Fuel cell health state estimation method and system

Abstract

The invention relates to the technical field of fuel cell state evaluation, in particular to a fuel cell health state estimation method and system, and the method comprises the steps: continuously collecting the monomer polarization curve, pile AC impedance spectrum and gas flow channel pressure pulsation time sequence data of a target fuel cell in multiple operation cycles; ohm, activation and concentration polarization component decomposition are carried out on the polarization curve, a corresponding voltage loss component is extracted, correlation matching calculation is carried out on the corresponding voltage loss component and an impedance spectrum, and a polarization loss-impedance characteristic mapping relation is constructed. And extracting a pressure fluctuation mode synchronous with the polarization curve time sequence, and identifying an extra concentration difference loss component caused by the gas diffusion barrier by correlating the concentration difference polarization loss voltage. And integrating the component, the mapping relation and a reference aging model to calculate a comprehensive health state index. The multi-aging superposition state of the membrane electrode, the proton exchange membrane and the gas diffusion layer can be accurately reflected, aging inducement distinguishing is refined, and the accuracy of health state characterization is improved.

Description

Technical Field

[0001] This invention relates to the field of fuel cell condition assessment technology, and in particular to a method and system for estimating the health status of fuel cells. Background Technology

[0002] Current technologies for estimating the health status of fuel cells mostly rely on single polarization curve data or AC impedance spectroscopy data for analysis. Some technologies perform simple polarization type classification on the polarization curves, only obtaining basic polarization loss values. Pressure pulsation time-series data of the gas flow channel is often used alone for gas transport status monitoring, without being combined with electrochemical data. Conventional technologies, when analyzing the aging state of fuel cells, can only obtain overall polarization loss results, failing to establish a corresponding correlation between different polarization loss components and impedance spectral characteristics, and also failing to perform detailed processing for concentration polarization loss. The timing of pressure data acquisition and polarization curve acquisition is not synchronized, making it impossible to distinguish the different causes of concentration polarization loss.

[0003] During fuel cell operation, ohmic polarization, activation polarization, and concentration polarization correspond to different aging mechanisms. A single data dimension cannot accurately correspond to various aging characteristics, and impedance spectroscopy data cannot be directly converted into polarization loss components, which can lead to biases in aging mechanism analysis. Gas diffusion layer blockage and poor gas flow path transmission can generate additional concentration polarization losses. Conventional techniques cannot identify this type of specific loss component, and health status assessment cannot integrate the multiple aging effects of membrane electrode catalytic activity decline, proton exchange membrane aging, and gas diffusion layer blockage. It is necessary to establish a correlation between polarization loss components and impedance spectra, and at the same time extract synchronous temporal pressure fluctuation patterns and correlate them with concentration polarization losses to identify additional concentration loss components caused by gas diffusion obstacles. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for estimating the health status of fuel cells.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a fuel cell health status estimation method, comprising:

[0006] Continuously acquire dynamic electrochemical data sets of the target fuel cell over multiple complete operating cycles, including cell polarization curves, stack internal AC impedance spectra, and gas flow channel pressure pulsation time sequences.

[0007] The polarization curves of the battery cells are subjected to component decomposition processing of ohmic polarization, activation polarization and concentration polarization to extract a set of voltage loss components characterizing different aging mechanisms. The set of voltage loss components includes ohmic polarization loss voltage, activation polarization loss voltage and concentration polarization loss voltage.

[0008] By performing correlation matching calculations between the voltage loss component set and the internal AC impedance spectrum of the stack, a polarization loss-impedance characteristic mapping relationship is established. This polarization loss-impedance characteristic mapping relationship is used to deduce the corresponding polarization loss component from the impedance spectrum.

[0009] The pressure fluctuation pattern synchronized with the acquisition time of the polarization curve of the battery cell is extracted from the pressure pulsation time sequence of the gas flow channel, and the pressure fluctuation pattern is correlated with the concentration polarization loss voltage to identify the additional concentration loss component caused by gas diffusion obstacles.

[0010] By integrating the additional concentration loss component, the polarization loss-impedance characteristic mapping relationship, and the preset benchmark aging model, the comprehensive health status index of the target fuel cell is calculated. The comprehensive health status index reflects the superimposed effect of membrane electrode catalytic activity decay, proton exchange membrane dry aging, and gas diffusion layer blockage.

[0011] As a further aspect of the present invention, the step of performing component decomposition processing on the polarization curve of the battery cell for ohmic polarization, activation polarization, and concentration polarization to extract a set of voltage loss components characterizing different aging mechanisms includes:

[0012] In the high current density linear region of the cell polarization curve, the slope of the curve is obtained by linear regression fitting, and the voltage loss corresponding to the slope of the curve is calculated as the ohmic polarization loss voltage.

[0013] In the low current density initial region of the cell polarization curve, the activation polarization process is nonlinearly fitted using the Tafel equation, and the overpotential obtained from the fitting is calculated as the activation polarization loss voltage.

[0014] The ohmic polarization loss voltage and the activation polarization loss voltage are subtracted sequentially from the total voltage loss of the cell polarization curve, and the remaining voltage loss is calculated as the concentration polarization loss voltage.

[0015] The ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage corresponding to different operating times are recorded and arranged in chronological order to form the voltage loss component set.

[0016] As a further aspect of the present invention, the step of using the voltage loss component set and the internal AC impedance spectrum of the stack to perform correlation matching calculation to establish a polarization loss-impedance characteristic mapping relationship includes:

[0017] Relaxation time distribution analysis was performed on the internal AC impedance spectrum of the fuel cell stack to separate the characteristic impedance arcs corresponding to the charge transfer process, proton conduction process and mass transport process;

[0018] The characteristic frequency and characteristic resistance value are extracted from the characteristic impedance arc. The characteristic resistance value of the high-frequency region corresponding to the charge transfer process is subjected to linear regression analysis with the activation polarization loss voltage to obtain the first mapping coefficient.

[0019] The second mapping coefficient is obtained by performing a linear regression analysis between the characteristic resistance value in the mid-frequency region corresponding to the proton conduction process and the ohmic polarization loss voltage.

[0020] A trend correlation analysis is performed between the low-frequency characteristic impedance corresponding to the mass transfer process and the concentration polarization loss voltage to obtain the third mapping coefficient;

[0021] The first mapping coefficient, the second mapping coefficient, and the third mapping coefficient are integrated to construct the polarization loss-impedance characteristic mapping relationship.

[0022] As a further aspect of the present invention, the step of extracting the pressure fluctuation pattern synchronized with the acquisition time of the polarization curve of the battery cell from the gas flow channel pressure pulsation time sequence, and performing correlation analysis between the pressure fluctuation pattern and the concentration polarization loss voltage to identify the additional concentration loss component caused by gas diffusion obstacles includes:

[0023] In the gas flow channel pressure pulsation time sequence, lock the time segment that completely overlaps with the data acquisition window of the battery cell polarization curve;

[0024] Perform a fast Fourier transform on the pressure pulsation data of the time segment to calculate its energy proportion in the low-frequency band;

[0025] The low-frequency energy ratio is compared with the reference pressure pulsation spectrum. When the low-frequency energy ratio exceeds a set threshold, it is determined that the time segment has a pressure fluctuation pattern related to gas diffusion barriers.

[0026] The concentration polarization loss voltage of this time segment is subtracted from the concentration polarization loss voltage under the reference state without diffusion barriers, and the difference is quantized as the additional concentration loss component.

[0027] As a further aspect of the present invention, the step of performing a fast Fourier transform on the pressure pulsation data of the time segment to calculate its energy proportion in the low-frequency band includes:

[0028] The pressure pulsation data of the time segment is preprocessed with zero mean to obtain a centered pressure pulsation signal;

[0029] The Hanning window function is applied to the centered pressure pulsation signal, and then a fast Fourier transform is performed to obtain the pressure pulsation spectrum.

[0030] In the pressure pulsation spectrum, the frequency band below the set cutoff frequency is defined as the low frequency band. The sum of the squares of the amplitudes of all frequency components in the low frequency band is calculated to obtain the low frequency band energy.

[0031] The total energy of the pressure pulsation spectrum is obtained by summing the squares of the amplitudes of all frequency components across the entire frequency band.

[0032] The low-frequency energy is divided by the full-frequency energy to obtain the low-frequency energy percentage.

[0033] As a further aspect of the present invention, the calculation of the comprehensive health status index of the target fuel cell by fusing the additional concentration loss component, the polarization loss-impedance characteristic mapping relationship, and a preset benchmark aging model includes:

[0034] The reference polarization loss components and reference impedance characteristics of the fuel cell under standard operating conditions and brand-new conditions are read from the reference aging model.

[0035] By inputting the currently acquired internal AC impedance spectrum of the fuel cell into the polarization loss-impedance characteristic mapping relationship, the mapped polarization loss component in the current state can be calculated.

[0036] From the current actual set of voltage loss components, subtract the values ​​corresponding to the mapped polarization loss components to obtain the polarization loss residuals that cannot be directly mapped by impedance;

[0037] The additional concentration loss component is added to the concentration polarization portion of the polarization loss residual to obtain the total aging loss component related to gas diffusion.

[0038] The mapped polarization loss component and the total aging loss component are compared with the corresponding components in the benchmark aging model, and the comprehensive health status index is calculated by a weighted fusion algorithm.

[0039] As a further aspect of the present invention, the step of subtracting the values ​​corresponding to the mapped polarization loss components from the current actual set of voltage loss components to obtain polarization loss residuals that cannot be directly mapped by impedance includes:

[0040] Read the actual values ​​of the ohmic polarization loss voltage, the actual value of the activation polarization loss voltage, and the actual value of the concentration polarization loss voltage from the current actual set of voltage loss components;

[0041] From the mapped polarization loss components obtained by impedance spectrum mapping, read the mapped ohmic polarization loss voltage, the mapped active polarization loss voltage, and the mapped concentration polarization loss voltage;

[0042] Subtract the mapped ohmic polarization loss voltage from the actual value of the ohmic polarization loss voltage to obtain the ohmic polarization loss residual;

[0043] Subtracting the mapped activation polarization loss voltage from the actual value of the activation polarization loss voltage yields the activation polarization loss residual.

[0044] Subtract the mapped concentration polarization loss voltage from the actual value of the concentration polarization loss voltage to obtain the concentration polarization loss residual;

[0045] The ohmic polarization loss residual, the activation polarization loss residual, and the concentration polarization loss residual are combined to form the polarization loss residual.

[0046] As a further aspect of the present invention, it also includes a step of extrapolating and predicting aging trends based on the comprehensive health status index and historical operating data:

[0047] The comprehensive health status index of the target fuel cell at multiple historical time points is obtained to form a health status time series.

[0048] The health status time series is smoothed and filtered to eliminate short-term fluctuation noise, resulting in a smooth health status series that reflects the long-term trend.

[0049] Curve fitting is performed on the smoothed health state sequence to obtain an aging trend function describing the evolution of health state over time.

[0050] Using the aging trend function, the predicted health status of the target fuel cell at a future specified time point is extrapolated and predicted.

[0051] The predicted health status value is compared with a preset failure threshold. When the predicted health status value is lower than the failure threshold, it is determined that the target fuel cell may reach the end of its life at a specified future time.

[0052] As a further aspect of the present invention, the step of smoothing and filtering the health state time series to eliminate short-term fluctuation noise and obtain a smooth health state series reflecting long-term trends includes:

[0053] The health status time series is processed using a moving average algorithm, and the length of the moving window is set to be greater than the typical period of short-term fluctuations in the health status.

[0054] Starting from the beginning of the health status time series, the arithmetic mean of the data points covered by the sliding window is calculated sequentially, and the arithmetic mean is used as the smoothed health status value corresponding to the center point of the window.

[0055] Move the sliding window backward by one data point and repeat the above steps of calculating the arithmetic mean until the entire health status time series has been processed;

[0056] Arrange the smoothed health status values ​​corresponding to the center points of all windows in the original time order to form the smoothed health status sequence.

[0057] As a further aspect of the present invention, the present invention also includes a fuel cell health state estimation system, the system comprising a main control processor and a cooperative data memory, the cooperative data memory being electrically coupled to the main control processor and interacting with it for instructions and data, the cooperative data memory being used to store program instructions and a reference database, and the main control processor being used to read and execute the program instructions in the cooperative data memory to realize the entire process of the fuel cell health state estimation method described above.

[0058] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0059] By performing component decomposition processing of ohmic polarization, activation polarization, and concentration polarization on the polarization curves of individual cells, it is possible to extract the ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage corresponding to different aging mechanisms. Correlation matching calculations are performed on this set of voltage loss components and the internal AC impedance spectrum of the fuel cell stack, allowing a one-to-one correspondence between the polarization loss components and the impedance spectrum characteristics. The constructed polarization loss-impedance characteristic mapping relationship can directly deduce the corresponding polarization loss components from the impedance spectrum data, forming a quantifiable correspondence between the electrochemical characteristic changes corresponding to different aging mechanisms. This refines the analytical dimensions of fuel cell aging characteristics and aligns with the actual changes in the internal electrochemical aging of fuel cells.

[0060] Extracting the pressure fluctuation pattern synchronized with the acquisition time of the cell polarization curve by the gas flow channel pressure pulsation time sequence allows the pressure fluctuation characteristics to maintain temporal consistency with the changes in concentration polarization loss voltage. Correlation analysis of this pressure fluctuation pattern with concentration polarization loss voltage can separate the loss component caused by gas diffusion obstacles from the overall concentration polarization loss, accurately identify the additional concentration loss component caused by gas diffusion layer blockage and poor gas transmission, distinguish the different causes of concentration polarization loss, and make the characterization of concentration polarization loss more consistent with the actual working conditions of gas transmission inside the fuel cell, fully reflecting the changes in electrochemical performance loss caused by gas transmission failure. Attached Figure Description

[0061] Figure 1 This is a flowchart of a fuel cell health status estimation method according to the present invention;

[0062] Figure 2A flowchart establishing the polarization loss-impedance characteristic mapping relationship;

[0063] Figure 3 A graph showing the evolution of the proportion of low-frequency energy in pressure pulsations and the additional concentration loss component;

[0064] Figure 4 This is a time-series analysis diagram of pressure pulsation in the gas flow channel;

[0065] Figure 5 A time series and aging trend analysis chart of the comprehensive health status index of fuel cells. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0067] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0068] See Figure 1The study continuously acquires dynamic electrochemical data sets of the target fuel cell over multiple complete operating cycles. These data sets include individual cell polarization curves, stack internal AC impedance spectra, and gas flow channel pressure pulsation time sequences. The individual cell polarization curves are decomposed into ohmic polarization, activation polarization, and concentration polarization components to extract voltage loss components characterizing different aging mechanisms. These voltage loss components include ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage. Correlation matching calculations are performed between the voltage loss component sets and the stack internal AC impedance spectra to establish a polarization loss-impedance characteristic mapping relationship. This mapping relationship is used to deduce the corresponding polarization loss components from the impedance spectra. Pressure fluctuation patterns synchronized with the acquisition time of the individual cell polarization curves are extracted from the gas flow channel pressure pulsation time sequences. Correlation analysis is then performed between these pressure fluctuation patterns and concentration polarization loss voltage to identify additional concentration loss components caused by gas diffusion barriers. By integrating the additional concentration loss component, the polarization loss-impedance characteristic mapping relationship, and the preset benchmark aging model, the comprehensive health status index of the target fuel cell is calculated. This comprehensive health status index reflects the superposition effect of membrane electrode catalytic activity decay, proton exchange membrane dry aging, and gas diffusion layer blockage.

[0069] In one embodiment of the present invention, see [reference] Figure 2 The example scenario is set as the target fuel cell operating in multiple constant current discharge stages. At each stage, the cell polarization curve, stack internal AC impedance spectrum, and gas flow channel pressure pulsation time sequence are collected. Data comparison is performed by comparing the decomposition results with the baseline components under the initial healthy state to highlight the changes. In some embodiments, the cell polarization curve is decomposed into ohmic polarization, activation polarization, and concentration polarization components. In the high current density linear region of the cell polarization curve, the curve slope is obtained through linear regression fitting. The voltage loss corresponding to the curve slope is calculated as the ohmic polarization loss voltage. For example, data points are selected in the range where the current density is higher than 0.5 A / cm² for least squares fitting, and the product of the slope of the fitted line and the operating current is the ohmic polarization loss voltage. In the low current density initial region of the cell polarization curve, the activation polarization process is nonlinearly fitted using the Tafel equation. The overpotential obtained from the fitting is calculated as the activation polarization loss voltage. The Tafel equation is expressed as:

[0070]

[0071] in: Represents activation overpotential, and It is a constant obtained through nonlinear least squares fitting. This represents the current density. From the total voltage loss of the cell polarization curve, the ohmic polarization loss voltage and the activation polarization loss voltage are subtracted sequentially. The remaining voltage loss is calculated as the concentration polarization loss voltage. The ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage at different operating times are recorded and arranged in chronological order to form a voltage loss component set. Data comparison is achieved by overlaying the curves of the ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage over time in the voltage loss component set with a reference curve to visually reflect the aging progress.

[0072] Optionally, when establishing the polarization loss-impedance characteristic mapping relationship, relaxation time distribution analysis is performed on the AC impedance spectrum inside the fuel cell stack to separate characteristic impedance arcs corresponding to the charge transfer, proton conduction, and mass transport processes. Characteristic frequencies and characteristic resistance values ​​are extracted from these characteristic impedance arcs. Linear regression analysis is performed on the characteristic resistance values ​​in the high-frequency region corresponding to the charge transfer process and the activation polarization loss voltage to obtain the first mapping coefficient. For example, a linear fit is performed with the characteristic resistance value in the high-frequency region as the independent variable and the activation polarization loss voltage as the dependent variable; the slope of the fitted line is the first mapping coefficient. Linear regression analysis is performed on the characteristic resistance values ​​in the mid-frequency region corresponding to the proton conduction process and the ohmic polarization loss voltage to obtain the second mapping coefficient. Trend correlation analysis is performed on the characteristic impedance characteristics in the low-frequency region corresponding to the mass transport process and the concentration polarization loss voltage to obtain the third mapping coefficient. The trend correlation analysis uses the Pearson correlation coefficient to calculate the correlation strength between the amplitude of the characteristic impedance in the low-frequency region and the concentration polarization loss voltage; the third mapping coefficient is this correlation coefficient. The first, second, and third mapping coefficients are integrated to construct a polarization loss-impedance characteristic mapping relationship, which is stored in the system as a coefficient matrix. In specific implementations, data comparison in example scenarios also includes comparing the polarization loss components derived from the impedance spectrum through the mapping relationship under different operating cycles with the actual decomposed polarization loss components, and calculating the root mean square error to evaluate the mapping accuracy. In some embodiments, the relaxation time distribution analysis uses a regularization algorithm to process the AC impedance spectrum data inside the fuel cell stack, and the separated characteristic impedance arcs determine the characteristic frequency range of each process through peak identification. It can be understood that the polarization loss-impedance characteristic mapping relationship allows the ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage to be indirectly estimated solely from the AC impedance spectrum inside the fuel cell stack, reducing the dependence on frequent polarization curve testing.

[0073] In one embodiment of the present invention, the example scenario is set as a target fuel cell undergoing a stepped load cycle test. Gas flow channel pressure pulsation time-series sequence and cell polarization curves are simultaneously acquired. Data comparison involves comparing the additional concentration loss component calculated under the condition where gas diffusion obstacles are identified with the component under the baseline condition without such obstacles. In some embodiments, a pressure fluctuation pattern synchronized with the cell polarization curve acquisition time is extracted from the gas flow channel pressure pulsation time-series sequence. Within the gas flow channel pressure pulsation time-series sequence, a time segment completely overlapping with the cell polarization curve data acquisition window is locked. This synchronization is achieved through unified timestamp alignment, ensuring a strict temporal correspondence between the pressure pulsation data and the polarization curve data. Fast Fourier Transform is performed on the pressure pulsation data of the time segment to calculate its energy proportion in the low-frequency band. The low-frequency energy proportion is compared with the baseline pressure pulsation spectrum. When the low-frequency energy proportion exceeds a set threshold, it is determined that there is a pressure fluctuation mode related to gas diffusion barrier in the time segment. The concentration polarization loss voltage of the time segment is subtracted from the concentration polarization loss voltage under the baseline state without diffusion barrier. The difference is quantified as the additional concentration loss component. Data comparison can intuitively show that the additional concentration loss component is positive when there is diffusion barrier, while it approaches zero under normal conditions.

[0074] Optionally, the specific process of performing a Fast Fourier Transform (FFT) on the pressure pulsation data of a time segment and calculating the low-frequency energy proportion includes: performing zero-mean preprocessing on the pressure pulsation data of the time segment to obtain a centered pressure pulsation signal; applying a Hanning window function to the centered pressure pulsation signal; and then performing a FFT operation to obtain the pressure pulsation spectrum. In the pressure pulsation spectrum, the frequency band below a set cutoff frequency is defined as the low-frequency band. The sum of the squares of the amplitudes of all frequency components within the low-frequency band is calculated to obtain the low-frequency band energy. The sum of the squares of the amplitudes of all frequency components within the full-frequency band is calculated to obtain the full-frequency band energy. The low-frequency band energy is divided by the full-frequency band energy to obtain the low-frequency band energy proportion. The formula for calculating the low-frequency band energy proportion is expressed as:

[0075]

[0076] in: This represents the calculated proportion of low-frequency energy. Represents frequency, Represents frequency The amplitude of the pressure pulsation spectrum at that location. This represents the set cutoff frequency. This represents the Nyquist frequency. In some embodiments, the reference pressure pulsation spectrum is calculated using pressure data from a fuel cell operating stably under healthy, diffusion-free conditions, following the same process, and a spectrum database is established. The threshold is set based on the upper confidence limit of the normal distribution obtained statistically from the reference database. It is understood that gas diffusion barriers can cause changes in the gas pressure fluctuation pattern within the flow channel, with increased low-frequency fluctuation energy. By quantifying the change in the proportion of low-frequency energy and correlating it with concentration polarization loss voltage, the aging loss component caused by gas diffusion problems can be specifically identified and separated.

[0077] In one embodiment of the present invention, the example scenario is set as follows: after the target fuel cell has undergone a period of operation test including multiple operating conditions, the system has acquired the dynamic electrochemical dataset at the current moment and completed the aforementioned component decomposition and feature extraction. The data comparison is performed by comparing the finally calculated comprehensive health status index with the health status estimated based on a single feature to demonstrate the comprehensiveness of the fusion method. In some embodiments, the fusion calculation begins by reading the reference polarization loss component and reference impedance characteristics of the fuel cell under standard operating conditions and new conditions from the reference aging model. The reference aging model is a database established in advance through experiments or simulations, which stores the reference values ​​of ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage of the new battery under standard temperature, pressure, humidity, and reactant gas stoichiometry, as well as the corresponding AC impedance spectrum characteristics of the stack. The current AC impedance spectrum inside the fuel cell is input into the polarization loss-impedance characteristic mapping relationship to calculate the mapped polarization loss component in the current state. Specifically, the first, second, and third mapping coefficients stored in the mapping relationship are used, combined with the high-frequency, mid-frequency, and low-frequency characteristic values ​​extracted from the current impedance spectrum, to calculate the mapped activation polarization loss voltage, the mapped ohmic polarization loss voltage, and the mapped concentration polarization loss voltage, respectively.

[0078] From the current actual voltage loss component set, subtract the values ​​corresponding to the mapped polarization loss components to obtain the polarization loss residuals that cannot be directly mapped by impedance. Add the additional concentration loss component to the concentration polarization portion of the polarization loss residuals to obtain the total aging loss component related to gas diffusion. This step aims to integrate all losses related to gas diffusion barriers, regardless of whether they are explicitly reflected in the impedance spectrum. Compare the mapped polarization loss component and the total aging loss component with their corresponding components in the baseline aging model, and calculate the comprehensive health status index using a weighted fusion algorithm. The expression for the weighted fusion algorithm is:

[0079]

[0080] in: This represents the calculated comprehensive health status index. , , These are pre-set weighting coefficients that satisfy... , This represents the difference between the mapped ohmic polarization loss voltage and the reference value of the ohmic polarization loss voltage in the reference model. This represents the reference value of the ohmic polarization loss voltage in the reference model. This represents the difference between the mapped activation polarization loss voltage and the baseline value of the activation polarization loss voltage in the reference model. This represents the reference value of the activation polarization loss voltage in the reference model. This represents the difference between the total aging loss component related to gas diffusion and the baseline value of the concentration polarization loss voltage in the benchmark model. This represents the baseline value of concentration polarization loss voltage in the benchmark model. It can be understood that the weighted fusion algorithm normalizes and weights the loss components from different sources, reflecting different aging mechanisms, ultimately consolidating multi-dimensional aging information into a single comprehensive health status index. Optional, weighting coefficients... , , The value is determined based on the degree of impact of different aging modes on the overall performance degradation of the fuel cell, and can be determined through expert knowledge or statistical analysis of historical data.

[0081] See Figure 3 In the gas diffusion barrier characterization system for fuel cell health status estimation, the proportion of low-frequency energy in pressure pulsations and the additional concentration loss component exhibit a highly synchronized monotonically increasing trend. Both reflect the evolution of gas diffusion layer blockage over time. Specifically, as the target fuel cell's operating time gradually accumulates from 0 hours to 900 hours, the proportion of low-frequency energy (left vertical axis) in the gas flow channel pressure pulsation time series, after Fast Fourier Transform, continuously increases from an initial 0.10 to 0.80. This change directly reflects the shift of the gas disturbance mode within the flow channel towards lower frequencies, a typical spectral characteristic of the gradual intensification of gas diffusion barriers. Simultaneously, the additional concentration loss component (right vertical axis), quantified from the difference between the concentration polarization loss voltage and the baseline state, also linearly increases from 0V to approximately 0.0145V. This component directly corresponds to the additional concentration polarization voltage loss caused by gas transport obstruction. From a mechanistic perspective, the increase in the proportion of low-frequency energy stems from physical degradation such as blockage of the gas diffusion layer pores and dust accumulation in the flow channels, which leads to the attenuation of the turbulent components of the gas flow and the concentration of pulsation frequencies to low frequencies. The additional concentration loss component is a direct manifestation of this hydrodynamic degradation in electrochemical performance. The strong correlation between the two verifies the core hypothesis that "pressure pulsation spectrum characteristics can serve as a non-invasive monitoring indicator of gas diffusion obstacles," providing a key hydrodynamic dimension input for the subsequent calculation of the comprehensive health status index by integrating impedance characteristics and polarization loss components.

[0082] In one embodiment of the present invention, the example scenario is set as follows: during a complete performance diagnosis of a fuel cell system, the system has completed voltage loss component decomposition based on the current test data and generated corresponding mapped polarization loss components using the real-time acquired internal AC impedance spectrum of the fuel cell stack through the established polarization loss-impedance characteristic mapping relationship. Data comparison is achieved by constructing a table containing actual measured values, mapped calculated values, and their differences to clearly display the polarization loss residuals. The table is named "Actual Values, Mapped Values, and Residuals of Polarization Loss Components Table". In some embodiments, the actual values ​​of ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage are read from the current actual voltage loss component set. These actual values ​​are directly calculated from the synchronously acquired cell polarization curves through component decomposition processing. From the mapped polarization loss components obtained by impedance spectrum mapping, the mapped ohmic polarization loss voltage, mapped activation polarization loss voltage, and mapped concentration polarization loss voltage are read. These mapped values ​​are estimated values ​​derived from the characteristic parameters of the current internal AC impedance spectrum of the fuel cell stack based on the polarization loss-impedance characteristic mapping relationship.

[0083] Subtracting the mapped ohmic polarization loss voltage from the actual value of the ohmic polarization loss voltage yields the ohmic polarization loss residual, which can be calculated using the following expression:

[0084]

[0085] in: This represents the calculated ohmic polarization loss residual. This represents the actual value of the ohmic polarization loss voltage read from the voltage loss component set. This represents the mapped ohmic polarization loss voltage read from the mapped polarization loss component. Subtracting the mapped activation polarization loss voltage from the actual value of the activation polarization loss voltage yields the activation polarization loss residual. Subtracting the mapped concentration polarization loss voltage from the actual value of the concentration polarization loss voltage yields the concentration polarization loss residual. It can be understood that the polarization loss residual reflects the deviation between the actually measured performance loss and the performance loss indirectly estimated through impedance spectroscopy. This deviation may originate from aging mechanisms or measurement noise that the impedance characteristics fail to fully capture. Optionally, after obtaining each residual through the above subtraction calculations, the residual values ​​can be standardized, for example, by dividing by the corresponding actual value of the polarization loss voltage, to obtain relative residuals for cross-dimensional comparison. Combining the ohmic polarization loss residuals, activation polarization loss residuals, and concentration polarization loss residuals constitutes a complete set of polarization loss residuals. This set of polarization loss residuals is stored in vector form, represented as... ,in Represents the set of polarization loss residuals. Represents the residual of Ohmic polarization loss. Represents the activation polarization loss residual. This represents the residual of concentration polarization loss. Referring to Table 1, in specific implementation, data comparison can be reflected in Table 1, which lists the actual values, mapped values, and calculated residual values ​​of the three polarization losses at a given operating time. Through numerical comparison, it can be intuitively seen which polarization process has a larger mapping deviation.

[0086] Table 1: Actual values, mapped values, and residuals of polarization loss components

[0087] Polarization loss type Actual value (mV) Mapping value (mV) Residual value (mV) Ohmic polarization loss voltage 45.2 42.1 3.1 Activation polarization loss voltage 120.5 118.7 1.8 Concentration polarization loss voltage 85.7 78.3 7.4

[0088] In some embodiments, the process of reading the actual values ​​of the voltage loss component set and the mapped polarization loss component values ​​is performed synchronously to ensure time reference consistency. It is understood that the merged polarization loss residual will be used in subsequent calculations of the total aging loss component related to gas diffusion, and the concentration polarization loss residual will be added to the additional concentration loss component. Optionally, after each calculation, the system can record a historical sequence of polarization loss residuals to track the changing trends of loss components that cannot be mapped by impedance.

[0089] See Figure 4 In the time-series analysis of pressure pulsation in fuel cell gas flow channels, the extraction of pressure fluctuation patterns and the correlation analysis of concentration polarization loss are the core steps in identifying gas diffusion obstacles. Specifically, both the actual pressure pulsation curve (solid line) and the reference pressure fluctuation curve (dashed line) exhibit periodic oscillations around the standard pressure value (dotted dashed line, 100 kPa), reflecting the dynamic balance between gas supply and consumption during fuel cell operation. From the time-series characteristics, the amplitude of the actual pressure pulsation gradually increases over time, with the peak value rising from approximately 106 kPa to approximately 108 kPa, and the trough value decreasing from approximately 97 kPa to approximately 92 kPa. In contrast, the reference pressure fluctuation maintains a stable amplitude (approximately ±5 kPa). This amplitude difference is a typical manifestation of pressure pulsation distortion caused by aging phenomena such as gas diffusion layer blockage. At the frequency domain analysis level, the time segment synchronized with the polarization curve acquisition window needs to undergo zero-mean preprocessing and Hanning windowing. Then, a fast Fourier transform is used to obtain the pressure pulsation spectrum, and the energy proportion in the low-frequency band is calculated. When the low-frequency energy proportion of the actual pressure pulsation is significantly higher than that of the reference pressure pulsation, it can be determined that there is a gas diffusion obstacle. The difference between the concentration polarization loss voltage corresponding to this time segment and the reference state is the additional concentration loss component. This component will be merged with the concentration polarization part in the polarization loss residual and finally participate in the calculation of the comprehensive health status index.

[0090] In one embodiment of the present invention, the example scenario is set as a vehicle fuel cell system operating for one year. Its comprehensive health status index is periodically calculated and recorded, forming a health status time series containing twenty-six data points. Data comparison is achieved by plotting the original health status time series, the smoothed series, and the finally fitted aging trend curve on the same chart to visually demonstrate the effects of data smoothing and trend extraction. In some embodiments, aging trend extrapolation prediction is performed based on the comprehensive health status index and historical operating data to obtain the comprehensive health status index of the target fuel cell at multiple historical time points, forming a health status time series. Each data point in the series is associated with a specific calculation timestamp. The health status time series is smoothed and filtered to eliminate short-term fluctuation noise, resulting in a smoothed health status series reflecting long-term trends. The smoothing and filtering process uses a moving average algorithm, setting the length of the sliding window to be greater than the typical period of short-term health status fluctuations. For example, if the health status index fluctuates periodically due to daily start-stop cycles, the length of the sliding window is set to cover the number of data points covering several fluctuation periods. Curve fitting is performed on a smoothed health state sequence to obtain an aging trend function describing the evolution of health state over time. The form of the aging trend function can be selected based on the downward trend of the sequence, such as a linear function, exponential function, or power function. The function parameters are determined using the least squares method, and its expression can be:

[0091]

[0092] in: Represents time Health status index Representative at reference time Initial health status index, This represents the aging rate constant obtained through fitting. This is a natural constant. Using an aging trend function, the predicted health status of the target fuel cell at a future specified time point is extrapolated. This predicted health status is compared with a preset failure threshold. When the predicted health status is lower than the failure threshold, it is determined that the target fuel cell may reach the end of its lifespan at that future specified time point. The failure threshold can be set according to application requirements; for example, when the comprehensive health status index drops to 80% of its new state, it is determined that early warning maintenance is required.

[0093] Optionally, the specific steps for smoothing the health status time series to obtain a smoothed health status sequence include: processing the health status time series using a moving average algorithm; setting the length of the moving window, which must be greater than the typical period of short-term fluctuations in the health status; for example, if slight fluctuations in the health status index are observed with a monthly cycle, the moving window length is set to cover the number of data points spanning three months. Starting from the beginning of the health status time series, the arithmetic mean is calculated for each data point covered by the moving window, and this arithmetic mean is used as the smoothed health status value corresponding to the center point of the window. The moving window is then moved forward by one data point, and the above steps of calculating the arithmetic mean are repeated until the entire health status time series has been processed. All smoothed health status values ​​corresponding to the center points of the windows are arranged in their original chronological order to form a smoothed health status sequence. It can be understood that smoothing effectively filters out random fluctuations in the health status index caused by short-term operating condition fluctuations, measurement noise, etc., making the sequence more clearly reflect the long-term monotonic trend of performance degradation.

[0094] See Figure 5 In the comprehensive health status assessment and lifespan prediction system for fuel cells, the time-series analysis and aging trend fitting of the health status index are the core links for quantifying performance degradation and providing early warning of failure. Specifically, the original health status index sequence is calculated by fusing multi-cycle electrochemical characteristics, intuitively reflecting the health level resulting from the superposition of multiple mechanisms such as membrane electrode catalytic activity decay, proton exchange membrane aging, and gas diffusion layer blockage. The moving average smoothing sequence, by setting an appropriate window length, effectively filters out short-term operating condition fluctuations and measurement noise, highlighting long-term performance degradation patterns. The index-fitted aging trend function, based on the smoothed sequence, determines the aging rate constant using the least squares method, achieving accurate modeling of the health status evolution over time. The dashed line in the figure represents the preset failure warning threshold (80%). When the health status index falls below this threshold, the fuel cell enters a risk range of insufficient performance. As can be seen from the time-series evolution, the original health status index fluctuated significantly in the early stage of the monitoring period. After smoothing by moving average, the trend became clearer. The final fitted index aging trend curve completely depicted the entire process of gradual decline from the initial health status (about 99%) to the end of the lifespan. It can be used to extrapolate and predict the specific time node when the fuel cell reaches the failure threshold, providing a quantitative basis for preventive maintenance and life management.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for estimating the health status of a fuel cell, characterized in that, The method includes: Continuously acquire dynamic electrochemical data sets of the target fuel cell over multiple complete operating cycles, including cell polarization curves, stack internal AC impedance spectra, and gas flow channel pressure pulsation time sequences. The polarization curves of the battery cells are subjected to component decomposition processing of ohmic polarization, activation polarization and concentration polarization to extract a set of voltage loss components characterizing different aging mechanisms. The set of voltage loss components includes ohmic polarization loss voltage, activation polarization loss voltage and concentration polarization loss voltage. By performing correlation matching calculations between the voltage loss component set and the internal AC impedance spectrum of the stack, a polarization loss-impedance characteristic mapping relationship is established. This polarization loss-impedance characteristic mapping relationship is used to deduce the corresponding polarization loss component from the impedance spectrum. The pressure fluctuation pattern synchronized with the acquisition time of the polarization curve of the battery cell is extracted from the pressure pulsation time sequence of the gas flow channel, and the pressure fluctuation pattern is correlated with the concentration polarization loss voltage to identify the additional concentration loss component caused by gas diffusion obstacles. By integrating the additional concentration loss component, the polarization loss-impedance characteristic mapping relationship, and the preset benchmark aging model, the comprehensive health status index of the target fuel cell is calculated. The comprehensive health status index reflects the superimposed effect of membrane electrode catalytic activity decay, proton exchange membrane dry aging, and gas diffusion layer blockage.

2. The fuel cell health status estimation method according to claim 1, characterized in that, The process involves decomposing the polarization curves of the individual battery cells into ohmic polarization, activation polarization, and concentration polarization components to extract a set of voltage loss components characterizing different aging mechanisms, including: In the high current density linear region of the cell polarization curve, the slope of the curve is obtained by linear regression fitting, and the voltage loss corresponding to the slope of the curve is calculated as the ohmic polarization loss voltage. In the low current density initial region of the cell polarization curve, the activation polarization process is nonlinearly fitted using the Tafel equation, and the overpotential obtained from the fitting is calculated as the activation polarization loss voltage. The ohmic polarization loss voltage and the activation polarization loss voltage are subtracted sequentially from the total voltage loss of the cell polarization curve, and the remaining voltage loss is calculated as the concentration polarization loss voltage. The ohmic polarization loss voltage, activation polarization loss voltage, and concentration polarization loss voltage corresponding to different operating times are recorded and arranged in chronological order to form the voltage loss component set.

3. The fuel cell health status estimation method according to claim 2, characterized in that, The step of establishing a polarization loss-impedance characteristic mapping relationship by performing correlation matching calculations between the voltage loss component set and the internal AC impedance spectrum of the stack includes: Relaxation time distribution analysis was performed on the internal AC impedance spectrum of the fuel cell stack to separate the characteristic impedance arcs corresponding to the charge transfer process, proton conduction process and mass transport process; The characteristic frequency and characteristic resistance value are extracted from the characteristic impedance arc. The characteristic resistance value of the high-frequency region corresponding to the charge transfer process is subjected to linear regression analysis with the activation polarization loss voltage to obtain the first mapping coefficient. A second mapping coefficient is obtained by performing a linear regression analysis between the characteristic resistance value in the mid-frequency region corresponding to the proton conduction process and the ohmic polarization loss voltage. A trend correlation analysis is performed between the low-frequency characteristic impedance corresponding to the mass transfer process and the concentration polarization loss voltage to obtain the third mapping coefficient; The first mapping coefficient, the second mapping coefficient, and the third mapping coefficient are integrated to construct the polarization loss-impedance characteristic mapping relationship.

4. The fuel cell health status estimation method according to claim 3, characterized in that, The step of extracting the pressure fluctuation pattern synchronized with the acquisition time of the cell polarization curve from the gas flow channel pressure pulsation time sequence, and performing correlation analysis between the pressure fluctuation pattern and the concentration polarization loss voltage to identify the additional concentration loss component caused by gas diffusion obstacles includes: In the gas flow channel pressure pulsation time sequence, lock the time segment that completely overlaps with the data acquisition window of the battery cell polarization curve; Perform a fast Fourier transform on the pressure pulsation data of the time segment to calculate its energy proportion in the low-frequency band; The low-frequency energy ratio is compared with the reference pressure pulsation spectrum. When the low-frequency energy ratio exceeds a set threshold, it is determined that the time segment has a pressure fluctuation pattern related to gas diffusion barriers. The concentration polarization loss voltage of this time segment is subtracted from the concentration polarization loss voltage under the reference state without diffusion barriers, and the difference is quantized as the additional concentration loss component.

5. The fuel cell health status estimation method according to claim 4, characterized in that, The step of performing a Fast Fourier Transform on the pressure pulsation data of the time segment to calculate its energy proportion in the low-frequency band includes: The pressure pulsation data of the time segment is preprocessed with zero mean to obtain a centered pressure pulsation signal; The Hanning window function is applied to the centered pressure pulsation signal, and then a fast Fourier transform is performed to obtain the pressure pulsation spectrum. In the pressure pulsation spectrum, the frequency band below the set cutoff frequency is defined as the low frequency band. The sum of the squares of the amplitudes of all frequency components in the low frequency band is calculated to obtain the low frequency band energy. The total energy of the pressure pulsation spectrum is obtained by summing the squares of the amplitudes of all frequency components across the entire frequency band. The low-frequency energy is divided by the full-frequency energy to obtain the low-frequency energy percentage.

6. The fuel cell health status estimation method according to claim 5, characterized in that, The method integrates the additional concentration loss component, the polarization loss-impedance characteristic mapping relationship, and a preset benchmark aging model to calculate the comprehensive health status index of the target fuel cell, including: The reference polarization loss components and reference impedance characteristics of the fuel cell under standard operating conditions and brand-new conditions are read from the reference aging model. By inputting the currently acquired internal AC impedance spectrum of the fuel cell into the polarization loss-impedance characteristic mapping relationship, the mapped polarization loss component in the current state can be calculated. From the current actual set of voltage loss components, subtract the values ​​corresponding to the mapped polarization loss components to obtain the polarization loss residuals that cannot be directly mapped by impedance; The additional concentration loss component is added to the concentration polarization portion of the polarization loss residual to obtain the total aging loss component related to gas diffusion. The mapped polarization loss component, the total aging loss component, and the corresponding components in the benchmark aging model are compared, and the comprehensive health status index is calculated using a weighted fusion algorithm.

7. The fuel cell health status estimation method according to claim 6, characterized in that, The step of subtracting the values ​​corresponding to the mapped polarization loss components from the current actual set of voltage loss components to obtain the polarization loss residuals that cannot be directly mapped by impedance includes: Read the actual values ​​of the ohmic polarization loss voltage, the actual value of the activation polarization loss voltage, and the actual value of the concentration polarization loss voltage from the current actual set of voltage loss components; From the mapped polarization loss components obtained by impedance spectrum mapping, read the mapped ohmic polarization loss voltage, the mapped active polarization loss voltage, and the mapped concentration polarization loss voltage; Subtract the mapped ohmic polarization loss voltage from the actual value of the ohmic polarization loss voltage to obtain the ohmic polarization loss residual; Subtracting the mapped activation polarization loss voltage from the actual value of the activation polarization loss voltage yields the activation polarization loss residual. Subtract the mapped concentration polarization loss voltage from the actual value of the concentration polarization loss voltage to obtain the concentration polarization loss residual; The ohmic polarization loss residual, the activation polarization loss residual, and the concentration polarization loss residual are combined to form the polarization loss residual.

8. The fuel cell health status estimation method according to claim 7, characterized in that, It also includes the step of extrapolating and predicting aging trends based on the comprehensive health status index and historical operating data: The comprehensive health status index of the target fuel cell at multiple historical time points is obtained to form a health status time series. The health status time series is smoothed and filtered to eliminate short-term fluctuation noise, resulting in a smooth health status series that reflects the long-term trend. Curve fitting is performed on the smoothed health state sequence to obtain an aging trend function describing the evolution of health state over time. Using the aging trend function, the predicted health status of the target fuel cell at a future specified time point is extrapolated and predicted. The predicted health status value is compared with a preset failure threshold. When the predicted health status value is lower than the failure threshold, it is determined that the target fuel cell may reach the end of its life at a specified future time.

9. The fuel cell health status estimation method according to claim 8, characterized in that, The smoothing and filtering process applied to the health state time series to eliminate short-term fluctuation noise and obtain a smooth health state series reflecting long-term trends includes: The health status time series is processed using a moving average algorithm, and the length of the moving window is set to be greater than the typical period of short-term fluctuations in the health status. Starting from the beginning of the health status time series, the arithmetic mean of the data points covered by the sliding window is calculated sequentially, and the arithmetic mean is used as the smoothed health status value corresponding to the center point of the window. Move the sliding window backward by one data point and repeat the above steps of calculating the arithmetic mean until the entire health status time series has been processed; Arrange the smoothed health status values ​​corresponding to the center points of all windows in the original time order to form the smoothed health status sequence.

10. A fuel cell health status estimation system, characterized in that, The system includes a main control processor and a collaborative data memory. The collaborative data memory is electrically coupled to the main control processor and interacts with it for instructions and data. The collaborative data memory is used to store program instructions and a reference database. The main control processor is used to read and execute the program instructions in the collaborative data memory to implement the entire process of the fuel cell health state estimation method according to any one of claims 1 to 9.

Images