A rapid electrochemical impedance test method applied to low-temperature start of fuel cell

Abstract

The application discloses a kind of fast electrochemical impedance test methods applied to fuel cell low temperature start, belong to fuel cell detection technical field.The method is aimed at the problem that traditional electrochemical impedance spectrum test time is too long, cannot satisfy the difficulty of quick real-time monitoring in low temperature start process, first, by simulating water flooding and cathode blockage fault at normal temperature, construct multi-working condition impedance database;Further combined with equivalent circuit modeling and machine learning algorithm, 3~4 characteristic frequency bands most sensitive to fault state are selected from full-band data;Finally, in the process of fuel cell low temperature start, only the disturbance signal of these characteristic frequency bands is applied to the battery for fast scanning.The method shortens the single test time from traditional 2~5 minutes to 10~15 seconds, realizes the real-time, accurate diagnosis to membrane electrode icing and liquid water blockage and other key states, and provides an effective tool for optimizing fuel cell low temperature start control strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fuel cell detection, and particularly relates to a rapid electrochemical impedance test method applied to low-temperature starting of a fuel cell. BACKGROUND

[0002] In recent years, countries around the world have begun to pay close attention to and research renewable energy. Among many renewable energies, hydrogen energy is regarded as one of the most likely energy to replace traditional fossil fuels in the future as a clean secondary energy. Fuel cells, as a carrier of hydrogen energy, are widely concerned due to their high efficiency and no pollution. Water generated by electrochemical reaction may freeze under low-temperature starting conditions, which may cause rapid performance degradation or damage of the fuel cell system. Under this condition, supercooled water and more water vapor are easily generated inside the fuel cell, which causes abnormal internal state of the fuel cell stack and affects the normal operation of the fuel cell. The above conditions not only reduce the cold start performance of the fuel cell, but also shorten the service life. Therefore, it is necessary to monitor the state of the fuel cell in real time under low-temperature conditions, so as to adjust the operating conditions according to the monitoring results, so as to improve the performance of cold start. The structural characteristics of the fuel cell make it difficult to directly measure the internal state. Under the low-temperature starting state, only the single-cell voltage can be used for state detection, and the temperature and humidity of the fuel cell stack port have a large gap with the internal state of the fuel cell stack, and excessive dependence on these measured values may mislead the internal state of the fuel cell stack.

[0003] During the low-temperature starting process, whether the membrane temperature can rise above 0℃ is the key to the success of cold start, and ice accumulation and supercooled water are the main factors affecting the cold start performance. Ice mainly exists in the cathode side catalyst layer of the fuel cell membrane electrode, and ice crystals will block the pores in the catalyst layer that are originally used for oxygen transmission, causing part of the oxygen to fail to reach the reaction interface through the catalyst layer during the low-temperature starting process, resulting in an oxygen starvation failure, i.e., the oxygen transmission impedance becomes large. The supercooled water theory in the low-temperature starting process has become a research hotspot in recent years. The randomness and unclear nucleation mechanism hinder the further research of low-temperature starting. Supercooled water refers to water that remains in a liquid state below the freezing point (0℃). Under the influence of changes in the internal state of the fuel cell stack, supercooled water may randomly nucleate, and then drive the surrounding supercooled water to freeze rapidly. Supercooled water may cause a water flooding failure in the fuel cell during the low-temperature starting process, which is similar to the ice accumulation phenomenon and may block the catalyst layer. At present, there is no effective observation method for the above-mentioned ice accumulation and supercooled water phenomenon and other internal states during the low-temperature starting process of the fuel cell, and it is necessary to develop related methods and tools.

[0004] EIS (Electrochemical Impedance Spectroscopy) is a key diagnostic tool in the field of fuel cells. It can separate different scale loss mechanisms (such as ohmic polarization, activation polarization, and concentration polarization) by applying a small amplitude sinusoidal AC perturbation signal and analyzing the system response, and realize performance optimization and state management from micro-materials to system level, which is a core tool from material design to system control. In practical application, EIS is often used to diagnose the waterlogging, dryness and other membrane water content states and the life status of the single cell of the stack. The detection time of EIS is 2-5 minutes, and after many scholars' research, it has initially possessed the potential of real vehicle application. However, the internal state of the stack under low-temperature starting condition is extremely unstable compared with the normal running process, which is manifested as rapid accumulation of low-temperature ice over time and rapid increase of supercooled water with the change of membrane humidity. The above characteristics are contrary to the condition that the internal state of the system to be tested is stable required by EIS test. Therefore, the biggest problem to be solved by the prior art is how to perform the corresponding EIS test in a short time when the fuel cell is in a relatively stable internal state during the low-temperature starting process of the fuel cell. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a rapid electrochemical impedance test method applied to low-temperature starting of a fuel cell, which acquires a characteristic frequency suitable for low-temperature starting condition through a special working condition, greatly shortens the time required for EIS test, realizes rapid EIS test of the fuel cell during low-temperature starting process, and acquires internal impedance information of the fuel cell during low-temperature starting process.

[0006] The purpose of the present application is achieved by the following technical scheme:

[0007] The present application provides a rapid electrochemical impedance test method applied to low-temperature starting of a fuel cell, comprising the following steps:

[0008] S1, fuel cell reference condition determination and basic data construction: determining the EIS alternating current amplitude of the fuel cell to be tested under the rated working condition, and performing multi-condition full-frequency EIS test on the fuel cell to be tested, and synchronously collecting the physical parameters and impedance data of the fuel cell;

[0009] S2, fuel cell normal temperature fault simulation and impedance characteristic calibration based on fault decomposition: simulating the low-temperature supercooled water state of the fuel cell under the low-temperature starting condition by the normal temperature waterlogging condition, and simulating the low-temperature ice accumulation state by the cathode carbon powder blockage experiment; respectively performing full-frequency EIS test, and extracting the impedance characteristic data of the fuel cell under different degrees of waterlogging and different degrees of cathode blockage state;

[0010] S3, performing fuel cell equivalent circuit fitting and characteristic frequency extraction: building a fuel cell equivalent circuit model, using the impedance characteristic data obtained in S2 to perform model parameter fitting of the fuel cell equivalent circuit model; based on the model parameters of the fitted fuel cell equivalent circuit model, identifying the most sensitive feature frequency band set to the fault category and severity through machine learning, for subsequent rapid EIS testing;

[0011] S4, performing low-temperature start-up condition fuel cell rapid EIS test: during the low-temperature start-up process of the fuel cell, only a sinusoidal disturbance signal of the characteristic frequency band extracted in S3 is applied to the battery, the voltage and current responses are collected, and the complex impedance value is calculated to realize rapid EIS test.

[0012] Further, the S1 comprises:

[0013] S11. Determine the rated operating condition of the fuel cell under test;

[0014] S12. Perform a pre-experiment under the rated operating condition to determine the EIS alternating current amplitude under the rated operating condition;

[0015] S13. Perform full-band EIS testing on the fuel cell under test under multiple stable operating points formed by different temperature, current density, stoichiometric ratio and humidity, and simultaneously collect voltage, temperature, humidity, pressure and flow to establish basic impedance data including Nyquist and Bode plots.

[0016] Further, the S2 comprises:

[0017] S21. At room temperature, simulate water flooding caused by low-temperature supercooled water by adjusting the cathode boundary conditions, perform simulation experiments of the fuel cell under test in the low-current density range under normal temperature conditions, and perform full-band EIS testing under different water flooding conditions to obtain complete Nyquist and Bode plot data under each stable operating condition, thereby obtaining impedance characteristic data under different water flooding levels;

[0018] S22. Simulate the state of ice accumulation blocking the cathode catalyst layer under the low-temperature start-up condition of the fuel cell by performing a cathode carbon powder blocking experiment, and perform EIS testing under constant current density to establish a correspondence between the degree of cathode blocking and impedance characteristics.

[0019] Further, in S22, the cathode carbon powder blocking experiment is performed by spraying different amounts of carbon powder solution on the surface of the cathode catalyst layer to form simulated ice accumulation states with 20%, 40%, 60%, 80% and 100% pore coverage.

[0020] Further, the S3 comprises:

[0021] S31. Construct an eight-parameter fuel cell equivalent circuit model comprising a high-frequency inductance-resistance series, a medium-frequency resistance-constant phase element parallel, and a low-frequency R-L-C parallel resonant circuit;

[0022] S32. Use the impedance characteristic data of the fuel cell in different degrees of flooding and different degrees of cathode blockage state extracted by S2 respectively to perform model parameter fitting of the fuel cell equivalent circuit model through an optimization algorithm;

[0023] S33. Based on the fitted model parameters of the fuel cell equivalent circuit model, combine machine learning and statistical analysis to identify the most sensitive feature frequency band set for fault categories and states, which is used for subsequent rapid EIS testing.

[0024] Further, the S32 comprises:

[0025] Standardize and correct the full-band EIS test data of the fuel cell in different degrees of flooding and different degrees of cathode blockage state obtained in step S2;

[0026] Set parameter boundary constraints according to physical significance for the fuel cell equivalent circuit model constructed in step S31;

[0027] Use a hybrid strategy combining particle swarm optimization algorithm and gradient descent algorithm to iteratively optimize the model parameters of the fuel cell equivalent circuit model to minimize the weighted residual sum of squares between the model calculated impedance and the experimentally measured impedance;

[0028] Calculate the chi-square value, root mean square error, and average relative error of this fitting, and generate a smooth fitting curve based on a logarithmic uniform distribution frequency.

[0029] Further, the S33 comprises:

[0030] Use complex nonlinear least squares method to fit the impedance characteristic data of each group of full-band EIS test collected in S2 to the fuel cell equivalent circuit model respectively, to obtain the fuel cell impedance characteristic data under each working condition after fitting;

[0031] Based on the fuel cell impedance characteristic data under each working condition after fitting, construct a multi-dimensional full-working-condition EIS fault characteristic database covering fuel cell health benchmarks, different degrees of flooding, and different degrees of cathode blockage state;

[0032] Use the multi-dimensional full-working-condition EIS fault characteristic database as the training set, use random forest or gradient boosting tree algorithm to train the full-band fuel cell impedance characteristic data, and construct a state classification model;

[0033] By calculating the feature importance weight of each feature frequency in the state classification, feature frequencies with high discriminative power for fault categories and severity are selected, and finally a set of discrete feature frequencies including high frequency, medium frequency and low frequency is determined.

[0034] The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells provided by this invention has the following significant advantages compared with existing technologies:

[0035] 1. A breakthrough in testing speed has been achieved, meeting real-time requirements: By simplifying the traditional full-band scanning (0.1Hz~10kHz, taking 2-5 minutes) to scanning only 3-4 core characteristic frequencies, the time for a single electrochemical impedance spectroscopy test has been compressed to 10-15 seconds. This order-of-magnitude speed improvement fundamentally solves the problem that traditional methods cannot capture the rapid dynamic changes in internal states (such as ice accumulation and supercooled water) during low-temperature startup, making real-time online monitoring and diagnosis possible.

[0036] 2. Improved accuracy and specificity of low-temperature start-up diagnostics: The characteristic frequencies are not arbitrarily selected, but rather chosen from full-band data based on ambient temperature fault simulation experiments and machine learning algorithms, identifying the most sensitive frequencies to "flooding" and "icing" faults. Therefore, the limited data obtained through rapid testing has a high information concentration, directly and effectively reflecting key state changes (e.g., diagnosing icing through high-frequency impedance abrupt changes, and diagnosing water blockage through low-frequency impedance trends), avoiding interference from redundant information in full-band data, resulting in clearer diagnostic logic and more reliable conclusions.

[0037] 3. Provides a complete and transferable solution and data foundation: This invention not only provides a final rapid testing method, but also constructs a complete technical process from basic data construction → fault simulation and calibration → model fitting → feature extraction. The established mapping relationship of "fault state - equivalent circuit parameters - characteristic frequency" has universality and can provide a valuable data foundation and feature extraction paradigm for low-temperature start-up research and control strategy development of similar fuel cells.

[0038] 4. Provides direct data support for optimizing low-temperature start-up control strategies: Rapidly acquired impedance parameters (high-frequency ohmic impedance, polarization impedance) can serve as real-time feedback inputs to the control system. This enables adaptive and preventative control based on internal states (rather than solely on terminal voltage or temperature). For example, when an abnormal increase in membrane impedance is detected (indicating icing), heating or load strategies can be adjusted in advance, potentially improving the success rate of low-temperature starts, shortening start-up time, and reducing damage to the battery.

[0039] 5. The method balances the diagnostic efficiency and system complexity: while achieving the core goal of fast and accurate diagnosis, the requirement for bandwidth and sweep speed of the test hardware (such as an electrochemical workstation or a dedicated impedance spectrometer) is reduced due to the need for excitation and sampling at only a few fixed frequencies, which is beneficial for the development of low-cost, embedded online diagnostic modules and promotes the application of the technology in practical scenarios such as vehicle-mounted scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of a rapid electrochemical impedance test method for low-temperature start-up of a fuel cell according to an embodiment of the present application.

[0041] Figure 2 is a Bode plot impedance statistical diagram for EIS testing under different water-flooded conditions in an embodiment of the present application.

[0042] Figure 3 is a Bode plot phase angle statistical diagram for EIS testing under different water-flooded conditions in an embodiment of the present application.

[0043] Figure 4 is a schematic diagram of a fuel cell equivalent circuit model in an embodiment of the present application.

[0044] Figure 5 is a schematic diagram of a fuel cell equivalent circuit model parameter fitting result in an embodiment of the present application.

[0045] Figure 6 is a schematic diagram of a comparison between full-band EIS test and rapid EIS test data in step S4 in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0047] The purpose of the present application is to provide a rapid electrochemical impedance measurement method for low-temperature start-up of a fuel cell, to solve the problem of difficulty in obtaining internal information of the fuel cell under this working condition and the problem of inability to meet the performance improvement of cold start. The rapid electrochemical impedance test method for low-temperature start-up of a fuel cell proposed in the present application is universal for all stacks. In this embodiment, a certain type of fuel cell single cell is taken as an example to describe the method.

[0048] The present embodiment is a rapid electrochemical impedance test method for low-temperature start-up of a fuel cell, as shown in Figure 1 , comprising the following specific steps:

[0049] S1, fuel cell reference working condition determination and basic data construction: determine the EIS alternating current amplitude of the measured fuel cell under the rated working condition, and perform multi-working condition full frequency EIS test on the measured fuel cell, and synchronously collect the physical parameters and impedance data of the fuel cell; specifically, step S1 comprises:

[0050] S11. Determine the rated working condition of the measured fuel cell: for the measured fuel cell monomer, determine the rated working condition in combination with technical data and bench test. In this embodiment, the rated working condition is: working temperature 75℃, current density 1.0A / cm 2 , cathode / anode stoichiometric ratio 2.0 / 1.5, cathode / anode back pressure 1.1 / 1.0 Bar, and relative humidity 50%.

[0051] S12. Pre-experiment under rated working condition to determine the EIS alternating current amplitude under rated working condition: pre-experiment under rated working condition, adjust the AC disturbance amplitude with small step, finally select 5% of the load current as the best disturbance signal to balance the test signal-to-noise ratio and linearity.

[0052] S13. Multi-working condition full frequency EIS test of the measured fuel cell, synchronously collect the physical parameters and impedance data of the fuel cell: under different temperature, current density, stoichiometric ratio and humidity combination of multiple stable working condition points, use the electrochemical workstation to perform full frequency EIS test (0.1Hz~10kHz, 10 points per decade) on the measured fuel cell, synchronously collect voltage, temperature and humidity, pressure and flow, and establish basic impedance data including Nyquist diagram and Bode diagram.

[0053] S2, normal temperature fault simulation and impedance characteristic extraction of fuel cell based on fault decomposition: simulate the low-temperature supercooled water state of fuel cell under low-temperature starting condition by normal temperature water flooding working condition, and simulate the low-temperature icing state by cathode carbon powder blockage experiment; respectively perform full frequency EIS test, and extract impedance characteristic data of fuel cell under different degrees of water flooding and different degrees of cathode blockage state.

[0054] This step aims to simulate the "liquid water" and "icing" state of fuel cell under low-temperature starting by normal temperature experiment, and perform impedance characteristic extraction; specifically comprising:

[0055] S21. Under normal temperature, simulate water flooding caused by low-temperature supercooled water by adjusting the cathode boundary condition, perform simulation experiment of water flooding state of the measured fuel cell in low current density interval under normal temperature working condition, perform full frequency EIS test under different water flooding working conditions, and thus obtain impedance characteristic data under different water flooding degrees:

[0056] In the water flooding simulation experiment, the relative humidity and flow rate of the cathode / air side were adjusted to induce the fuel cell into different degrees of water flooding state at room temperature. By using higher gas relative humidity, lower gas mass flow rate and higher current density, the fuel cell can be induced into different degrees of water flooding state. For example, in the fixed current density (300 mA / cm 2 ) operating condition of the fuel cell, the cathode humidity was increased to 90%~100%, and the cathode flow rate was reduced, and the cathode stoichiometric ratio was reduced from 2 to 1.5. Compared with the condition of moderate cathode humidity (50%), the water produced by the fuel cell under the same electric density did not change, but the water brought by the cathode gas was more, and the decrease of flow rate led to the decrease of sweeping ability, which caused the water to be unable to completely flow out and remained in the cathode catalyst layer, resulting in water flooding failure. By controlling the above variables, including adjusting by reducing the cathode flow rate, increasing the cathode humidity, etc., and verifying by EIS test, the fuel cell can be induced into different degrees of water flooding state after experiencing the normal state of membrane wetting, so as to simulate and study the oxygen starvation failure condition. The water flooding condition leads to the trend of the decrease of the output voltage of the fuel cell, and the decrease of the output voltage of the fuel cell to different degrees represents different degrees of water flooding condition.

[0057] Under different water flooding conditions, the EIS test of the full frequency band was carried out by using the electrochemical workstation, the complete Nyquist graph and Bode graph data under each stable condition were obtained, and the impedance characteristic data of the fuel cell under different water flooding degrees were obtained. The results show that the high-frequency ohmic resistance remains basically unchanged; the impedance arc gradually increases in the medium and low frequency range, and the low-frequency impedance rises; the tail of the Nyquist graph extends outward in the low frequency, and the phase valley deepens. These characteristics show that the hindering of the water flooding state to the mass transfer process is significantly enhanced. The impedance and phase angle data of the Bode graph are shown in Figure 2 、 Figure 3 .

[0058] S22. The carbon powder blocking experiment of the fuel cell cathode was used to simulate the state of ice blocking the cathode catalyst layer under the low-temperature starting condition of the fuel cell, and the EIS test under the constant current density was carried out, and the corresponding relationship between the cathode blocking degree and the impedance characteristics was established:

[0059] The embodiment realizes the equivalent simulation of different degrees of "ice blockage" by spraying carbon powder on the cathode side of the membrane electrode (MEA) at room temperature. Six fuel cell monomers with consistent initial open circuit voltage and performance are selected to construct a gradient blockage test scheme including one control group and five experimental groups. By spraying different amounts of carbon powder solution, carbon powder deposition layers with 20%, 40%, 60%, 80%, and 100% pore coverage are formed to simulate different degrees of "cathode blockage" and physically simulate different degrees of ice crystallization in the cathode catalyst layer under low-temperature starting conditions.

[0060] Specifically: six fuel cell monomer samples with consistent initial open circuit voltage and performance are selected, and all sample membrane electrodes are sandwich structures (anode and cathode catalyst layers are sprayed on both sides of the proton exchange membrane). One of them is used as a normal control group (untreated), and the remaining five are used as experimental groups to simulate different degrees of cathode blockage. Carbon powder is dissolved in ethanol at a predetermined mass concentration and ultrasonically oscillated for 30-60 minutes to form a uniform suspension. Under the condition that the membrane electrode temperature is controlled at 85°C, different amounts of carbon powder solution are uniformly sprayed on the surface of the cathode catalyst layer of the experimental group samples. By precisely controlling the spraying amount, carbon powder deposition layers with 20%, 40%, 60%, 80%, and 100% pore coverage are formed in the five experimental groups to simulate the gradient blockage state caused by low-temperature ice accumulation. In this embodiment, the 85°C temperature can accelerate ethanol evaporation, avoiding liquid droplet residue, while preventing proton exchange membrane crease damage. The carbon powder stably adheres after spraying and does not fall off during subsequent gas purging and operation.

[0061] Under stable conditions at room temperature, full-band EIS testing is performed on the six samples (1 control group + 5 experimental groups), focusing on the analysis of: high-frequency impedance changes: extracting ohmic impedance to analyze the changes in fuel cell ohmic impedance caused by different blockage degrees; middle and low-frequency impedance changes: analyzing the change trend of charge transfer impedance and mass transfer impedance to analyze the influence of oxygen starvation failure of different blockage degrees on fuel cell low-frequency proton transfer impedance; impedance spectrum feature extraction: recording the Nyquist and Bode graph features under different blockage degrees.

[0062] Based on the test data, the quantitative relationship between the blockage degree and the impedance characteristics is established:

[0063] High-frequency impedance law: as the blockage degree increases from 0% to 100%, the high-frequency ohmic impedance shows a monotonous decreasing trend, and the change amount and the blockage ratio approximately show a quasi-linear relationship; physical explanation: this law is consistent with the impedance reduction phenomenon caused by low-temperature ice blockage gas transmission path, verifying the effectiveness of the room temperature carbon powder blockage simulation; similar high-frequency impedance change trend (first decrease and then increase) is observed in the -10°C low-temperature starting experiment, further confirming the rationality of the simulation method.

[0064] S3, performing fuel cell equivalent circuit fitting and characteristic frequency extraction: an equivalent circuit model of the fuel cell is built, the impedance characteristic data obtained in S2 are used to perform model parameter fitting of the equivalent circuit model of the fuel cell, and the most sensitive feature frequency band set to the fault category and severity is identified through machine learning, which is used for subsequent rapid EIS testing.

[0065] This step aims to quantify and separate different impedances and extract high, medium and low frequency characteristic frequencies according to the collected impedance data. The core role and significance of the equivalent circuit is that it simplifies the complex fuel cell system involving multiple physical fields (electrochemistry, fluid mechanics, thermodynamics) into a circuit network composed of standard electronic components (such as resistors, capacitors, inductors, etc.), thereby quantifying the internal information of the system. Specifically, step S3 includes:

[0066] S31. Building an equivalent circuit model of the fuel cell:

[0067] The equivalent circuit is preferably an improved Randles circuit, as shown in Figure 4 , including electrolyte high-frequency resistance, double-layer capacitance, charge transfer resistance, and diffusion impedance elements; when there is water flooding or oxygen starvation, a mass transfer resistance is further introduced to reflect additional mass transfer loss, and an eight-parameter fuel cell equivalent circuit model is constructed, which includes high-frequency inductance resistance in series, medium-frequency resistance-constant phase element in parallel, and low-frequency R-L-C parallel resonance circuit.

[0068] The semi-empirical formula of impedance is:

[0069] ;

[0070] In the formula, w is the angular frequency; j is the imaginary unit; R is the ohmic resistance; L is the high-frequency inductance; a is the dispersion index; is a constant phase element, which is a specific expression of the CPE element; R ct is the charge transfer resistance; C represents the equivalent capacitance; L mt is the low-frequency inductance; R mt is the mass transfer resistance. The model is used to fit the experimental data of EIS testing, and chi-square test is used to analyze the results.

[0071] S32. Using the impedance characteristic data of the fuel cell in different degrees of water flooding and different degrees of cathode blockage extracted in S2, the model parameter fitting of the equivalent circuit model of the fuel cell is performed through an optimization algorithm:

[0072] The full-band EIS test data of the fuel cell in different degrees of flooding and different degrees of cathode blockage state obtained in step S2 are standardized and corrected, effective frequency band data is screened and retained to prevent excessive inductance from causing poor fitting accuracy of the equivalent circuit; the fuel cell equivalent circuit model constructed in step S31 is set with parameter boundary constraints according to physical significance; a hybrid strategy combining particle swarm optimization algorithm and gradient descent algorithm is used to iteratively optimize the model parameters of the fuel cell equivalent circuit model to minimize the weighted residual sum of squares between the model calculation impedance and the experimental measurement impedance; after completing the model parameter fitting, the chi-square value, root mean square error and average relative error of this fitting are calculated to evaluate the fitting goodness, and a smooth fitting curve is generated based on the logarithmic uniform distribution frequency, and finally the data results including the smooth fitting data worksheet, the original experimental data worksheet and the embedded optimal parameter set, error analysis results and Nyquist / Bode comparison are derived. The model parameter values (such as ohmic resistance, charge transfer resistance, etc.) of the fuel cell equivalent circuit model under each working condition are obtained. These model parameter values reflect the electrochemical characteristics of the battery under different states.

[0073] The main role of this step is to realize the physical meaning decoupling of impedance spectrum: by converting the EIS spectrum into specific resistance and capacitance parameters, the high-frequency ohmic impedance change and the low-frequency mass transfer impedance change caused by carbon powder blockage (simulating ice accumulation) can be effectively separated and quantified, as shown in Figure 5 , thereby establishing a mapping relationship between the blockage degree and the specific physical parameters inside the battery, providing a quantitative theoretical basis for the classification of working conditions and the extraction of characteristic frequencies.

[0074] S33. Feature frequency extraction based on sensitivity analysis:

[0075] The purpose of this step is to identify the most sensitive feature frequency band set to the fault category and state (severity) based on the fitted model parameters of the fuel cell equivalent circuit, combined with machine learning and statistical analysis, for subsequent rapid EIS testing.

[0076] Specifically, each set of full-frequency EIS test data collected in the S2 step is fitted to the above-mentioned fuel cell equivalent circuit model by using a complex nonlinear least squares method to obtain impedance characteristic data of the fuel cell under each operating condition. Water flooding and blockage mainly affect the mass transfer and ohmic impedance, and at the same time, subtle changes in the charge transfer impedance can be observed in the medium frequency region (1-100 Hz). By fitting the parameter changes, the medium frequency characteristic frequency sensitive to the charge transfer is extracted. Based on the full-frequency EIS test data of each operating condition after fitting, a multi-dimensional full-condition EIS fault feature database covering the fuel cell health benchmark, different degrees of water flooding and different degrees of cathode blockage state is constructed. Taking the multi-dimensional full-condition EIS fault feature database as the training set, taking the operating condition as the label, and taking the real and imaginary parts of the impedance at each frequency point as the features, an integrated learning algorithm such as random forest or gradient boosting tree is used for training to construct a state classification model. The algorithm automatically calculates and outputs the contribution degree (feature importance weight) of each frequency point to the correct classification. The higher the weight of the frequency point, the greater the role it plays in distinguishing different fault states. By calculating the feature importance weight of each frequency point in state classification, 3-4 characteristic frequency bands with high recognition of fault categories and severity are selected. Ensure that the selected 3-4 characteristic frequency bands can represent the key dynamics of high frequency (reflecting ohmic impedance), medium frequency (reflecting charge transfer impedance) and low frequency (reflecting mass transfer impedance) respectively, and finally determine a set of discrete characteristic frequency band set containing high frequency, medium frequency and low frequency.

[0077] As preferred, the statistical screening uses analysis of variance (ANOVA) to evaluate the significance of impedance differences under different simulated operating conditions; and then principal component analysis (PCA) is used to reduce the dimensionality of the data, and the frequency points corresponding to the principal components with a contribution rate of >85% are selected.

[0078] S4, perform a low-temperature start-up operating condition fuel cell rapid EIS test: during the low-temperature start-up process of the fuel cell, only a sinusoidal disturbance signal at the characteristic frequency band in the characteristic frequency band set extracted in S3 is applied to the battery, the voltage and current responses are collected, and the complex impedance value is calculated to realize rapid EIS testing.

[0079] S41. Low-temperature environment pretreatment and state initialization:

[0080] The fuel cell stack is placed in an environmental chamber, and a purging process is performed to remove residual moisture in the flow channel and membrane electrode. Subsequently, the temperature of the environmental chamber is lowered to a preset cold start temperature (-10℃-30℃), and the temperature is kept at this temperature for 2-4 hours to ensure that each part inside the fuel cell stack reaches a thermal equilibrium state and is in an initial state without liquid water residue.

[0081] S42. Constant current cold start loading:

[0082] After the low-temperature soaking is completed, the reaction gas is introduced into the fuel cell, and the fuel cell is discharged under a constant current density (0.05 A / cm 2 0.2 A / cm 2 ) working condition according to a preset cold start strategy, and enters the cold start process.

[0083] S43. Fast EIS real-time scanning based on characteristic frequencies:

[0084] During the cold start process, the fast EIS test program is triggered at a preset time interval. During the signal application process, the EIS test no longer performs full-frequency scanning (usually 0.1 Hz-10 kHz), but only applies a sinusoidal disturbance current signal of 3-4 characteristic frequency bands extracted in the S3 step to the fuel cell; the voltage and current response signals at each characteristic frequency are collected, and the corresponding complex impedance values are calculated.

[0085] S44. Impedance parameter analysis and ice accumulation state diagnosis:

[0086] Since the test is only for a small number of characteristic frequencies, the single EIS test period is compressed to within 10-15 seconds. The system uses the collected impedance characteristic data to calculate the high-frequency impedance (membrane impedance) and polarization impedance of the fuel cell at the current time. An effective data acquisition method is provided for real-time state observation of the subsequent fuel cell low-temperature start. The general diagnosis logic is to judge whether there is icing phenomenon in the membrane electrode by monitoring the mutation of the high-frequency impedance; and to judge whether the liquid water blockage, i.e., the existence of supercooled water, occurs in the catalyst layer or the gas diffusion layer by the change trend of the low-frequency impedance.

[0087] The following is the effectiveness experiment and result analysis of the method of the present embodiment:

[0088] To prove the effectiveness and accuracy of the method proposed in the present application in practical application, before performing the rapid EIS test of the fuel cell in the low-temperature starting condition, the effectiveness verification of the rapid EIS test method in the normal-temperature condition is performed: after extracting the characteristic frequency, the operating conditions of the fuel cell are changed, and the EIS tests in the full frequency domain and the characteristic frequency are performed in the operating conditions different from all the previous operating conditions, and the difference between the fuel cell impedances measured by the two tests is compared. The above test experiment is performed for multiple times to verify the effectiveness of the EIS test using the characteristic frequency. In different operating conditions of the fuel cell, such as high-humidity low-oxygen, low-humidity low-oxygen, high-humidity high-oxygen, and low-humidity high-oxygen, the fuel cell is in different fault conditions such as water flooding and oxygen starvation. Combined with the external characteristics of the fuel cell, the fault is judged. Then, the full-band and rapid EIS tests are compared to verify whether the characteristic frequency EIS test can reflect whether the fuel cell is in a fault state by means of a small number of frequency bands. In completely new operating conditions different from the modeling data (such as changing the temperature and humidity combination), the traditional full-band EIS (as the true value) and the rapid EIS test of the present application are performed at the same time. Figure 6 As shown in the figure, the full-band EIS and rapid EIS test data of the fuel cell under the rated operating condition are compared. The comparison shows that the impedance key parameters measured by the rapid EIS are highly consistent with the full-band test results, and can effectively distinguish different fault states such as high-humidity low-oxygen and low-humidity high-oxygen, proving that only collecting the characteristic frequency can meet the fault diagnosis requirements, and at the same time meet the stringent requirements of real-time (10-15s) in the low-temperature starting process. Compared with the traditional full-band EIS test which takes 2-3 minutes, the present embodiment only needs to perform rapid EIS measurement at the extracted 3-4 characteristic frequency bands, so as to complete the fault recognition. The single test time can be shortened to 10-15 seconds, meeting the real-time requirement of rapid diagnosis in the low-temperature starting process of the fuel cell.

[0089] To study the low-temperature starting difficulty of the fuel cell caused by the liquid water and ice accumulation of the membrane electrode in the low-temperature cold starting process, the comparative experiment is performed under the constant current density cold starting condition, and the key impedance parameters are recorded by combining the rapid EIS test. The cold starting experiment is repeated for multiple times under the same experimental conditions, and the rapid EIS test is performed at different cold starting operating points, and the EIS results under different cold starting operating conditions are recorded. Through the rapid EIS test corresponding to different cold starting operating conditions, the impedance characteristics of the fuel cell under different ice accumulation states can be obtained. This provides an experimental basis for the diagnosis and real-time control of the low-temperature starting of the fuel cell in the subsequent stack fuel cell low-temperature starting condition.

Claims

1. A rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells, characterized in that, Includes the following steps: S1. Determination of fuel cell baseline operating conditions and construction of basic data: Determine the amplitude of EIS alternating current of the fuel cell under test under rated operating conditions, and conduct EIS tests on the fuel cell under test under multiple operating conditions and full frequency bands, while simultaneously collecting physical parameters and impedance data of the fuel cell. S2. Conduct room temperature fault simulation and impedance characteristic calibration of fuel cells based on fault decomposition: simulate the low temperature subcooled water state of fuel cell under low temperature start-up conditions by room temperature water flooding condition, simulate the low temperature ice accumulation state by cathode carbon powder blockage experiment, and conduct full-band EIS test to extract impedance characteristic data of fuel cells under different degrees of water flooding and different degrees of cathode blockage. S3. Perform fuel cell equivalent circuit fitting and characteristic frequency extraction: Build a fuel cell equivalent circuit model, and use the impedance characteristic data obtained in S2 to fit the model parameters of the fuel cell equivalent circuit model; based on the model parameters of the fitted fuel cell equivalent circuit model, use machine learning to identify the set of characteristic frequency bands that are most sensitive to fault categories and severity for subsequent rapid EIS testing. S4. Perform rapid EIS test on fuel cell under low temperature start-up conditions: During the low temperature start-up of the fuel cell, only apply the sinusoidal disturbance signal of the characteristic frequency band extracted in S3 to the battery, collect the voltage and current response, calculate the complex impedance value, and realize rapid EIS test.

2. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 1, characterized in that, S1 includes: S11. Determine the rated operating conditions of the fuel cell under test; S12. Conduct a preliminary experiment under rated operating conditions to determine the amplitude of the EIS alternating current under rated operating conditions; S13. Under multiple stable operating conditions composed of different combinations of temperature, current density, stoichiometry and humidity, perform full-band EIS testing on the fuel cell under test, and simultaneously collect voltage, temperature, humidity, pressure and flow rate to establish basic impedance data including Nyquist plot and Bode plot.

3. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 1, characterized in that, S2 includes: S21. At room temperature, the flooding caused by low-temperature supercooled water is simulated by adjusting the cathode boundary conditions. The flooding state of the fuel cell under test in the low current density range under room temperature conditions is simulated. Full-band EIS test is carried out under different flooding conditions to obtain complete Nyquist and Bode plot data under each stable condition, thereby obtaining impedance characteristic data under different flooding degrees. S22. The state of ice accumulation blocking the cathode catalyst layer under low-temperature start-up conditions of fuel cells is simulated by carbon powder blockage experiment of fuel cell cathode, and EIS test is carried out under constant current density to establish the correspondence between the degree of cathode blockage and impedance characteristics.

4. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 3, characterized in that, In S22, the cathode carbon powder blockage experiment is carried out by spraying different amounts of carbon powder solution onto the surface of the cathode catalyst layer to form a simulated ice accumulation state with 20%, 40%, 60%, 80%, and 100% pore coverage.

5. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 1, characterized in that, S3 includes: S31. Construct an eight-parameter equivalent circuit model for a fuel cell, including a series high-frequency inductor and resistor, a parallel mid-frequency resistor and constant phase angle element, and a parallel low-frequency RLC resonant circuit. S32. Using the impedance characteristic data of the fuel cell extracted in S2 under different degrees of water flooding and different degrees of cathode blockage, the model parameters of the equivalent circuit model of the fuel cell are fitted by an optimization algorithm. S33. Based on the model parameters of the fitted fuel cell equivalent circuit model, combined with machine learning and statistical analysis, the set of characteristic frequency bands most sensitive to fault categories and states is identified for subsequent rapid EIS testing.

6. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 5, characterized in that, S32 includes: The full-band EIS test data of the fuel cell under different degrees of water flooding and different degrees of cathode blockage obtained in step S2 are standardized and corrected. For the fuel cell equivalent circuit model constructed in step S31, parameter boundary constraints are set according to physical meaning; A hybrid strategy combining particle swarm optimization and gradient descent is used to iteratively optimize the model parameters of the equivalent circuit model of the fuel cell in order to minimize the weighted sum of squared residuals between the calculated impedance and the experimentally measured impedance. Calculate the chi-square value, root mean square error, and average relative error of the fit, and generate a smooth fitting curve based on the logarithmic uniform distribution frequency.

7. The rapid electrochemical impedance spectroscopy method for low-temperature start-up of fuel cells as described in claim 5, characterized in that, S33 includes: The impedance characteristic data of each group of full-band EIS tests collected by S2 are fitted into the equivalent circuit model of the fuel cell using the complex nonlinear least squares method to obtain the fuel cell impedance characteristic data under each working condition after fitting. Based on the fitted fuel cell impedance characteristic data for each operating condition, a multi-dimensional full-condition EIS fault characteristic database covering fuel cell health benchmark, different degrees of water flooding and different degrees of cathode blockage is constructed. Using the multidimensional full-condition EIS fault feature database as the training set, a state classification model is constructed by training the full-frequency fuel cell impedance feature data using random forest or gradient boosting tree algorithms. By calculating the feature importance weight of each feature frequency in the state classification, feature frequencies with high discriminative power for fault categories and severity are selected, and finally a set of discrete feature frequencies including high frequency, medium frequency and low frequency is determined.

Images