved operating conditions. The health status characteristics may include gas velocity, gas pressure, gas mass fraction, liquid water saturation, modal water, ion potential, electron potential, stack temperature, voltage and other characteristics of the battery to be diagnosed.

[0189] The health diagnosis module 20 is used to input the health status characteristics into the trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model.

[0190] The health status feature is the input data of the trained fuel cell health status. The health status is the output data obtained after processing the input data of the trained fuel cell health status. The output data is used to evaluate the health status of the battery to be diagnosed.

[0191] The fuel cell health status diagnosis and estimation system based on polarization characteristics of the present invention extracts the health status characteristics of the battery to be diagnosed under these typical working conditions by simulating a variety of typical working conditions and constructing a proton exchange membrane fuel cell model based on a multi-physical field based on multiple conditions such as gas, water, electricity, heat, and aging. After that, the health status characteristics are input as input data into the trained fuel cell health status diagnosis model to obtain the health status output by the fuel cell health status diagnosis model. In addition, since the fuel cell health status diagnosis model is also constructed based on a multi-physical field based on multiple conditions during the training process, the proton exchange membrane fuel cell model is obtained to extract the sample polarization health status characteristics of the sample battery under typical working conditions. At the same time, the sample polarization health status characteristics use the sample activation loss to characterize the hydrogen leakage loss, the sample ohmic loss to characterize the membrane resistance increase loss, and the sample concentration loss to characterize the electrochemical active surface area loss, so as to represent the degradation mechanism of the sample battery, and use the radial basis function as the discrete basis to extract the DRT of the sample electrochemical impedance spectrum. The DRT can identify polarization processes with different time constants in complex electrochemical systems. The time constants and impedance sizes of different electrochemical processes can be accurately separated and quantified according to the position and area of ​​the peak in the DRT, so as to directly obtain the time scale distribution of the system under study. The fuel cell health status diagnosis model will also extract the polarization resistance characteristics represented by the DRT, use the polarization resistance characteristics as fault state characteristics, and accurately classify the health status characteristics through the least squares support vector machine. This can not only accurately diagnose the health status of the fuel cell, but also has low implementation difficulty and cost and strong interpretability, thereby reducing the dependence of high-precision fuel cell health status diagnosis on expert experience.

[0192] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610 (processor), a communication interface 620 (Communications Interface), a memory 630 (memory) and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic command in the memory 630 to execute a fuel cell health status diagnosis and estimation method based on polarization characteristics, the method comprising:

[0193] A proton exchange membrane fuel cell model is constructed based on a multi-physical field with multiple conditions, and the health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions;

[0194] Inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model;

[0195] The fuel cell health status diagnosis model is trained by the following steps:

[0196] Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery;

[0197] Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum;

[0198] A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of the sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature is used as a fault state feature;

[0199] Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature;

[0200] The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

[0201] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memor), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0202] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, the computer can execute the fuel cell health status diagnosis and estimation method based on polarization characteristics provided by the above methods, the method comprising:

[0203] A proton exchange membrane fuel cell model is constructed based on a multi-physical field with multiple conditions, and the health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions;

[0204] Inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model;

[0205] The fuel cell health status diagnosis model is trained by the following steps:

[0206] Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery;

[0207] Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum;

[0208] A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of the sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature is used as a fault state feature;

[0209] Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature;

[0210] The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

[0211] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the above-mentioned method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics, the method comprising:

[0212] A proton exchange membrane fuel cell model is constructed based on a multi-physical field with multiple conditions, and the health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions;

[0213] Inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model;

[0214] The fuel cell health status diagnosis model is trained by the following steps:

[0215] Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery;

[0216] Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum;

[0217] A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of the sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature is used as a fault state feature;

[0218] Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature;

[0219] The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

[0220] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or perform equivalent conversions on some of the technical features therein. However, these modifications or conversions do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics, characterized in that: The method comprises: A proton exchange membrane fuel cell model is constructed based on a multi-physical field with multiple conditions, and the health status characteristics of the battery to be diagnosed are extracted from the proton exchange membrane fuel cell model under preset typical working conditions; Inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model; The fuel cell health status diagnosis model is trained by the following steps: Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery; Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum; A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of a sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature of the sample is used as a fault state feature; Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature; The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

2. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 1, characterized in that: The feature fusion of the sample high-frequency impedance feature and the sample polarization resistance feature to obtain the sample fusion feature specifically includes: Sort the obtained sample polarization resistance characteristics and the sample high-frequency impedance characteristics in ascending order according to the magnitude of the frequency; The sorted sample polarization resistance characteristics and the sample high-frequency impedance characteristics are sequentially connected to obtain a sample fusion characteristic.

3. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 1, characterized in that: The sample polarization health status characteristic includes a sample voltage of the fuel cell, and the calculation formula of the sample voltage is: E=U0-η act -or ohm -or con Where, E represents the sample voltage; η act represents the sample activation loss; η ohm represents the sample ohmic loss; η con represents the sample concentration loss; U0 represents the open circuit voltage.

4. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 3, characterized in that: The method of adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum specifically includes: Adding a preset sinusoidal excitation voltage to the sample voltage to obtain a sinusoidal excitation current; Obtaining an electrochemical impedance spectrum of the sample according to a sinusoidal excitation voltage and a sinusoidal excitation current; According to the preset adaptive weight factor, the high-frequency impedance characteristics of the sample corresponding to each frequency band are extracted from the electrochemical impedance spectrum of the sample.

5. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 4, characterized in that: The adaptive weight factor is obtained by minimizing the deviation of a preset number of the sample high-frequency impedance features and a preset number of sample polarization resistance features corresponding to the sample high-frequency impedance features.

6. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 1, characterized in that: The method adopts radial basis function as a discrete basis, extracts relaxation time distribution of sample electrochemical impedance spectrum based on radial basis function, extracts sample polarization resistance characteristics characterized by relaxation time distribution, and uses the polarization resistance characteristics as fault state characteristics, specifically including: Fitting the sample battery to a parallel array of several resistors and an infinite number of resistors and capacitors to obtain a non-negative distribution function describing the time relaxation characteristics of the electrochemical system of the sample battery, and converting the non-negative distribution function to obtain a logarithmic term for adapting the frequency of the electrochemical impedance spectrum; Using radial basis function as a discrete basis, and discretizing the logarithmic term based on the radial basis function to obtain a relaxation time distribution extraction model; Fitting the sample electrochemical impedance spectrum using the relaxation time distribution extraction model to extract the relaxation time distribution of the sample electrochemical impedance spectrum; Each internal resistance polarization process of the sample battery is characterized as a corresponding relaxation time distribution peak, the sample polarization resistance characteristics characterized by the relaxation time distribution are extracted, and the area of ​​each relaxation time distribution peak is used as a corresponding fault state feature.

7. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 1, characterized in that: The method of using the least squares support vector machine to classify the sample fusion features and train the fuel cell health status diagnosis model specifically includes: A least squares support vector machine is constructed using a Gaussian function, and an initial fuel cell health status diagnosis model is constructed based on the least squares support vector machine; The sample fusion features are classified using a fuel cell health status diagnosis model, and the parameters of the fuel cell health status diagnosis model are adjusted using a Lagrangian function.

8. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 7, characterized in that: The method of using the fuel cell health status diagnosis model to classify the sample fusion features and using the Lagrangian function to adjust the parameters of the fuel cell health status diagnosis model specifically includes: A least squares support vector machine is constructed based on a preset Gaussian function to obtain an initial fuel cell health status diagnosis model; Using an initial fuel cell health status diagnosis model to classify the sample fusion features, and obtain several groups of health status; The Lagrangian function of the initial fuel cell health status diagnostic model is constructed, and the bias and weight of the initial fuel cell health status diagnostic model are adjusted using the Lagrangian function until the prediction error of the health status is less than a preset value, thereby obtaining a trained fuel cell health status diagnostic model.

9. The method for diagnosing and estimating the health status of a fuel cell based on polarization characteristics according to claim 7, characterized in that: The expression of the fuel cell health status diagnosis model is: Among them, y(x) represents the assessed health status; K(x l ,x) represents the kernel function, σ represents the kernel parameter; b represents the weight of the fuel cell health status diagnosis model; N represents the total number of health status types; a i represents the i-th Lagrangian coefficient.

10. A fuel cell health status diagnosis and estimation system based on polarization characteristics, characterized in that: The system comprises: A feature extraction module is used to construct a proton exchange membrane fuel cell model based on a multi-physical field with multiple conditions, and to extract health status features of the battery to be diagnosed from the proton exchange membrane fuel cell model under preset typical working conditions; A health diagnosis module, used for inputting the health status feature into a trained fuel cell health status diagnosis model to obtain the health status of the battery to be diagnosed evaluated by the fuel cell health status diagnosis model; The fuel cell health status diagnosis model is trained by the following steps: Extracting a sample polarization health status characteristic of a sample battery from the proton exchange membrane fuel cell model under a preset typical working condition; in the sample polarization health status characteristic, a sample activation loss is used to characterize hydrogen leakage loss, a sample ohmic loss is used to characterize membrane resistance increase loss, and a sample concentration loss is used to characterize electrochemical active surface area loss, and the sample activation loss, the sample ohmic loss, and the sample concentration loss are used together to represent the degradation mechanism of the sample battery; Adding a preset sinusoidal excitation voltage to the sample polarization health state characteristic, extracting the sample electrochemical impedance spectrum of the fuel cell, and extracting the sample high-frequency impedance characteristic from the sample electrochemical impedance spectrum; A radial basis function is used as a discrete basis, and a relaxation time distribution of an electrochemical impedance spectrum of a sample is extracted based on the radial basis function, a polarization resistance feature of the sample characterized by the relaxation time distribution is extracted, and the polarization resistance feature of the sample is used as a fault state feature; Performing feature fusion on the sample high-frequency impedance feature and the sample polarization resistance feature to obtain a sample fusion feature; The sample fusion features are classified using a least squares support vector machine, and the fuel cell health status diagnosis model is obtained through training.

Images