A method for determining fuel cell stack flow channel parameters

Abstract

The application provides a kind of determination method of fuel cell stack flow channel parameter, method includes the following steps: collecting the dimensional dispersion of flow field plate in manufacturing process, diffusion layer compaction non-uniformity, sealing springback difference and other variables, and disturbance data such as load fluctuation, gas temperature and humidity, environmental pressure change in operating environment, establish joint uncertainty set;Set channel width, rib width, channel depth, group parameters as to be optimized flow channel parameters, construct performance index set including pressure drop, saturation, temperature gradient, voltage fluctuation, oxygen utilization rate, and set corresponding opportunity constraint confidence threshold;Based on water, gas, heat, electric multi-physical coupling model, the mapping relationship between parameters, uncertainty and performance is constructed, and a random coupling evaluation model is formed;Solve target parameter combination by opportunity constraint optimization algorithm.Considering average performance and robustness, the consistency, reliability and manufacturability of fuel cell flow channel design can be improved.

Description

Technical Field

[0001] This invention relates to the field of fuel cell stack design technology, and in particular to a method for determining the flow channel parameters of a fuel cell stack. Background Technology

[0002] The fuel cell stack is a key component that converts hydrogen and oxygen from the air into electrical energy through an electrochemical reaction. Its performance and lifespan largely depend on the design and parameter settings of the internal flow channel structure of the stack. The flow channel parameters directly affect gas distribution, water management characteristics, and pressure drop levels, thereby determining the power output efficiency and operational stability of the stack.

[0003] In existing technologies, the determination of flow channel parameters mainly relies on single-point optimization under fixed operating conditions, that is, adjusting the flow channel size under nominal operating conditions through simulation or experiment to meet pressure drop and mass transfer requirements. However, this method has the following shortcomings:

[0004] First, the above methods ignore the tolerance deviations in the manufacturing process, such as insufficient processing precision of the flow field plate, uneven compaction of the gas diffusion layer, and differences in the rebound of the sealing ring. These factors can cause the actual flow channel geometry to deviate from the design value. Second, they lack consideration for the uncertainties of the operating environment, such as frequent load fluctuations, seasonal changes in gas temperature and humidity, and changes in ambient air pressure with altitude.

[0005] These uncertainties often cause abnormal water and heat distribution inside the fuel cell stack, resulting in droplet blockage, waterlogging, or voltage fluctuations, which in turn lead to performance degradation and shortened lifespan. Therefore, we propose a method for determining the flow channel parameters of a fuel cell stack. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for determining the flow channel parameters of a fuel cell stack, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for determining the flow channel parameters of a fuel cell stack includes the following steps:

[0009] S1. Collect manufacturing process data and operating environment data to establish an uncertainty set; determine the flow channel parameter set and performance index set, and set the opportunity constraint threshold that enables the performance index to meet the constraints under a predetermined confidence level.

[0010] S2. Based on the multi-physical mechanisms of water, heat, gas and electricity, establish the mapping relationship between flow channel parameters, uncertainties and performance indicators, form a stochastic coupled evaluation model with parameter and uncertainty samples as input and performance indicator distribution as output, and use experimental data for calibration and accuracy verification.

[0011] S3. Establish opportunity constraints for performance indicators and define optimization objective functions, including at least the expected optimal performance and quantile performance improvement, so as to balance average performance and robustness under extreme conditions.

[0012] S4. Optimize the stochastic coupling evaluation model using random sampling, scenario approximation, or determinism methods to obtain a set of candidate parameters that satisfy the chance constraints. Then, based on the feasibility probability, expected performance, quantile redundancy, and manufacturing cost, perform a comprehensive ranking to determine the target parameter solution.

[0013] S5. Verify the performance distribution of the target parameter solution under independent sample sets and extreme working conditions, calculate the confidence interval and safety margin; when the performance touches the constraint threshold, trigger backtracking adjustment and generate the corresponding manufacturing tolerance window and assembly clamping force recommendation range.

[0014] S6. Deconstruct and solidify the verified target parameters into a standard version, establish the correspondence between parameters, tolerances, and operating conditions, and form a judgment procedure that includes inspection thresholds, factory consistency judgment, in-service monitoring, and anomaly rollback strategies for mass production and field applications.

[0015] S1 specifically refers to: acquiring statistical data on the manufacturing process, including dimensional variations in the flow field plate, non-uniformity of gas diffusion layer compaction, and manufacturing uncertainties such as differences in seal springback; acquiring historical data on the operating environment, including operational uncertainties such as load fluctuations, inlet and outlet gas temperature and humidity, and changes in ambient air pressure.

[0016] By combining manufacturing uncertainties with operational uncertainties, an uncertainty set is established as the basis for subsequent evaluation and optimization; the set of flow channel parameters to be optimized is determined, including at least channel width, rib width, channel depth, channel grouping parameters, and target pressure drop quantile;

[0017] Set a set of performance indicators, which includes at least pressure drop, local liquid phase saturation, temperature gradient, single cell voltage fluctuation, and cathode oxygen utilization rate; set a target confidence threshold for opportunity constraints, which is not lower than a set value.

[0018] S2 specifically refers to: establishing a mapping relationship between flow channel parameters, uncertainty samples, and performance indicators based on multiple physical mechanisms of water, heat, gas, and electricity;

[0019] A stochastic coupling evaluation model is formed, which takes flow channel parameters and uncertainty samples as inputs and outputs corresponding predicted values ​​of performance indicators.

[0020] The random coupling evaluation model is calibrated using in-service test data or bench test data, and the model accuracy is verified within the set operating conditions to ensure that the prediction results meet the preset accuracy requirements.

[0021] S3 specifically refers to: establishing opportunity constraints such that, under a set of uncertainties, the probability of a performance indicator meeting a set upper or lower limit is not less than a confidence threshold.

[0022] Under the premise of satisfying the opportunity constraints, define an optimization objective function, wherein the optimization objective includes at least minimizing the expected loss;

[0023] The worst quantile performance improvement is introduced into the optimization objective, where the worst quantile is the 95th quantile performance, to improve overall performance and consistency.

[0024] S4 specifically involves using a random sampling method to solve the chance constraints of the stochastic coupled evaluation model.

[0025] The stochastic coupling evaluation model is supplemented by a scenario approximation method or an equivalent deterministic method to obtain a set of candidate parameters that satisfy the chance constraints.

[0026] The candidate parameter set is sorted based on factors including the feasibility probability at the target confidence level, the expected loss, and the manufacturing and implementation cost.

[0027] Choose the parameter combination that ranks highly and has excellent overall performance as the objective parameter solution.

[0028] S5 specifically involves: verifying the solution of the target parameters on a sampling set independent of the solution process, and obtaining confidence intervals for pressure drop, liquid saturation, temperature gradient and voltage fluctuation;

[0029] Boundary verification of the objective parameter solution is performed under extreme combined operating conditions to evaluate whether each performance index satisfies the chance constraint conditions.

[0030] When any performance index touches the opportunity constraint threshold under boundary conditions, a backtracking adjustment process is triggered to update the target parameter solution.

[0031] While backtracking and adjusting, a manufacturing tolerance suggestion window and an assembly clamping force suggestion range are generated to match the target parameter solution.

[0032] S6 specifically involves: solidifying the target parameters for verification into flow channel parameter templates; establishing the correspondence between parameters, tolerances, and operating conditions to form a judgment procedure;

[0033] The judgment procedure should clearly define the incoming material inspection items, assembly inspection items, and the threshold for judging factory consistency.

[0034] The judgment procedure further specifies in-service monitoring indicators and anomaly rollback strategies to ensure that the target parameter solution can be implemented in mass production and field applications.

[0035] The beneficial effects of this invention are as follows:

[0036] This invention systematically collects and models manufacturing process errors (dispersion in flow field plate dimensions, non-uniformity of diffusion layer compaction, and differences in seal rebound) and operating environment fluctuations (load fluctuations, temperature and humidity changes, and air pressure changes). It not only provides statistical distribution and outlier handling methods, but also establishes a normalization and correlation matrix model, thereby ensuring that the uncertainty set has true representativeness. Compared with traditional flow channel solutions based on single-point design, this invention can maintain stable performance with more than 95% operating condition coverage, significantly improving consistency across batch production and reliability across application scenarios.

[0037] This invention, for the first time, establishes a one-to-one correspondence between the set of flow channel parameters (channel width, rib width, depth, and grouping method) and the set of performance indicators (pressure drop, liquid saturation, temperature gradient, voltage fluctuation, and oxygen utilization rate), clarifying the direct impact mechanism of geometric parameter changes on gas transport, heat transfer management, and electrochemical reactions. For example, parallel grouping can reduce the overall pressure drop and improve the uniformity of gas distribution, while series grouping can extend the gas residence time and improve oxygen utilization rate. Through this causal mapping relationship, the shortcomings of traditional black-box optimization that cannot be explained are avoided, making the design verifiable and engineering-guided.

[0038] This invention proposes a chance-based constraint mechanism based on Monte Carlo simulation, explicitly requiring each performance index to meet constraints with a confidence level of no less than 95%. Through a comprehensive constraint formula, it ensures that if a single index fails to meet the standard, the entire stack is deemed unqualified. This effectively avoids the hidden danger of average performance being acceptable but failure occurring under extreme operating conditions, a problem present in traditional methods. This ensures the fuel cell stack maintains high robustness under varying operating environments. By combining multi-index constraints with chance-based constraint judgment, this invention can effectively suppress liquid phase flooding while ensuring reasonable voltage drop, reducing the internal temperature gradient of the fuel cell stack, minimizing voltage fluctuations, and ensuring high oxygen utilization. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of a method for determining the flow channel parameters of a fuel cell stack according to the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Example 1: As Figure 1 As shown in the figure, this embodiment provides a method for determining the flow channel parameters of a fuel cell stack, including the following steps:

[0042] S1. Uncertainty Modeling and Index Setting: Collect manufacturing process data (including flow field plate size dispersion, diffusion layer compaction non-uniformity, and seal springback differences) and operating environment data (including load fluctuation, gas temperature and humidity, and ambient air pressure changes) to establish an uncertainty set; determine the flow channel parameter set and performance index set, and set opportunity constraint thresholds that enable the performance indexes to meet constraints under predetermined confidence conditions.

[0043] S2. Construction of stochastic coupling evaluation model: Based on the multi-physical mechanism of water, heat, gas and electricity, establish the mapping relationship between flow channel parameters, uncertainties and performance indicators, form a stochastic coupling evaluation model with parameter and uncertainty samples as input and performance indicator distribution as output, and use experimental data for calibration and accuracy verification.

[0044] S3. Opportunity Constraints and Robustness Objective Definition: Establish opportunity constraints for performance indicators and define the optimization objective function, which should include at least the expected optimal performance and quantile performance improvement, so as to balance average performance and robustness under extreme conditions.

[0045] S4. Opportunity Constraint Optimization Solution: The stochastic coupling evaluation model is optimized using random sampling, scenario approximation, or determinism methods to obtain a set of candidate parameters that satisfy the opportunity constraints. The solution is then determined by comprehensively ranking the feasible probability, expected performance, quantile redundancy, and manufacturing cost.

[0046] S5. Independent Validation and Boundary Scan: Validate the performance distribution of the target parameter solution under independent sample sets and extreme conditions, calculate the confidence interval and safety margin; when the performance touches the constraint threshold, trigger backtracking adjustment and generate a corresponding manufacturing tolerance window and assembly clamping force recommendation range.

[0047] S6. Parameter solidification and release: The verified target parameters are solidified into standard versions, and the correspondence between parameters, tolerances and operating conditions is established. Judgment procedures including inspection thresholds, factory consistency judgment, in-service monitoring and abnormal rollback strategies are formed for mass production and field applications.

[0048] S1 specifically includes the following sub-steps:

[0049] S110. Obtain manufacturing process statistics: Obtain historical sample data of the fuel cell stack manufacturing process, including:

[0050] Dimensional variation of flow field plates: Inspect no less than 50 batches of flow field plates using a coordinate measuring machine (CMM), statistically analyze the deviation range of channel width and depth, and form a dimensional error distribution curve;

[0051] Gas diffusion layer compaction nonuniformity: Using standard compression tests and thickness measuring instruments, the compaction thickness difference of no less than 30 samples was measured to obtain the mean and standard deviation of nonuniformity;

[0052] Seal rebound variation: The residual deformation rate of no less than 20 batches of seals is measured through compression rebound test to form the distribution range of rebound performance.

[0053] Finally, a manufacturing uncertainty database was established, with each indicator providing its mean, standard deviation, and maximum deviation to ensure data quantifiability and repeatability. It is worth noting that before establishing the manufacturing uncertainty model, the collected sample data from the flow field plate, diffusion layer, and seals were fitted with a normal distribution, and the distribution's rationality was verified using a chi-square test. If the distribution deviated from normality, an empirical distribution function was used. Outliers were removed using the 3σ criterion or completed using linear interpolation to ensure statistical stability and modeling consistency.

[0054] S120. Obtain historical operating environment data: Collect environmental data of the fuel cell stack during long-term operation or bench testing, including:

[0055] Load power curve: recorded by the fuel cell power sensor, with a minimum of 1000 hours of operating data;

[0056] Inlet and outlet gas temperature and humidity: collected in real time by temperature and humidity sensors, covering typical summer and winter operating conditions, with a sampling interval of ≤1 minute;

[0057] Ambient air pressure: Monitored by barometer for at least one year, recording daily averages and extreme values. Data are uniformly stored as time series files, and the mean, variance, and typical extreme values ​​are extracted to characterize operational uncertainties. The environmental data acquisition adopts a sampling frequency of 1Hz, and the original time series is filtered by moving average and imputed for missing values ​​to ensure data continuity and noise resistance.

[0058] S130. Establish an uncertainty set: Combine the data obtained in S110 and S120 to form a joint uncertainty set: represent the machining dimensional error using an interval set, such as the channel width error. ;

[0059] Fluctuations in operating load are represented by a probability distribution, such as a normal distribution with a mean of 1.0 and a variance of 0.1. Pressure changes are represented by the maximum-minimum values ​​of measured time series data as boundaries. Each uncertainty is input into the model in the form of an "interval or distribution," ensuring the set definition is specific and operable. Each uncertainty variable is normalized before entering the joint set, and its correlation is modeled using a correlation coefficient matrix, thus avoiding inaccuracies caused by the independence assumption. This set is representative and can accurately reflect the characteristics of coupled disturbances between manufacturing and the environment.

[0060] S140. Determine the set of flow channel parameters: Set the flow channel parameters required for optimization, including:

[0061] Channel width: 0.8-1.2mm; Rib width: 0.5-1.0mm; Channel depth: 0.5-1.5mm; Channel grouping parameters: parallel or series arrangement;

[0062] Target pressure drop quantile: 95th percentile value ≤ 500 Pa; the range is determined by industry experience and preliminary simulation results, so that the optimization does not deviate from engineering common sense, but covers the design space.

[0063] A causal mapping was performed to investigate the influence mechanism between channel geometry parameters and performance indicators. Parallel channels help reduce overall pressure drop and improve gas distribution uniformity, while series channels can extend the reaction path and improve oxygen utilization.

[0064] S150. Determine the set of performance indicators: Define a set of indicators for evaluating the performance of the fuel cell flow channel, including at least: pressure drop. The pressure was obtained through numerical simulation or experimental pressure measurement, with a threshold of 500 Pa.

[0065] Local liquid phase saturation (Dimensionless 0-1): Measured by numerical simulation or neutron imaging, threshold ≤ 0.7;

[0066] Temperature gradient (K / cm): Obtained by thermocouple array or infrared thermometer, threshold ≤5K / cm;

[0067] Single cell voltage fluctuation (mV): collected by the stack voltage sensor, threshold ≤20mV;

[0068] Cathode oxygen utilization rate (%): Calculated from inlet and outlet gas concentration measurements, threshold ≥85%;

[0069] Each indicator has a measurement method, unit, and threshold, making it verifiable and quantifiable; Example indicator statistics table (used to support subsequent probability determination):

[0070]

[0071] S160, Set the confidence threshold for opportunity constraints:

[0072] Based on the uncertainty set established in S130 and the performance indicators in S150, opportunity constraints are constructed: under uncertainty conditions, the probability that each performance indicator meets the threshold is not lower than a preset confidence threshold. To ensure engineering applicability, this embodiment sets the confidence threshold to 95%.

[0073] Specifically, this includes: Sample size and statistical methods: Monte Carlo random sampling is used to generate no fewer than N=1000 samples from the uncertainty set U. For each candidate parameter combination... The S220 stochastic coupling evaluation model was used to calculate the performance index sequence. The empirical probability of each performance index meeting the threshold was statistically analyzed and used as the basis for determining the chance constraint.

[0074] Single-index probability constraint expression:

[0075]

[0076] Indicates pressure drop ( The probability of a Pa value less than or equal to 500 Pa is greater than or equal to 95%.

[0077]

[0078] Indicates local liquid phase saturation ( The probability of 0.70 being less than or equal to 0.70 is greater than or equal to 95%.

[0079]

[0080] Representing temperature gradient The probability of being less than or equal to 5 K / cm is greater than or equal to 95%.

[0081]

[0082] Indicates single-cell voltage fluctuation The probability of being less than or equal to 20mV is greater than or equal to 95%.

[0083]

[0084] Indicates cathode oxygen utilization rate A probability greater than or equal to 85% must be greater than or equal to 95%.

[0085] in For pressure drop, For local liquid phase saturation, For temperature gradient, Ripple represents voltage fluctuation. For cathode oxygen utilization rate; comprehensive judgment criteria: to ensure that all indicators meet the confidence level requirements, a comprehensive chance constraint is set:

[0086]

[0087] in Indicates pressure drop The empirical probability of satisfying the threshold; Indicates local liquid phase saturation The empirical probability of satisfying the threshold; Representing temperature gradient The empirical probability of satisfying the threshold; This represents the empirical probability that the voltage fluctuation Ripple meets the threshold. Indicates oxygen utilization rate The empirical probability of satisfying the threshold.

[0088] Formula logic: Calculate the empirical probability for each of the five performance indicators; take the minimum value among these five probabilities; require that this minimum value is still not lower than 95%; if the condition is met, the parameter combination is determined to be "feasible", otherwise it is determined to be "infeasible".

[0089] For example: with N=1000 samples, a candidate solution is obtained as follows: =0.972, =0.964, =0.983, =0.953, =0.961, then the minimum value 0.953 ≥ 0.95, satisfying the chance constraint condition, and this parameter combination is judged as a qualified solution. By setting a 95% confidence threshold and establishing a complete probability judgment system, it is ensured that the optimized solution performs stably under most operating conditions, avoiding failure under extreme perturbations while only meeting the standard under a single operating condition, thus significantly enhancing the robustness and reliability of the design. The Monte Carlo sampling size N ≥ 1000, and a convergence judgment criterion is established through variance analysis: when the probability estimate fluctuation of 5 iterations is less than 0.5% after increasing the sample size, the result is considered convergent. Random number generation adopts the Mersenne Twister algorithm with a fixed random seed to ensure reproducibility.

[0090] S2 specifically includes the following sub-steps:

[0091] S210. Establishing Multi-Physical Mechanisms: Based on the working principle of fuel cell stacks, establish coupling mechanisms encompassing water, heat, gas, and electricity, specifically including:

[0092] Fluid flow and pressure drop: A three-dimensional CFD model was used, with boundary conditions including inlet gas flow rate of 0.5-2.0 slpm, inlet gas temperature of 25-80℃, and outlet pressure at or slightly above atmospheric pressure.

[0093] Heat transfer and temperature distribution: Establish heat conduction and convection models, considering the heat transfer coefficient of the reactor cooling water jacket. ;

[0094] Liquid water generation and migration: A two-phase flow model is used to describe the generation, accumulation, and migration of permeate water from the membrane electrode reaction within the flow channel; boundary conditions include a permeate rate of 10. -6 -10 -5 kg / s·cm². Key physical properties are disclosed as follows: interfacial tension is 0.072 N / m (water-gas interface at 25℃), gas diffusion layer porosity ranges from 0.5 to 0.7, and liquid water diffusion coefficient is approximately 2.6 × 10⁻⁶. -9 m² / s;

[0095] Electrochemical reaction and voltage output: A semi-empirical electrochemical model (Tafel equation + ohmic resistance) is adopted, with an input current density of 0.2-1.0 A / cm², output single cell voltage fluctuation, and typical ohmic resistance of 0.05-0.1 Ω·cm². Through the coupling of the above four mechanisms, a complete physical chain of "flow channel parameters - uncertainty factors - performance indicators" is formed.

[0096] S220. Forming a stochastic coupling evaluation model: Based on the multi-physics mechanism established in S210, a stochastic coupling evaluation model is formed, specifically as follows:

[0097] Input: Set of flow channel parameters (following S140) and set of uncertainties (following S130).

[0098] Random sampling size: At least N=1000 uncertain samples are generated using the Monte Carlo method;

[0099] Output: Distribution results of the corresponding performance indicators, including mean, standard deviation, and 5th, 50th, and 95th percentile values.

[0100] This model provides both average performance predictions and reflects performance distribution characteristics, laying the foundation for subsequent opportunity-constrained optimization.

[0101] S230. Model Calibration: Calibrate the stochastic coupling evaluation model using experimental data to ensure that the model prediction results are consistent with the measured results. This includes:

[0102] Data sources: Bench tests: collect data from no less than 100 sets of different gas flow rates, humidity and temperature conditions; In-service operation: extract no less than 500 hours of monitoring data, covering typical operating conditions such as full load, partial load and low temperature and high humidity.

[0103] Calibration method: Least square fitting is used to correct key parameters such as gas diffusivity, heat transfer coefficient, and reaction rate constant.

[0104] Calibration standards: Voltage drop error ≤ ±10Pa; Voltage fluctuation error ≤ ±5mV; Temperature gradient error ≤ ±0.5K / cm; Liquid phase saturation error ≤ ±0.05; Cathode oxygen utilization error ≤ ±2%.

[0105] If all the above criteria are met, the model calibration is successful and can proceed to accuracy verification.

[0106] S240. Accuracy Verification: The accuracy of the stochastic coupling evaluation model calibrated via S230 is verified, specifically including:

[0107] Validation method: Five-fold cross-validation is used, dividing the entire dataset into 80% training and 20% validation, and repeating the process for five rounds.

[0108] Validation data size: The validation set should contain at least 20 independent operating conditions to ensure statistical robustness.

[0109] Validation metric: On the validation set, the errors between predicted and measured values ​​must simultaneously satisfy the following:

[0110] Pressure drop error ≤ ±10Pa; liquid phase saturation error ≤ ±0.05; temperature gradient error ≤ ±0.5K / cm; voltage fluctuation error ≤ ±5mV; cathode oxygen utilization error ≤ ±2%.

[0111] When all indicators are within the acceptable error range, the model accuracy verification is deemed successful, and it can be used as the basic model for chance-constrained optimization. In multi-physics coupled solutions, a commercial CFD solver (such as ANSYS Fluent) is used. Mesh independence analysis is completed through a level 3 mesh refinement test, with the error controlled within 2%. The time step setting satisfies the Courant condition, and the residual convergence threshold is set to 1×10⁻⁶. -5 This ensures the stability of the calculation results.

[0112] S3 specifically includes the following sub-steps:

[0113] S310. Establish opportunity constraints: Under the uncertainty set U, for candidate parameter combinations Conduct sample-based evaluation and establish opportunity constraints for performance indicators:

[0114] Input elements: Uncertainty set: from S130; Performance indicators and thresholds: from S150 (pressure drop ≤ 500 Pa, liquid saturation ≤ 0.70, temperature gradient ≤ 5.0 K / cm, voltage fluctuation ≤ 20 mV, oxygen utilization ≥ 85%); Confidence threshold: 95% (inherited from S160).

[0115] Sampling and calculation methods: Sampling size: No less than N=1000 samples are generated using the Monte Carlo method; Calculation method: The S220 stochastic coupling evaluation model is run under each sample to obtain the index sequence.

[0116] Probability determination formula:

[0117]

[0118] in This represents the empirical probability of achieving the pressure drop target; N is the total number of samples, usually N≥1000, derived from Monte Carlo sampling of an uncertain set; It is a vector of design parameters (such as channel width, rib width, depth, etc.). This represents the i-th uncertainty sample (manufacturing tolerance, environmental fluctuations, operating condition disturbances, etc.); This represents the pressure drop value calculated under given design parameters and uncertainty samples; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise.

[0119] Calculation method: Sampling: Randomly generate N samples from the uncertainty set; ; Calculation: For each sample, calculate the pressure drop based on the stochastic coupling evaluation model (S220). Judgment: If If the result is positive, it is recorded as 1; otherwise, it is recorded as 0. The average is calculated by summing the results of all samples and dividing by N to obtain the probability of meeting the standard. The above formula expresses the probability that the pressure drop is ≤500Pa under all random disturbances. The larger the value, the more robust the design is under various uncertain conditions; when... A value of ≥0.95 means that at least 95% of operating conditions can meet the pressure drop constraint.

[0120]

[0121] in Empirical probability representing the liquid phase saturation index; This represents the local liquid phase saturation under given parameters and uncertainties; the threshold of 0.70 indicates that the liquid water content does not exceed 70%; it represents the probability that the liquid phase saturation does not exceed 70% under various manufacturing and operational uncertainties; if A value of ≥0.95 indicates that "flooding" will not occur in most cases, and the reactor interface remains unobstructed.

[0122]

[0123] in This represents the empirical probability of achieving the temperature gradient target. This represents the temperature gradient (in K / cm) calculated under given parameters and uncertainties; the threshold of 5.0 K / cm characterizes the maximum permissible temperature non-uniformity within the fuel cell stack; the above formula represents the probability that the internal temperature gradient of the fuel cell stack is ≤5.0 K / cm under uncertainties such as manufacturing tolerances and environmental fluctuations; it reflects the thermal uniformity and heat dissipation capacity of the system, avoiding aging or performance degradation of the membrane electrode due to local overheating; if A value ≥0.95 indicates that the temperature distribution is uniform under most operating conditions, and the design is feasible.

[0124]

[0125] in This represents the empirical probability that voltage fluctuations meet the standard. This represents the voltage fluctuation amplitude (in mV) of a single cell under given design parameters and uncertainties; the threshold is 20mV: the voltage fluctuation is specified to be no more than 20mV; the above formula represents the probability that the voltage fluctuation of the fuel cell stack will not exceed 20mV under various uncertainties; small voltage fluctuation indicates stable fuel cell stack output, which is beneficial to the reliable operation of the power supply system and load equipment; if A value of ≥0.95 indicates that the voltage is stable under most operating conditions, meeting the robustness design objective.

[0126]

[0127] in This represents the empirical probability of achieving the oxygen utilization rate target. This represents the cathode oxygen utilization rate (%) calculated under design parameters and uncertainty conditions; threshold 85%: stipulates that the oxygen utilization rate must be greater than or equal to 85%; the above formula represents the probability that the oxygen utilization rate of the fuel cell stack is greater than or equal to 85% under various uncertainty conditions; high oxygen utilization rate means that oxygen resources are fully utilized, improving energy efficiency and reducing operating costs; A value of ≥0.95 indicates that the oxygen utilization efficiency of the fuel cell stack meets the standard under most operating conditions, and the design is robust.

[0128] The comprehensive constraints are as follows: Judgment rule: If the above conditions are met, the parameter combination is judged as "feasible"; otherwise, it is "infeasible".

[0129] S320. Define the robust optimization objective: Under the premise of satisfying the chance constraints in S310, further define the robust optimization objective function to balance average performance and manufacturing cost:

[0130] Input data: mean and standard deviation of performance metrics (from S220 sample statistics); manufacturing implementation cost model (based on channel width, rib width, channel depth, and process yield / processing cost as a function);

[0131] Weight set It can be determined through the Analytic Hierarchy Process (AHP), design review, or grid search.

[0132] Objective function construction: For indicators that are "the smaller the better" (pressure drop, temperature gradient, voltage fluctuation), the mean is used directly; for indicators that are "the larger the better" (oxygen utilization rate), the negative sign is taken to make them "the smaller the better"; all indicators are linearly normalized to the [0,1] interval to avoid inconsistencies in dimensions.

[0133] Weighted summation objective:

[0134] in This represents the objective function value, used for comprehensive evaluation of design parameters. The overall performance of the fuel cell stack; This represents the expected or average pressure drop (Pa), and the smaller the value, the better; This represents the expected or average value of the temperature gradient (K / cm), and the smaller the value, the better; This represents the expected or average value (mV) of voltage fluctuation; the smaller the value, the better. The value represents the expected or average value (%) of cathode oxygen utilization. The larger the value, the better. Therefore, a negative sign is used to make it "smaller is better" in the objective function. Cost represents the manufacturing or operating cost, which can be in monetary value or a dimensionless normalized index. The weights of each performance indicator reflect the importance of different indicators and are generally determined through expert experience or normalization methods, and must satisfy the following conditions: And can be normalized to .

[0135] The above formula uses a weighted summation method to unify multiple performance indicators into a single objective function; among them, pressure drop, temperature gradient, voltage fluctuation, and cost are indicators that are better when minimized, so they are directly weighted positively; oxygen utilization rate is an indicator that is better when maximized, so it is weighted negatively; minimizing... This means weighing various performance aspects and selecting the overall optimal design solution.

[0136] Constraints: Comparison is only made on parameter combinations deemed feasible by S310. If none of these options are feasible, then it is necessary to backtrack to the extended parameter range of S140 / S210 or improve the model.

[0137] Example result: With weights w=(0.30,0.15,0.25,0.20,0.10), a certain parameter set obtains J=0.395, which is the optimal value in the feasible set.

[0138] S330, Introduction of quantile performance improvement: To avoid relying solely on mean optimization and neglecting tail risk, a 95th quantile (or 5th quantile) indicator is further introduced for constraint improvement:

[0139] Input elements:

[0140] Percentile values ​​for each performance index: from statistical results of the S220 sample.

[0141] Quantile threshold setting: 95th quantile voltage drop ; 95th percentile voltage fluctuation 95th quantile temperature gradient 95th quantile saturation 5th quantile oxygen utilization .

[0142] Filtering and Judgment: In the feasible solution set of S310, solutions that do not meet the quantile threshold are further eliminated; define the quantile improvement metric:

[0143]

[0144] in This indicates a correction signal or error value, used to determine whether adjustment is triggered; This indicates a pre-set traffic threshold (i.e., the theoretically required minimum or target traffic). This represents the actual system flow detected by the sensor; the formula above represents the difference between the actual operating flow and the set threshold; when ,but This indicates that the actual gas supply is insufficient, requiring compensation or triggering an alarm; when ,but This indicates that the actual gas supply exceeds the threshold, and it may be necessary to reduce the supply to prevent waste or overload; when ,but This indicates that the system is in a critically stable state.

[0145] Parallel solution determination rule: If multiple solutions have similar J values ​​(difference ≤ 0.01), the one with the better quantile redundancy vector is selected first; if they are still the same, the one with lower manufacturing cost is selected. The weights of the weighted summation objective function can be adjusted according to different application scenarios (such as focusing on ripples for automotive applications and focusing on cost and lifespan for stationary power supplies), but this does not affect the consistency of the technical implementation process and judgment logic.

[0146] Example result: Solution A: The quantile threshold is satisfied; Solution B: Elimination occurs when the threshold is exceeded. Solution A is ultimately selected as the objective parameter solution.

[0147] S4 specifically includes the following sub-steps:

[0148] S410, Random Sampling and Solver Settings: Input: Randomized Coupled Evaluation Model after S220 calibration, Uncertainty Set (S130), Design Parameter Range (S140), Chance Constraints and Robust Objectives (S310-S330).

[0149] Sampling size: The Monte Carlo method is used, with the number of samples N = 1000-5000, which is adaptively increased to ensure convergence of results. A random seed (e.g., seed = 2025) is set to ensure reproducibility.

[0150] Initial design library generation: Continuous parameters (channel width, rib width, channel depth): Latin hypercube sampling (LHS) combined with grid method is used, with 5 levels for each parameter;

[0151] Discrete parameters (channel grouping method): Full enumeration is used;

[0152] An initial candidate solution set D0 with a size of approximately 150-300 groups is formed.

[0153] Solver selection and parallelization: Parallel grid search is used as the baseline; local optimization algorithms are superimposed when necessary. Parallelism is recommended to be ≥8 threads. The optimization process uses Latin hypercube sampling combined with particle swarm optimization (PSO). PSO parameters are set as follows: population size 20, number of iterations 30, inertia weight 0.8, and individual / global learning factors both 1.5. Convergence is accelerated through parallel computing strategies. Early termination condition: If 80% of candidate solutions have been calculated and the optimal objective value has improved by <1% in the last 50 iterations, then early termination is performed.

[0154] S420. Scene Approximation and Equivalent Determinization: Objective: To approximate the random space using finite representative scenes, thereby improving solution efficiency. Representative Scene Selection: Perform K-means clustering on N=1000 uncertain samples, select K=20 cluster centers as representative scenes, and record the weights. .

[0155] Weighted evaluation: The model is run on 20 representative scenarios to output weighted mean, variance and quantile statistics; for conservative constraints (such as pressure drop ≤500Pa), the worst-case scenario is used to replace the probability constraints to form equivalent deterministic conditions.

[0156] Consistency check: For the top 30 designs that pass the scenario approximation test, a full sample of N=1000 is reviewed. If the pass rate is ≥95%, the design is considered valid. If it is lower than the threshold, K is increased to 30 or stratified sampling is used.

[0157] S430, Candidate Parameter Set Generation: Hard Constraint Screening: Based on the N=1000 full samples or scenario approximation verification, calculate the empirical probability that each performance index meets the threshold for the initial design library D0 and its local optimization extended solution.

[0158] Judgment criteria: When all performance indicators are satisfied

[0159]

[0160] When the quantile improvement threshold (S330) is met, the solution is included in the candidate set C; size control: if |C|>100, the top 100 in the comprehensive score are retained; if |C|=0, the single non-critical threshold is relaxed by 5% and resampling is performed; output: candidate parameter set C, size 10-100, covering the entire design space.

[0161] S440, Sorting and Multi-Indicator Scoring: Input: Performance metric statistics, manufacturing cost estimates, and quantile redundancy of candidate set C. Normalization: Normalize each performance metric and cost to [0,1] to form comparable quantities.

[0162] Objective function: Comprehensive scoring function

[0163]

[0164] in This represents the comprehensive optimization evaluation function, used to ultimately determine the design parameters. The advantages and disadvantages; This represents the weighted summation objective function, which already includes performance indicators such as pressure drop, temperature gradient, voltage fluctuation, oxygen utilization rate, and cost. Represents robustness metrics, which measure the stability of a design under uncertainty, such as the variance of performance fluctuations or the probability of constraint satisfaction; It represents a default risk indicator, measuring the severity of design violations of constraints, such as the probability of exceeding limits under opportunity constraints; , It is a tradeoff coefficient that adjusts the importance of the robustness enhancement term and the risk penalty term respectively, and is generally taken as a non-negative real number.

[0165] The above formula is based on the objective function. Add a modified comprehensive evaluation function: This indicates that if the design is more robust (lower volatility, higher confidence level), an additional reward will be given; This means that if the design violates the constraints (such as the probability of meeting the standard is less than 95%), it will be penalized; in this way, the three aspects of "performance, robustness, and risk" are integrated into a unified optimization index.

[0166] Sorting rules: Sort by S value in ascending order, and retain the top K best solutions (usually K=10); if the difference in S value is ≤0.01, then compare quantile margin, manufacturing cost, and process margin in turn.

[0167] S450, Target Parameter Selection and Filing: Feasibility Review: For the top 10 ranked options, check the manufacturing process limits and gauge capabilities item by item. If a parameter approaches the limit with a margin of <10%, it is marked as "yellow light". Prioritize the option with a margin of ≥10%.

[0168] Target Solution Determination: If the S-value difference between the first and second-ranked solutions is ≥0.02, it is directly determined as the target parameter solution; if the difference between the first 2-3 solutions is <0.02, it is simultaneously determined as an equivalent solution and enters small-batch A / B verification trial production. Archiving and Traceability: Generate a version number (e.g., FC-Chan-V1.3) for the final target solution, and solidify the random seed, sample set hash, solver configuration, weight vector, and running log. Output: A list of target parameter solutions or equivalent solutions, and a data package containing "parameters-tolerances-operating conditions-quantile redundancy-inspection items-rollback strategies".

[0169] S5 specifically includes the following sub-steps:

[0170] S510 Independent Verification Experiment Design: Experimental Objective: To verify the effectiveness and stability of the target parameter solution determined in S450 under actual operating conditions.

[0171] Validation operating conditions: Load range: 0.2-1.0 A / cm², covering low, medium, and high loads; Operating environment: Gas temperature and humidity 20-90%RH, ambient temperature 0-40℃, air pressure 95-105 kPa; Dynamic operating conditions: Step load, periodic load, and random fluctuation load (power change rate ≥10% / s). Sample size: No less than 20 independent experimental operating conditions, each operating condition lasting ≥2 hours, with a cumulative validation time ≥100 hours. Measurement methods: Data is collected using a high-precision differential pressure gauge (accuracy ±1Pa), thermocouple array (accuracy ±0.1K), voltage sensor (accuracy ±1mV), and oxygen concentration sensor (accuracy ±0.1%).

[0172] S520, Performance Comparison and Threshold Check: Comparison Method: Compare the experimentally measured performance indicators with the thresholds set in S150 item by item.

[0173] Judgment criteria: Pressure drop ≤ 500 Pa; Liquid phase saturation ≤ 0.70; Temperature gradient ≤ 5.0 K / cm; Voltage fluctuation ≤ 20 mV; Cathode oxygen utilization rate ≥ 85%; Pass condition: In 20 independent experimental conditions, the compliance rate of each indicator is ≥ 95%, then it is judged as "verification passed"; Example result: For a certain target solution, under 20 conditions, the pressure drop compliance rate is 100%, the voltage fluctuation compliance rate is 95%, and the oxygen utilization rate is 100%, so it is judged as passed.

[0174] S530, Anomaly Diagnosis and Retrospective Adjustment: Anomaly Identification: If the compliance rate of a certain indicator is <95%, it is determined to be an "abnormal indicator"; Diagnosis Method: Plot the residual distribution curve for the abnormal indicator and analyze the error concentration area; Combine with the operation log to determine the source of the anomaly (such as extreme fluctuations in ambient air pressure, insufficient rebound of seals).

[0175] Retrospective adjustment strategy: If a single indicator slightly exceeds the limit (<5%), adjust the corresponding physical parameters (such as heat transfer coefficient and diffusivity) in the model and recalibrate in the S220-S240 process; if multiple indicators systematically exceed the limit, retrospectively go back to S140 to expand the parameter space or reset the quantile threshold (S330).

[0176] Iteration limit: The number of backtracking adjustments shall not exceed 3. If it still fails, the target solution shall be deemed invalid and S410-S450 shall be run again to generate a new solution. For candidate flow channel parameters that do not meet the performance indicators, a maximum of three rounds of backtracking iteration adjustment are allowed. The triggering condition is that the deviation between the actual measured value and the simulation prediction value exceeds the threshold. Backtracking adjustment is based on sensitivity analysis and prioritizes the adjustment of two highly sensitive variables, pressure drop and flow channel depth, to ensure efficient convergence.

[0177] S540, Final Confirmation and Archiving: Final Confirmation Conditions: After independent verification and retrospective adjustment through S510-S530, the performance index compliance rate is ≥95% across the entire operating range.

[0178] Archiving mechanism: Generate a version number for the final solution (e.g., FC-Final-V2.0), solidify model parameters, experimental data, and backtracking adjustment records; archived content includes: parameters-tolerances-experimental conditions-compliance rate-error distribution-number of iterations.

[0179] Traceability: Any future optimizations or reviews can be traced back to the specific sample set, experimental batch, and backtracking adjustment records.

[0180] Output results: The final parameter solution is determined and a "Verification and Backtracking Report" is generated as technical support for the patent implementation.

[0181] S6 specifically includes the following sub-steps:

[0182] S610 Long-term operation monitoring and data acquisition: Monitoring objective: To monitor the long-term operating status of the final parameter solution in the actual fuel cell stack and ensure stable performance throughout the entire life cycle.

[0183] Data collected variables include voltage drop. Liquid phase saturation Temperature gradient Voltage fluctuations, Ripple, Oxygen utilization rate And external environmental variables (temperature, humidity, air pressure).

[0184] Sampling frequency and period: Instantaneous operating condition: sampling frequency ≥ 1Hz; Average operating condition: mean and standard deviation of rolling window for 1 minute; Long-term trend: archiving period ≥ 1 hour, cumulative monitoring duration ≥ 5000 hours.

[0185] Data storage: Data is stored in the form of timestamp sequences, supporting backtracking analysis and anomaly tracing.

[0186] S620. Anomaly Detection and Diagnosis: Judgment Method: Dynamic threshold determination and trend analysis of monitoring indicators.

[0187] If Δp > 500 Pa, or Ripple > 20 mV, or This triggers a Level 1 alarm.

[0188] If Q95(Δp)>480Pa, or Q95(Ripple)>18mV, or If the threshold is less than 86%, a level 2 alarm will be triggered.

[0189] Diagnostic methods: Identify deviation trends by comparing residual distribution and moving average curves; compare operation logs during abnormal periods to distinguish between environmental disturbances (such as sudden drops in air pressure) and structural failures (such as aging of seals).

[0190] Anomaly Classification: Environmental disturbances: generally short-term fluctuations; Structural degradation: manifested as long-term deviations or continuous exceedances. Operational monitoring uses a 1Hz frequency to collect voltage drop and voltage fluctuations in real time. The control logic is simple PID control with threshold triggering. When the deviation exceeds the set value, the optimization algorithm module of S410–S450 is automatically re-called to achieve adaptive iterative optimization.

[0191] S630, Adaptive Maintenance and Parameter Correction: Adaptive Trigger Condition: Adaptive maintenance is activated when the duration of a Level 2 alarm is ≥10 min, or the compliance rate is <95%.

[0192] Correction methods: Slight deviation (threshold exceeds <5%): Adjust operating conditions, such as increasing intake flow rate and reducing load current; Moderate deviation (5-10%): Adjust physical parameters (diffusivity, heat transfer coefficient) in the model and rerun S220-S240 calibration; Severe deviation (>10%): Backtrack to S410-S450 and regenerate candidate solutions;

[0193] Execution method: A closed-loop control system is adopted to automatically issue adjustment commands (such as adjusting fan speed and coolant flow) and record operation logs.

[0194] S640, Continuous Optimization and Version Iteration: Rolling Optimization Cycle: A comprehensive evaluation is conducted every 1000 hours of cumulative operation or every quarter, and the statistical distribution and quantile thresholds are updated;

[0195] Version iteration mechanism: If the new parameter solution generated in the new round of optimization improves the objective function J by ≥5% compared with the current solution, it will be upgraded to a new version (e.g., V2.1→V2.2), and the model, samples and validation data will be solidified;

[0196] Archiving and Traceability: A unique version number is generated for each iteration, and the archive includes: runtime data, exception records, corrected parameters, and iteration results to ensure future traceability;

[0197] Final output: Achieve long-term stable optimization of fuel cell flow channel parameters, forming a closed-loop system of "monitoring-diagnosis-correction-optimization".

[0198] Example 2: This example provides a system for determining the flow channel parameters of a fuel cell stack, including:

[0199] The uncertainty modeling module is used to acquire manufacturing process statistics and historical operating environment data, combine manufacturing uncertainty and operational uncertainty to form an uncertainty set, and set an opportunity constraint threshold under the condition that the confidence level is not less than 95%.

[0200] The stochastic coupling evaluation model module is used to establish the mapping relationship between flow channel parameters, uncertainties and performance indicators based on multiple physical mechanisms of water, heat, gas and electricity. It takes parameter and uncertainty samples as inputs and outputs the performance indicator distribution results, and performs calibration and accuracy verification through experimental data.

[0201] The Opportunity Constraint and Robust Objective Definition Module is used to establish opportunity constraints for performance metrics and define the optimization objective function, which includes at least the expected optimal performance and the 95th percentile performance improvement, in order to improve average performance and tail robustness.

[0202] The optimization solution module is used to perform chance-constrained optimization on the stochastic coupled evaluation model using random sampling, scenario approximation or equivalent determinism methods, generate a set of candidate parameters, and sort the candidate solutions based on feasible probability, expected performance, quantile redundancy and manufacturing cost to determine the target parameter solution;

[0203] The verification and backtracking module is used to verify the target parameter solution under independent sampling sets and extreme working conditions. When the performance index touches the threshold, it triggers backtracking adjustment and generates a manufacturing tolerance window and assembly clamping force recommendation range that matches the target parameter solution.

[0204] The parameter solidification and release module is used to solidify the verified target parameters into a standard version, establish the correspondence between parameters, tolerances and operating conditions, and form a judgment procedure that includes incoming material and assembly inspection items, factory consistency judgment thresholds, in-service monitoring indicators and abnormal rollback strategies, which serves as the basis for mass production and operation.

[0205] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0206] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0207] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0208] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0209] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0210] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0211] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0212] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0213] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0214] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining the flow channel parameters of a fuel cell stack, characterized in that, Includes the following steps: S1. Collect manufacturing process data and operating environment data to establish an uncertainty set; determine the flow channel parameter set and performance index set, and set the opportunity constraint threshold that enables the performance index to meet the constraints under a predetermined confidence level. S2. Based on the multi-physical mechanisms of water, heat, gas and electricity, establish the mapping relationship between flow channel parameters, uncertainties and performance indicators, form a stochastic coupled evaluation model with parameter and uncertainty samples as input and performance indicator distribution as output, and use experimental data for calibration and accuracy verification. S3. Establish opportunity constraints for performance indicators and define optimization objective functions, including at least the expected optimal performance and quantile performance improvement, so as to balance average performance and robustness under extreme conditions. S4. Optimize the stochastic coupling evaluation model using random sampling, scenario approximation, or determinism methods to obtain a set of candidate parameters that satisfy the chance constraints. Then, based on the feasibility probability, expected performance, quantile redundancy, and manufacturing cost, perform a comprehensive ranking to determine the target parameter solution. S5. Verify the performance distribution of the target parameter solution under independent sample sets and extreme working conditions, calculate the confidence interval and safety margin; when the performance touches the constraint threshold, trigger backtracking adjustment and generate the corresponding manufacturing tolerance window and assembly clamping force recommendation range. S1 specifically involves: acquiring statistical data on the manufacturing process, including manufacturing uncertainties such as dimensional variations in the flow field plate, non-uniformity of gas diffusion layer compaction, and differences in seal springback; acquiring historical data on the operating environment, including operational uncertainties such as load fluctuations, inlet and outlet gas temperature and humidity, and changes in ambient air pressure; combining manufacturing and operational uncertainties to establish an uncertainty set, which serves as the basis for subsequent evaluation and optimization; and determining the set of flow channel parameters to be optimized, wherein the flow channel parameters include at least channel width, rib width, channel depth, channel grouping parameters, and target pressure drop quantiles.

2. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, It also includes S6, which decrypts and solidifies the verified target parameters into a standard version, establishes the correspondence between parameters, tolerances, and operating conditions, and forms a judgment procedure that includes inspection thresholds, factory consistency judgment, in-service monitoring, and anomaly rollback strategies for mass production and field applications.

3. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, S1 further includes: setting a set of performance indicators, the performance indicators including at least pressure drop, local liquid phase saturation, temperature gradient, single cell voltage fluctuation and cathode oxygen utilization; setting a target confidence threshold for opportunity constraints, the confidence threshold being not lower than a set value.

4. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, S2 specifically refers to: Based on multiple physical mechanisms involving water, heat, gas, and electricity, a mapping relationship between flow channel parameters, uncertainty samples, and performance indicators is established. A stochastic coupling evaluation model is formed, which takes flow channel parameters and uncertainty samples as inputs and outputs corresponding predicted values ​​of performance indicators. The random coupling evaluation model is calibrated using in-service test data or bench test data, and the model accuracy is verified within the set operating conditions to ensure that the prediction results meet the preset accuracy requirements.

5. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, S3 specifically refers to: Establish opportunity constraints such that, under a set of uncertainties, the probability that the performance index meets the set upper or lower limit is not less than the confidence threshold. Under the premise of satisfying the opportunity constraints, define an optimization objective function, wherein the optimization objective includes at least minimizing the expected loss; The worst quantile performance improvement is introduced into the optimization objective, where the worst quantile is the 95th quantile performance, to improve overall performance and consistency.

6. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, S4 specifically refers to: A random sampling method is used to solve the chance constraint problem of the stochastic coupled evaluation model. The stochastic coupling evaluation model is supplemented by a scenario approximation method or an equivalent deterministic method to obtain a set of candidate parameters that satisfy the chance constraints. The candidate parameter set is sorted based on factors including the feasibility probability at the target confidence level, the expected loss, and the manufacturing and implementation cost. Choose the parameter combination that ranks highly and has excellent overall performance as the objective parameter solution.

7. The method for determining the flow channel parameters of a fuel cell stack according to claim 1, characterized in that, S5 specifically refers to: The solutions to the target parameters are validated on a sampling set independent of the solution process to obtain confidence intervals for pressure drop, liquid saturation, temperature gradient and voltage fluctuation. Boundary verification of the objective parameter solution is performed under extreme combined operating conditions to evaluate whether each performance index satisfies the chance constraint conditions. When any performance index touches the opportunity constraint threshold under boundary conditions, a backtracking adjustment process is triggered to update the target parameter solution. While backtracking and adjusting, a manufacturing tolerance suggestion window and an assembly clamping force suggestion range are generated to match the target parameter solution.

8. The method for determining the flow channel parameters of a fuel cell stack according to claim 2, characterized in that, S6 specifically refers to: The target parameters for verification are solidified into flow channel parameter versions; the correspondence between parameters, tolerances, and operating conditions is established to form a judgment procedure; The judgment procedure should clearly define the incoming material inspection items, assembly inspection items, and the threshold for judging factory consistency. The judgment procedure further specifies in-service monitoring indicators and anomaly rollback strategies to ensure that the target parameter solution can be implemented in mass production and field applications.

Images