A preparation method and system of a hydrogen fuel cell based on multi-link feedback control

Abstract

The present invention relates to a preparation method and system for a hydrogen fuel cell based on multi-link feedback control, the method comprising: uniformly coating a catalyst slurry on the surface of a proton exchange membrane through an electromagnetic field-assisted coating process to form cathode and anode catalyst layers; real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly during the hot pressing process, and dynamic adjustment of the short-term coating parameters of the electromagnetic field-assisted coating process based on the resistance data; online collection of electrochemical impedance spectroscopy data during the electrochemical activation stage, and reverse optimization of the catalyst slurry in the electromagnetic field-assisted coating process and the long-term process benchmark values ​​in the hot pressing process through a transfer learning algorithm; the system is used to implement the above method, and the system comprises: an electromagnetic field-assisted coating module, a hot pressing-resistance monitoring module, and an electrochemical activation-transfer learning optimization module. The present invention effectively solves the problems of uneven catalyst coating and solidified process parameters in the membrane electrode preparation process through a multi-link feedback control mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen fuel cell preparation, and in particular to a method and system for preparing a hydrogen fuel cell based on multi-link feedback control. Background Art

[0002] As a highly efficient and clean energy conversion device, hydrogen fuel cells hold broad application prospects in new energy vehicles, distributed power generation, and other fields. The membrane electrode (MEA), a core component of hydrogen fuel cells, has a manufacturing process that directly impacts its performance and lifespan. Currently, traditional MEA manufacturing methods primarily involve catalyst coating, hot pressing, and electrochemical activation. However, existing technologies present a significant technical challenge during the manufacturing process: uneven catalyst slurry coating and process parameter curing lead to uneven distribution of contact resistance at the membrane electrode interface, which in turn impacts battery performance and durability.

[0003] This problem mainly arises from two reasons:

[0004] First, during the catalyst coating stage, traditional coating processes (such as spraying and blade coating) make it difficult to precisely control the distribution of the catalyst slurry, which can easily lead to local agglomeration or uneven thickness, affecting the interface contact quality during subsequent hot pressing.

[0005] Secondly, the lack of a dynamic feedback mechanism between the various process links leads to the solidification of process parameters, which makes it difficult to adapt to fluctuations in material properties and equipment operation deviations, and cannot be dynamically adjusted according to real-time feedback; for example: the impedance spectrum data collected during the electrochemical activation stage is usually only used to evaluate battery performance, and fails to form a closed-loop optimization with the coating and hot pressing parameters. The defects of isolated process links and fragmented data make the optimization cycle of long-term process benchmark values ​​such as catalyst slurry ratio and hot pressing pressure long and the trial-and-error cost high, which ultimately leads to poor consistency of battery batches and performance improvement bottlenecks. Summary of the Invention

[0006] To this end, the present invention provides a method and system for preparing a hydrogen fuel cell based on multi-link feedback control. Through the multi-link feedback control mechanism, the problems of uneven catalyst coating and process parameter solidification during the membrane electrode preparation process are effectively solved.

[0007] To solve the above technical problems, the present invention provides a method for preparing a hydrogen fuel cell based on multi-link feedback control, comprising:

[0008] The catalyst slurry is evenly coated on the surface of the proton exchange membrane through an electromagnetic field assisted coating process to form cathode and anode catalyst layers;

[0009] Real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly during the hot pressing process, and dynamic adjustment of the short-term coating parameters of the electromagnetic field-assisted coating process based on the resistance data;

[0010] Electrochemical impedance spectroscopy data is collected online during the electrochemical activation stage, and the long-term process benchmark values ​​of the catalyst slurry and hot pressing process in the electromagnetic field-assisted coating process are reversely optimized through the transfer learning algorithm.

[0011] In one embodiment of the present invention, the electromagnetic field assisted coating process includes:

[0012] Before coating, the catalyst slurry is pre-sheared to form a shear densified structure;

[0013] During the coating process, a pulsed alternating electromagnetic field is applied to orient the platinum nanoparticles in the catalyst slurry along the direction of the magnetic field, and a dynamic rotating magnetic field is applied simultaneously to control the periodic rotation of the magnetic field direction, with the rotation angle alternately switching between 0 and 180 degrees.

[0014] In one embodiment of the present invention, during the coating process, an ultrasonic cavitation generator is used to destroy the ionomer-encapsulated bubbles in the slurry at a certain frequency.

[0015] In one embodiment of the present invention, during the coating process, a temperature gradient field is set in the coating area along the coating direction, gradually decreasing from 40°C at the starting end of the coating to 25°C at the ending end, with a gradient rate of 2-5°C / cm.

[0016] In one embodiment of the present invention, real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly includes:

[0017] A distributed resistance sensor array is embedded on the upper and lower template surfaces of the hot press, with each sensor arranged at equal intervals. The distributed resistance sensor array generates a resistance thermal map of each area of ​​the membrane electrode assembly in real time;

[0018] If the standard deviation of the resistance distribution exceeds 0.5 mΩ·cm², the electromagnetic field-assisted coating process is triggered and the electromagnetic field frequency, intensity or output mode is adjusted.

[0019] In one embodiment of the present invention, if the high resistance area is concentrated at the edge of the membrane, the electromagnetic field frequency is increased by 10% to 20%;

[0020] If the high resistance areas are scattered, increase the magnetic field strength by 0.1-0.2 T and increase the slurry viscosity by 5%-8%;

[0021] If the resistance low value areas and high value areas are distributed alternately, the electromagnetic field is switched to an intermittent pulse mode.

[0022] In one embodiment of the present invention, the implementation of the transfer learning algorithm includes the following steps:

[0023] Construct a convolutional neural network model, setting the input data to the charge transfer resistance and double layer capacitance from the electrochemical impedance spectroscopy and the real-time temperature during the activation stage;

[0024] The training dataset includes coating parameters from historical production, including platinum loading and ionomer content, hot pressing pressure and temperature, and the corresponding charge transfer resistance and double layer capacitance.

[0025] The outputs of the neural network model are the adjusted values ​​of platinum loading and ionomer content in the catalyst slurry, as well as the compensated values ​​of hot pressing temperature and pressure during the hot pressing process.

[0026] In one embodiment of the present invention, when the charge transfer resistance is greater than 0.25 Ω·cm² and the double layer capacitance is less than 20 mF / cm², the platinum loading is controlled to increase by 0.03-0.05 mg / cm², and the hot pressing temperature during hot pressing is adjusted to increase by 5°C.

[0027] When the charge transfer resistance is less than 0.15 Ω·cm² and the double layer capacitance is greater than 30 mF / cm², the ionomer content is controlled to be reduced by 5% to 8%, and the hot pressing pressure during hot pressing is adjusted to be reduced by 0.5 MPa.

[0028] When the charge transfer resistance and the double layer capacitance fluctuate nonlinearly, the electromagnetic field frequency in the electromagnetic field-assisted coating process is controlled to randomly perturb to break up the slurry agglomeration.

[0029] In one embodiment of the present invention, the reverse optimization of process parameters further includes the following collaborative strategies:

[0030] If the transfer learning model detects that the charge transfer resistance value drops rapidly at the initial stage of activation, an ultrasonic cavitation generator is used at a certain frequency to destroy the ionomer-encapsulated bubbles in the slurry, and a gradient pressure is applied during the hot pressing process;

[0031] If the transfer learning model detects that the double-layer capacitance value fluctuates periodically with activation time, the ratio of water and ethanol in the catalyst slurry is adjusted, and the hot pressing holding time of the hot pressing process is extended.

[0032] To solve the above technical problems, the present invention further provides a hydrogen fuel cell preparation system based on multi-link feedback control, comprising:

[0033] The electromagnetic field assisted coating module is used to evenly coat the catalyst slurry on the surface of the proton exchange membrane to form the cathode and anode catalyst layers, including:

[0034] Pulsed alternating electromagnetic field generators are placed on both sides of the coating machine roller;

[0035] Coating parameter dynamic adjustment unit, which receives external feedback instructions in real time and adjusts electromagnetic field parameters;

[0036] The hot pressing molding-resistance monitoring module is used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including:

[0037] Distributed resistance sensor arrays are embedded in the upper and lower template surfaces of the hot press;

[0038] The resistance data analysis unit sets trigger conditions based on the distributed resistance sensor array and provides coating parameter adjustment instructions;

[0039] A short-term parameter control interface transmits adjustment instructions to the coating parameter dynamic adjustment unit of the electromagnetic field assisted coating module;

[0040] The Electrochemical Activation-Transfer Learning Optimization module is used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmarks, including:

[0041] High-frequency impedance spectrum acquisition unit, configured in the activation equipment;

[0042] Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm;

[0043] Long-term parameter update interface, which writes the correction value into the slurry formula database of the electromagnetic field assisted coating module and the process parameter library of the molding module.

[0044] The above technical solution of the present invention has the following advantages over the prior art:

[0045] The method for preparing a hydrogen fuel cell based on multi-link feedback control described in the present invention systematically solves the core problem of the mismatch between process parameter curing and dynamic requirements by constructing a closed-loop feedback control system for the three links of "coating-hot pressing-activation".

[0046] First, the use of electromagnetic field-assisted technology in the coating stage can reduce slurry agglomeration. Combined with real-time resistance monitoring data during hot pressing, short-term coating parameters can be dynamically adjusted to further avoid the problem of local excessive thickness or thinness, thereby improving the structural consistency of the catalytic layer.

[0047] Secondly, during the electrochemical activation stage, the impedance spectrum data is reversely mapped to the previous process parameter space through a transfer learning algorithm, breaking through the limitations of the traditional trial-and-error method that relies on a fixed data set. The algorithm can identify the impact of differences in material properties from different batches on the process benchmark value.

[0048] Through the above-mentioned two cross-link parameter collaborative optimization mechanisms, the dual control of short-term process compensation and long-term benchmark iteration is achieved in principle, which not only ensures the manufacturing consistency of single products, but also continuously improves the process robustness through data closed loop, and finally forms an adaptive hydrogen fuel cell preparation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0050] Figure 1 It is a flowchart of the steps of the method for preparing a hydrogen fuel cell based on multi-link feedback control of the present invention;

[0051] Figure 2 It is a structural framework diagram of the hydrogen fuel cell preparation system based on multi-link feedback control of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention. Example 1

[0053] Reference Figure 1 As shown, the present invention discloses a hydrogen fuel cell preparation method based on multi-link feedback control. Its core lies in constructing a three-stage closed-loop process control system of "coating-hot pressing-activation". Through real-time data feedback and algorithm optimization, process guidance and dynamic adjustment of process parameters in different process stages are achieved. The specific implementation process and technical mechanism are as follows:

[0054] S10, electromagnetic field assisted coating process: The catalyst slurry is evenly coated on the surface of the proton exchange membrane through electromagnetic field assisted technology to form cathode and anode catalyst layers.

[0055] This stage uses the directional force of the electromagnetic field on the charged particles (such as platinum nanoparticles) in the catalyst slurry to reduce the capillary contraction differences during the slurry drying process, avoid cracks or holes in the catalytic layer, thereby improving the electronic conduction capacity of the catalytic layer and overcoming the agglomeration or sedimentation problems caused by differences in the rheological properties of the slurry in the traditional coating process; among them: the Lorentz force generated by the electromagnetic field can prompt the platinum particles to align along the preset direction, while suppressing the non-uniform wrapping of the ionomer and the catalyst, thereby forming an ordered three-phase reaction interface on the microscale; the directionally arranged catalyst particles can reduce the tortuosity of the proton transmission path, reduce the electronic conduction impedance inside the catalyst layer, and provide a structural catalyst layer substrate for subsequent hot pressing.

[0056] S20. Dynamic regulation in hot pressing: During the hot pressing stage, the interfacial contact resistance distribution of the membrane electrode assembly (MEA) is monitored in real time, and the short-term parameters of the electromagnetic field-assisted coating (such as electromagnetic field strength and frequency) are dynamically adjusted based on the resistance data.

[0057] When the hot press applies pressure, the quality of the interface contact between the proton exchange membrane and the catalyst layer directly affects the transmission efficiency of the reactant gas and protons. In this embodiment, a distributed resistance sensor array can be used to locate areas of poor contact (such as areas of high resistance). If an abnormally high local resistance is detected, it indicates that the catalytic layer and the proton exchange membrane in this area are not in close contact. In this case, the electromagnetic field force in this area can be targeted to enhance, causing the catalytic layer to undergo plastic deformation during the hot pressing process, filling the microscopic gaps at the interface and reducing the spatial discreteness of the contact resistance. This instant feedback mechanism overcomes the limitations of the fixed parameters of traditional hot pressing processes, links the hot pressing process with the coating process, and fundamentally solves the problem of uneven interface contact caused by coating problems.

[0058] S30. Reverse optimization of electrochemical activation: During the electrochemical activation stage, electrochemical impedance spectroscopy data is collected online, and the catalyst slurry formulation and long-term process benchmark values ​​for hot pressing (such as platinum loading, ionomer content, and hot pressing temperature) are reversely optimized through a transfer learning algorithm.

[0059] The charge transfer resistance and double-layer capacitance in the electrochemical impedance spectroscopy directly reflect the utilization rate of the active sites in the catalytic layer and the effective area of ​​the three-phase interface. The transfer learning algorithm establishes a multi-parameter coupled optimization model by correlating historical process parameters with electrochemical impedance spectroscopy data features. Utilizing a cross-stage correlation model between electrochemical impedance spectroscopy data and process parameters, it overcomes the data silo problem in the coating, hot pressing, and activation stages of the traditional process. Through charge transfer characteristics, it traces back to material ratio defects or other process defects, enabling precise correction of process parameters. For example, when an abnormal increase in charge transfer resistance is detected, the algorithm can reversely deduce that the platinum loading in the slurry is insufficient or the ionomer coating is too thick, thereby adjusting the long-term baseline parameters. This data-driven optimization approach breaks through the empirical reliance of traditional trial-and-error methods and achieves continuous iterative upgrades of process parameters at the system level.

[0060] Specifically, through steps S10 to S30, a dual control mechanism of "short-term compensation" and "long-term optimization" is formed: short-term adjustments during the hot pressing phase directly affect the current production unit, compensating for coating defects through immediate correction of electromagnetic field parameters; reverse optimization during the activation phase, based on cross-batch data accumulation, updates long-term process benchmarks to improve overall process robustness. The synergy of these two forms a closed-loop system of "real-time correction and historical learning," which not only ensures the quality consistency of individual products but also continuously improves the process upper limit through data accumulation.

[0061] Specifically, the invention of this application also improves the electromagnetic field assisted coating process, including:

[0062] Before coating, the catalyst slurry is pre-sheared to destroy the soft agglomerates of platinum nanoparticles in the slurry through mechanical shear force, and to promote the extension of the ionomer molecular chains along the shear direction to form a shear-densified structure. This step changes the rheological properties of the slurry, transforming it from an initial pseudoplastic fluid to a shear-thinning state. The shear densification effect generated by pre-shearing can reduce the air bubble encapsulation inside the slurry, while making the mixing of platinum particles and ionomer more uniform, reducing the stress concentration points caused by solvent volatilization during the drying process, and thus reducing the risk of cracking of the catalytic layer. In addition, this densified structure provides a stable medium environment for subsequent electromagnetic field action, avoiding fluctuations in coating thickness due to local viscosity differences in the slurry.

[0063] During the coating process, a pulsed alternating electromagnetic field is applied, leveraging the conductive properties of platinum nanoparticles to generate eddy currents in the alternating magnetic field. The Lorentz force generated by the interaction between the eddy currents and the magnetic field drives the platinum particles to align along the magnetic field. The aligned platinum particles form a continuous electron conduction path, effectively reducing the bulk resistance of the catalytic layer. The pulsed design (e.g., an intermittent magnetic field with a frequency of 1-10 kHz) prevents excessive particle migration and accumulation due to continuous force, while utilizing the alternating properties to maintain the dynamic balance of slurry fluidity and ensure linear controllable coating thickness. Compared to traditional static magnetic fields, alternating magnetic fields, by periodically changing direction, can cover a wider distribution area of ​​particles and reduce the loss of catalytically active surfaces caused by single orientation.

[0064] A dynamic rotating magnetic field is applied synchronously, causing the magnetic field direction to rotate periodically within the range of 0° to 180° (for example, switching direction every 5 seconds). This design breaks the excessive orientation of platinum particles in a single direction by perturbing the magnetic field in the spatial dimension. When the magnetic field direction rotates, the migration path of the platinum particles is deflected, forming a multi-dimensional interwoven network structure, which creates redundant conduction paths in the catalytic layer. When local areas suffer structural damage due to heat stress or activation, the multi-dimensional network can provide bypass conduction channels, improving battery operation stability. This structure microscopically increases the tortuosity factor of the three-phase reaction interface, expands the surface area of ​​the gas diffusion channel, and avoids the anisotropy of the mechanical strength of the catalytic layer caused by unidirectional arrangement. In addition, the alternating rotation angle also alleviates the edge effect of the slurry in the coating direction, improving the thickness uniformity of the coating layer across the width.

[0065] Based on the above basic electromagnetic field assisted coating process, the inventors further introduced ultrasonic cavitation effect and temperature gradient field control to further optimize and improve the electromagnetic field assisted coating process.

[0066] Specifically, the introduction of ultrasonic cavitation can further optimize the microscopic uniformity of the catalyst slurry and eliminate internal defects through the synergistic effect of acoustic and electromagnetic fields. During the coating process, an ultrasonic cavitation generator acts on the slurry layer at a specific frequency (e.g., 20-40 kHz). As the ultrasonic wave propagates through the slurry, it forms periodic compression and rarefaction areas, generating cavitation bubbles. When the bubbles collapse, the local instantaneous pressure can reach hundreds of MPa. This impact force can break up the overly thick coating of the ionomer on the catalyst particles and shatter any micron-sized bubbles remaining in the slurry.

[0067] In the electromagnetic field-assisted coating process, excessive wrapping of ionomers will hinder the exposure of active sites on platinum particles, while the presence of bubbles will lead to the formation of pore defects after the catalytic layer dries. The cavitation effect can directionally destroy the wrapping layer and bubbles, thereby improving the uniformity of ionomer distribution on the surface of platinum particles and reducing the structural voids after the catalytic layer is cured.

[0068] The cavitation effect of ultrasound and the pulsed alternating electromagnetic field form a temporal and spatial complementarity: in the spatial dimension, the cavitation effect of ultrasound acts across the entire thickness of the slurry, while the electromagnetic field primarily affects the alignment of platinum particles at the surface and near the surface. In the temporal dimension, the continuous action of ultrasound and the pulsed nature of the electromagnetic field form a dynamic superposition, continuously optimizing the microstructure during the slurry leveling phase. This synergistic mechanism avoids the problem of localized overtreatment caused by a single physical field. For example, while the electromagnetic field drives the migration of platinum particles, ultrasound simultaneously removes ionomer barriers along the migration path, thereby improving the efficiency of directional alignment.

[0069] Specifically, temperature gradient field control is introduced, and the leveling and curing process of the slurry is optimized through thermodynamic regulation. The temperature gradient is used to match the evolution law of the rheological properties of the slurry, and a linear temperature gradient is set from 40°C at the starting end to 25°C at the ending end in the coating direction, with a gradient rate of 2-5°C / cm; the higher temperature at the starting end reduces the viscosity of the slurry and enhances the fluidity, which is conducive to the rapid spreading of the slurry on the substrate surface; as the coating progresses, the temperature gradually decreases, causing the viscosity of the slurry to increase, suppressing the edge flow effect caused by gravity, and achieving consistent control of the coating thickness through temperature-viscosity coupling regulation; for example, in the high-temperature zone at the starting end, the shear-thinning properties of the slurry cooperate with the directional effect of the electromagnetic field to accelerate the arrangement of platinum particles; in the low-temperature zone at the ending end, the viscosity rebounds to fix the microstructure, preventing the arranged particles from secondary displacement due to the flow of the slurry.

[0070] The temperature gradient and the rotating magnetic field of the pulsed alternating electromagnetic field form a linkage effect: the high-temperature zone at the starting end: cooperates with the rapid switching of the rotating magnetic field (such as shortening the 0°→180° cycle), and uses low viscosity conditions to achieve multi-directional rapid migration of platinum particles; the low-temperature zone at the ending end: slows down the switching frequency of the magnetic field, so that the particles gradually stabilize their orientation in the high-viscosity medium; this temperature-magnetic field linkage control avoids the problem of orientation attenuation caused by the mismatch between the slurry curing rate and the electromagnetic response speed under a single temperature field.

[0071] Specifically, the present invention further discloses a process for real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly, comprising:

[0072] Micro-resistance sensors (for example, an array of interdigitated electrodes with a spacing of 5 mm) are embedded at equal intervals on the upper and lower template surfaces of the hot press to form a detection network covering the entire area of ​​the membrane electrode assembly. Each sensor measures the local contact resistance using a four-wire method to eliminate interference from wire resistance. When the hot press applies pressure, the microscopic contact state of the proton exchange membrane and the catalytic layer is mapped in real time to the spatial distribution of resistance values. The sensor array generates a resistance thermal map at a sampling frequency of 10-100 Hz, and intuitively displays the positional distribution of high-resistance areas (such as red patches) and low-resistance areas (such as blue areas) through color gradients.

[0073] The standard deviation threshold of the contact resistance distribution is set at 0.5 mΩ·cm². When the calculated actual standard deviation exceeds this threshold, it indicates that there is a significant uneven contact problem at the membrane electrode interface. At this time, the resistance data analysis unit sends an adjustment instruction to the electromagnetic field-assisted coating module.

[0074] Specifically, based on actual usage, the mapping rules between several common resistance distribution characteristics and electromagnetic field parameter adjustment are refined:

[0075] (1) When the high resistance area is concentrated at the edge of the membrane electrode (such as the annular red band in the thermal diagram), it is determined that the edge climbing effect in the coating stage causes the catalyst layer thickness to be thinned. At this time, the electromagnetic field frequency is increased by 10%-20% (for example, from 1kHz to 1.1-1.2kHz). The skin effect of the high-frequency electromagnetic field is used to enhance the lateral migration ability of the platinum particles on the surface of the slurry to the edge area. The high-frequency alternating magnetic field generates a higher eddy current density on the surface of the conductor, which prompts the platinum particles in the edge area to redistribute to fill the thickness gap, while suppressing the waste of slurry caused by excessive thickness in the center area.

[0076] (2) If the high-resistance areas are randomly scattered (e.g., discrete red spots appear in the thermal map), this indicates that there are localized agglomerations in the slurry or that the ionomer is too thick. In this case, the magnetic field strength is increased by 0.1-0.2T (e.g., from 0.5T to 0.6-0.7T) and the slurry viscosity is increased by 5%-8%. Increasing the magnetic field can break the van der Waals force binding of the agglomerates, prompting the platinum particles to peel from the agglomerates; while moderately increasing the viscosity (e.g., by adding a thickener in real time) can inhibit the secondary agglomeration of the particles during migration, ensuring the dispersion compensation effect.

[0077] (3) When the low resistance value area and the high resistance value area are distributed alternately (such as the red and blue striped areas appearing in the thermal diagram), it indicates that there are periodic slurry supply fluctuations or uneven surface tension of the substrate during the coating process. At this time, the electromagnetic field is switched to an intermittent pulse mode (such as 200ms pulse width alternating with 100ms interval), and the strong magnetic field of the pulse peak (such as 1.2T) is used to break through the interface adsorption barrier, while the intermittent period allows the slurry stress to relax, avoiding excessive migration and accumulation of particles caused by continuous high-intensity magnetic fields.

[0078] Specifically, the present invention further discloses an implementation process of the transfer learning algorithm, which includes the following steps:

[0079] A convolutional neural network is used as the core model of transfer learning. Its input layer is set as the key characteristic parameters of the electrochemical impedance spectroscopy: charge transfer resistance, double-layer capacitance and real-time temperature in the activation stage. The charge transfer resistance is selected to reflect the kinetic resistance of the catalyst surface reaction, and the double-layer capacitance represents the effective area of ​​the three-phase interface. The temperature parameter is used to correct the interference of the activation conditions on the electrochemical impedance spectroscopy data. The network hidden layer is designed as a spatiotemporal convolution module, which can simultaneously capture the frequency domain correlation of the charge transfer resistance-double-layer capacitance and the nonlinear coupling relationship with the process parameters.

[0080] The training dataset integrates multi-dimensional data from historical production, including:

[0081] Coating parameters: platinum loading (affects charge transfer resistance), ionomer content (affects double layer capacitance);

[0082] Hot pressing parameters: pressure (determines the tightness of the interface contact), temperature (affects the redistribution of the ionomer);

[0083] Performance data: Measured values ​​of charge transfer resistance and double layer capacitance for the corresponding batch.

[0084] Heterogeneous data (such as pressure in MPa and temperature in °C) are mapped to a unified feature space through normalization, enabling the model to identify implicit correlations between parameters.

[0085] The model output layer is set to the adjusted values ​​of the process parameters, including:

[0086] Platinum loading adjustment: directly related to the charge transfer resistance value, because platinum particles are the active site carriers of the charge transfer reaction;

[0087] Ionomer content adjustment: By adjusting the ionomer thickness, the proton transport path is affected, and the double layer capacitance is indirectly modified;

[0088] Hot pressing temperature / pressure compensation: Corrects the balance between interface contact resistance and catalyst layer porosity.

[0089] Specifically, based on actual usage, the corresponding rules for common electrochemical indicator combinations and process parameter adjustments are refined, and targeted optimization of process parameters is achieved through classification threshold setting:

[0090] (1) Combined regulation of high charge transfer resistance and low double layer capacitance. When the charge transfer resistance is greater than 0.25Ω·cm² and the double layer capacitance is less than 20 mF / cm², it is determined that insufficient platinum loading leads to a scarcity of active sites, and the ionomer is too thick to hinder the transport of protons and reactants. At this time, the following measures are taken: the platinum loading is increased by 0.03-0.05 mg / cm² to directly supplement the density of active sites and reduce the charge transfer activation energy; and the hot pressing temperature is increased by 5°C: this promotes the melting and redistribution of the ionomer during the hot pressing process, reducing its wrapping thickness on the platinum particles. This temperature increase can ensure that the glass transition temperature of the ionomer is exceeded according to model calculations, thereby optimizing the interface structure without causing membrane dehydration.

[0091] (2) Combined regulation of low charge transfer resistance and high double layer capacitance. When the charge transfer resistance is less than 0.15Ω·cm² and the double layer capacitance is greater than 30 mF / cm², it indicates that the ionomer content is too high, resulting in overload of the proton conduction channel and insufficient porosity of the catalyst layer. In this case, the following measures should be taken: reducing the ionomer content by 5%-8% to reduce the volume proportion of the ionomer and expand the gas diffusion channel; and reducing the hot pressing pressure by 0.5 MPa to alleviate the pore collapse of the catalyst layer caused by excessive pressure and restore part of the pore structure to balance the double layer capacitance and gas mass transfer efficiency.

[0092] (3) Strategies for dealing with abnormal nonlinear fluctuations. When the charge transfer resistance and double layer capacitance show nonlinear fluctuations (e.g., the charge transfer resistance first decreases and then increases with increasing frequency), it indicates that the slurry has dynamic agglomeration. At this time, the electromagnetic field frequency is controlled to be randomly disturbed (e.g., randomly fluctuate with a deviation of ±10%), and the unpredictability of the magnetic field direction is used to break the electrostatic adsorption balance of the platinum particles. The shear force field generated by the random disturbance can destroy the soft agglomerates formed in the slurry, while avoiding the aggravation of resonant agglomeration caused by regular frequency switching.

[0093] Furthermore, based on the above, the inventors also developed differentiated control strategies based on two typical abnormal modes commonly seen in production:

[0094] (1) When the transfer learning model detects that the rate of decrease of the charge transfer resistance value exceeds the threshold in the initial activation period (0-30 minutes), it indicates that there are micron-sized bubbles wrapped by ionomer in the slurry, and the following coordinated adjustments are triggered:

[0095] Ultrasonic cavitation intervention: Start an ultrasonic generator with adjustable frequency (20-35 kHz) and use the micro-jet generated by the cavitation effect to impact the bubble interface. The specific frequency selection (such as 28 kHz) can make the bubble resonate and collapse. The local pressure at the moment of collapse destroys the ionomer wrapping layer on the bubble and releases the trapped gas. At the same time, the cavitation effect causes the ionomer molecular chain to break and reorganize, forming a denser covering layer.

[0096] Hot pressing gradient pressurization: A three-stage pressurization strategy is adopted during the hot pressing stage:

[0097] Initial stage (0-10 seconds): Apply low pressure (0.5 MPa) to make the proton exchange membrane and the catalyst layer initially contact, to avoid bubbles from bursting under pressure and causing interface cracks;

[0098] Gradient pressure increase stage (10-30 seconds): The pressure increases from 0.1 MPa / s to 1.2 MPa. The gradual compression causes the residual microbubbles to migrate laterally along the membrane surface to the edge and be discharged.

[0099] Holding pressure stage (30-60 seconds): Maintain 1.2 MPa to allow the ionomer to flow fully and fill the interfacial voids.

[0100] (2) When the double layer capacitance value shows periodic fluctuations during the activation process (fluctuation amplitude > 15%, cycle of about 2-4 hours), it indicates that the volatility of the slurry solvent causes the dynamic change of the ionomer network structure. At this time, the following adjustments are performed:

[0101] Solvent ratio optimization: The volume ratio of water to ethanol in the slurry was adjusted from the initial 3:1 to 2.5:1 (±0.2). Increasing the ethanol ratio can reduce the surface tension of the solvent, slow down the shrinkage stress of the ionomer during the drying process, and inhibit the periodic structural relaxation caused by differences in solvent evaporation rates. At the same time, the rapid volatilization of ethanol forms micro-nano channels inside the catalytic layer, enhancing the gas diffusion capacity.

[0102] Extending the hot pressing holding time: Extending the hot pressing holding time from 60 seconds to 90-120 seconds allows the ionomer to fully creep and flow at high temperatures (130-140°C). Extending the holding time can eliminate the interfacial viscoelastic memory effect caused by solvent residue. Example 2

[0103] On the basis of Example 1, in order to implement the above method, refer to Figure 2As shown, the present application also discloses a hydrogen fuel cell preparation system based on multi-link feedback control, comprising:

[0104] The electromagnetic field assisted coating module is used to evenly coat the catalyst slurry on the surface of the proton exchange membrane to form the cathode and anode catalyst layers, including:

[0105] Pulsed alternating electromagnetic field generators are placed on both sides of the coating machine roller;

[0106] Coating parameter dynamic adjustment unit, which receives external feedback instructions in real time and adjusts electromagnetic field parameters;

[0107] The hot pressing molding-resistance monitoring module is used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including:

[0108] Distributed resistance sensor arrays are embedded in the upper and lower template surfaces of the hot press;

[0109] The resistance data analysis unit sets trigger conditions based on the distributed resistance sensor array and provides coating parameter adjustment instructions;

[0110] A short-term parameter control interface transmits adjustment instructions to the coating parameter dynamic adjustment unit of the electromagnetic field assisted coating module;

[0111] The Electrochemical Activation-Transfer Learning Optimization module is used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmarks, including:

[0112] High-frequency impedance spectrum acquisition unit, configured in the activation equipment;

[0113] Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm;

[0114] Long-term parameter update interface, which writes the correction value into the slurry formula database of the electromagnetic field assisted coating module and the process parameter library of the molding module.

[0115] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0117] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0118] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0119] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for preparing a hydrogen fuel cell based on multi-link feedback control, characterized in that: include: The catalyst slurry is evenly coated on the surface of the proton exchange membrane through an electromagnetic field-assisted coating process to form cathode and anode catalyst layers. During the coating process, a pulsed alternating electromagnetic field is applied to orient the platinum nanoparticles in the catalyst slurry along the direction of the magnetic field. During the hot pressing process, the interfacial contact resistance distribution of the membrane electrode assembly is monitored in real time, and the short-term coating parameters of the electromagnetic field-assisted coating process are dynamically adjusted based on the resistance data. This includes: embedding a distributed resistance sensor array on the upper and lower template surfaces of the hot press, with each sensor spaced evenly apart. The distributed resistance sensor array generates a real-time resistance thermal map of each region of the membrane electrode assembly. If the standard deviation of the resistance distribution exceeds 0.5 mΩ·cm², the electromagnetic field-assisted coating process is triggered, and the electromagnetic field frequency, intensity, or output mode are adjusted. Electrochemical impedance spectroscopy data is collected online during the electrochemical activation stage, and the long-term process benchmark values ​​of the catalyst slurry and the hot pressing process in the electromagnetic field-assisted coating process are reversely optimized through the transfer learning algorithm. The implementation of the transfer learning algorithm includes the following steps: constructing a convolutional neural network model, setting the input data as the charge transfer resistance, double-layer capacitance and real-time temperature in the electrochemical impedance spectroscopy and the activation stage; setting the training data set to include coating parameters in historical production, including: platinum loading and ionomer content, hot pressing pressure and hot pressing temperature and the corresponding charge transfer resistance and double-layer capacitance; the output of the neural network model is the adjustment value of the platinum loading and ionomer content in the catalyst slurry, and the compensation value of the hot pressing temperature and pressure in the hot pressing process.

2. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 1, characterized in that: The electromagnetic field assisted coating process includes: Before coating, the catalyst slurry is pre-sheared to form a shear densified structure; During the coating process, a dynamic rotating magnetic field is applied synchronously to control the direction of the magnetic field to rotate periodically, and the rotation angle is alternately switched between 0 and 180 degrees.

3. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 2, characterized in that: During the coating process, an ultrasonic cavitation generator is also used to destroy the ionomer-encapsulated bubbles in the slurry at a certain frequency.

4. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 2, characterized in that: During the coating process, a temperature gradient field is set in the coating area along the coating direction, gradually decreasing from 40°C at the starting end of the coating to 25°C at the ending end, with a gradient rate of 2-5°C / cm.

5. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 1, characterized in that: If the high resistance area is concentrated at the edge of the membrane, increase the electromagnetic field frequency by 10% to 20%; If the high resistance areas are scattered, increase the magnetic field strength by 0.1-0.2 T and increase the slurry viscosity by 5%-8%; If the resistance low value areas and high value areas are distributed alternately, the electromagnetic field is switched to an intermittent pulse mode.

6. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 1, characterized in that: When the charge transfer resistance is greater than 0.25 Ω·cm² and the double layer capacitance is less than 20 mF / cm², the platinum loading is controlled to increase by 0.03-0.05 mg / cm², and the hot pressing temperature is adjusted to increase by 5°C. When the charge transfer resistance is less than 0.15 Ω·cm² and the double layer capacitance is greater than 30 mF / cm², the ionomer content is controlled to be reduced by 5% to 8%, and the hot pressing pressure during hot pressing is adjusted to be reduced by 0.5 MPa. When the charge transfer resistance and the double layer capacitance fluctuate nonlinearly, the electromagnetic field frequency in the electromagnetic field-assisted coating process is controlled to randomly perturb to break up the slurry agglomeration.

7. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 1, characterized in that: Inverse optimization of process parameters also includes the following collaborative strategies: If the transfer learning model detects that the charge transfer resistance value drops rapidly at the initial stage of activation, an ultrasonic cavitation generator is used at a certain frequency to destroy the ionomer-encapsulated bubbles in the slurry, and a gradient pressure is applied during the hot pressing process; If the transfer learning model detects that the double-layer capacitance value fluctuates periodically with activation time, the ratio of water and ethanol in the catalyst slurry is adjusted, and the hot pressing holding time of the hot pressing process is extended.

8. A hydrogen fuel cell production system based on multi-link feedback control, characterized by: include: The electromagnetic field assisted coating module is used to evenly coat the catalyst slurry on the surface of the proton exchange membrane to form the cathode and anode catalyst layers, including: Pulsed alternating electromagnetic field generators are placed on both sides of the coating machine roller; Coating parameter dynamic adjustment unit, which receives external feedback instructions in real time and adjusts electromagnetic field parameters; The hot pressing molding-resistance monitoring module is used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including: Distributed resistance sensor arrays are embedded in the upper and lower template surfaces of the hot press; The resistance data analysis unit sets trigger conditions based on the distributed resistance sensor array and provides coating parameter adjustment instructions; A short-term parameter control interface transmits adjustment instructions to the coating parameter dynamic adjustment unit of the electromagnetic field assisted coating module; The Electrochemical Activation-Transfer Learning Optimization module is used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmarks, including: High-frequency impedance spectrum acquisition unit, configured in the activation equipment; Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm; Long-term parameter update interface, which writes the correction value into the slurry formula database of the electromagnetic field assisted coating module and the process parameter library of the molding module.

Images