Preparation method and system of hydrogen fuel cell based on multi-link feedback regulation

Abstract

The invention relates to a hydrogen fuel cell preparation method and system based on multi-link feedback regulation, and the method comprises the steps: uniformly coating the surface of a proton exchange membrane with catalyst slurry through an electromagnetic field auxiliary coating technology to form a cathode and anode catalyst layer; the interface contact resistance distribution of the membrane electrode assembly is monitored in real time in the hot press molding process, and short-term coating parameters of the electromagnetic field auxiliary coating technology are dynamically adjusted according to resistance data; electrochemical impedance spectroscopy data are collected on line in the electrochemical activation stage, and catalyst slurry in the electromagnetic field auxiliary coating process and a long-term process reference value in the hot press molding process are reversely optimized through a transfer learning algorithm; the system is used for realizing the method and comprises an electromagnetic field auxiliary coating module, a hot press molding-resistance monitoring module and an electrochemical activation-transfer learning optimization module. Through a multi-link feedback regulation and control mechanism, the problems of non-uniform catalyst coating and process parameter curing in the membrane electrode preparation process are effectively solved.

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 an efficient and clean energy conversion device, hydrogen fuel cells have broad application prospects in new energy vehicles, distributed power generation and other fields. Among them, membrane electrode (MEA) is the core component of hydrogen fuel cells, and its preparation process directly affects the performance and life of the battery. At present, the traditional membrane electrode preparation method mainly includes key links such as catalyst coating, hot pressing and electrochemical activation. However, there is a significant technical problem in the preparation process of the existing technology: uneven coating of catalyst slurry and curing of process parameters lead to uneven distribution of membrane electrode interface contact resistance, which in turn affects battery performance and durability.

[0003] This problem arises mainly from two reasons: First, during the catalyst coating stage, it is difficult to accurately control the distribution of the catalyst slurry using traditional coating processes (such as spraying and scraping), which can easily lead to local agglomeration or uneven thickness, affecting the interface contact quality of subsequent hot pressing. 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, but fails to form a closed-loop optimization with 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, ultimately leading to poor consistency of battery batches and performance improvement bottlenecks. Summary of the invention

[0004] To this end, the present invention provides a method and system for preparing a hydrogen fuel cell based on multi-link feedback control, which effectively solves the problems of uneven catalyst coating and process parameter solidification in the membrane electrode preparation process through a multi-link feedback control mechanism.

[0005] In order 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: 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; 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; Electrochemical impedance spectroscopy data is collected online during the electrochemical activation stage, and the long-term process benchmark values ​​in the catalyst slurry and hot pressing process in the electromagnetic field-assisted coating process are reversely optimized through the transfer learning algorithm.

[0006] In one embodiment of the present invention, the electromagnetic field assisted coating process comprises: Before coating, the catalyst slurry is pre-sheared to form a shear densified structure; 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.

[0007] In one embodiment of the present invention, during the coating process, an ultrasonic cavitation generator is also used to destroy the ionomer-encapsulated bubbles in the slurry at a certain frequency.

[0008] 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 coating to 25°C at the ending end, with a gradient rate of 2-5°C / cm.

[0009] In one embodiment of the present invention, real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly includes: A distributed resistance sensor array is embedded on the upper and lower template surfaces of the hot press, and each sensor is arranged at equal intervals. The distributed resistance sensor array is used to generate a resistance thermal map of each area of ​​the membrane electrode assembly in real time; If the standard deviation of the resistance distribution exceeds 0.5 mΩ·cm², the electromagnetic field assisted coating process is triggered to adjust the electromagnetic field frequency, intensity or output mode.

[0010] In one embodiment of the present invention, if the high resistance area is concentrated at the edge of the film, the frequency of the electromagnetic field is increased by 10% to 20%; If the high resistance area is 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 the intermittent pulse mode.

[0011] In one embodiment of the present invention, the implementation of the transfer learning algorithm includes the following steps: Construct a convolutional neural network model, and set the input data to be the charge transfer resistance, double layer capacitance and real-time temperature in the activation stage in the electrochemical impedance spectroscopy; The training data set is 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 outputs of the neural network model are the adjusted values ​​of platinum loading and ionomer content in the catalyst slurry, and the compensated values ​​of hot pressing temperature and pressure during hot pressing.

[0012] In one embodiment of the present invention, when the charge transfer resistance value is greater than 0.25 Ω·cm² and the double electric layer capacitance value 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 of hot pressing molding is adjusted to increase by 5°C; When the charge transfer resistance value is less than 0.15 Ω·cm² and the double layer capacitance value is greater than 30 mF / cm², the ionomer content is controlled to be reduced by 5% to 8%, and the hot pressing pressure of hot pressing molding 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 be randomly disturbed to break up the slurry agglomeration.

[0013] In one embodiment of the present invention, the reverse 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 to destroy the ionomer-encapsulated bubbles in the slurry at a certain frequency, 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 the activation time, the ratio of water to ethanol in the catalyst slurry is adjusted, and the hot pressing holding time of the hot pressing process is extended.

[0014] In order to solve the above technical problems, the present invention also provides a hydrogen fuel cell preparation system based on multi-link feedback control, comprising: 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: Pulse alternating electromagnetic field generators are placed on both sides of the coating machine roller; The coating parameter dynamic adjustment unit receives external feedback instructions in real time and adjusts the electromagnetic field parameters; Hot Pressing-Resistance Monitoring Module, used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including: A distributed resistance sensor array is embedded in the upper and lower template surfaces of the hot press; A resistance data analysis unit, which sets trigger conditions according to the distributed resistance sensor array and provides coating parameter adjustment instructions; A short-term parameter control interface transmits adjustment instructions to a coating parameter dynamic adjustment unit of an electromagnetic field assisted coating module; Electrochemical activation-transfer learning optimization module, used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmark values, including: A high-frequency impedance spectrum acquisition unit is configured in the activation device; Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm; Long-term parameter update interface, 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.

[0015] The above technical solution of the present invention has the following advantages compared with the prior art: 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 of the three links of "coating-hot pressing-activation".

[0016] 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 over-thickness or over-thinness, thereby improving the structural consistency of the catalytic layer.

[0017] Secondly, in the electrochemical activation stage, the impedance spectrum data is reversely mapped to the previous process parameter space through the transfer learning algorithm, breaking through the limitation of the traditional trial and error method relying on a fixed data set. The algorithm can identify the influence of differences in material properties of different batches on the process benchmark value.

[0018] Through the above-mentioned two cross-link parameter collaborative optimization mechanisms, the dual regulation of short-term process compensation and long-term benchmark iteration is achieved in principle, which not only ensures the manufacturing consistency of a single product, but also continuously improves the process robustness through a data closed loop, ultimately forming an adaptive hydrogen fuel cell preparation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 It is a flow chart of the steps of the method for preparing a hydrogen fuel cell based on multi-link feedback control of the present invention; 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

[0020] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention. Embodiment 1

[0021] Reference Figure 1As shown, the present invention discloses a method for preparing a hydrogen fuel cell based on multi-link feedback control, the core of which is to construct a three-stage closed-loop process control system of "coating-hot pressing-activation", and to achieve process guidance and dynamic adjustment of process parameters in different process stages through real-time data feedback and algorithm optimization. The specific implementation process and technical mechanism are as follows: 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.

[0022] In this stage, the electromagnetic field is used to exert a directional force on the charged particles (such as platinum nanoparticles) in the catalyst slurry, reducing the difference in capillary contraction during the drying process of the slurry, avoiding cracks or holes in the catalyst layer, thereby improving the electronic conduction capacity of the catalyst 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 arrange in a preset direction, while suppressing the non-uniform encapsulation of the ionomer and the catalyst, thereby forming an orderly three-phase reaction interface on a microscopic scale; the catalyst particles with a directional arrangement 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.

[0023] 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 electromagnetic field-assisted coating (such as electromagnetic field strength and frequency) are dynamically adjusted according to the resistance data.

[0024] When the hot press applies pressure, the interface contact quality between the proton exchange membrane and the catalyst layer directly affects the transmission efficiency of the reaction gas and protons. In this embodiment, the distributed resistance sensor array can locate the poor contact area (such as the high resistance area); if the local resistance is abnormally increased, it indicates that the fit between the catalyst layer and the proton exchange membrane in this area is insufficient. At this time, the electromagnetic field force in the local area can be targeted to promote the plastic deformation of the catalyst layer during the hot pressing process, fill the microscopic gaps in the interface, and reduce the spatial discreteness of the contact resistance. This instant feedback mechanism breaks through the limitations of the fixed parameters of the traditional hot pressing process, links the hot pressing process with the coating process, and solves the problem of uneven interface contact caused by coating problems from the root.

[0025] 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 ​​of hot pressing (such as platinum loading, ionomer content, hot pressing temperature) are reversely optimized through the transfer learning algorithm.

[0026] The charge transfer resistance and double-layer capacitance in the electrochemical impedance spectrum directly reflect the utilization rate of the active sites in the catalyst layer and the effective area of ​​the three-phase interface. The transfer learning algorithm establishes a multi-parameter coupled optimization model by associating historical process parameters with electrochemical impedance spectrum data features. The cross-stage correlation model between electrochemical impedance spectrum data and process parameters is used to break through the data island problem in the coating, hot pressing, and activation links in the traditional process. The charge transfer characteristics are traced back to material ratio defects or other process defects to achieve accurate 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, and then adjust the long-term baseline parameters. This data-driven optimization method breaks through the empirical dependence of the traditional trial and error method and realizes continuous iterative upgrades of process parameters from a system level.

[0027] Specifically, through the above steps S10 to S30, a dual control mechanism of "short-cycle compensation" and "long-cycle optimization" is formed: the short-term adjustment in the hot pressing stage directly acts on the current production unit, and the coating defects are compensated by the instant correction of the electromagnetic field parameters; the reverse optimization in the activation stage is based on the accumulation of cross-batch data, and the long-term process benchmark value is updated to improve the overall process robustness. The synergy of the two forms a closed-loop system of "real-time correction-historical learning", which not only ensures the quality consistency of a single product, but also continuously improves the process upper limit through data accumulation.

[0028] Specifically, the invention of the present application also improves the electromagnetic field assisted coating process, including: The catalyst slurry is pre-sheared before coating, and the soft agglomerates of platinum nanoparticles in the slurry are destroyed by mechanical shear force, and the ionomer molecular chains are caused to stretch along the shear direction to form a shear densification structure. This step changes the rheological properties of the slurry, causing it to transform from an initial pseudoplastic fluid to a shear thinning state. The shear densification effect produced by pre-shearing can reduce the 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, thereby reducing the risk of cracking of the catalyst layer. In addition, this densified structure provides a stable medium environment for the subsequent electromagnetic field action, avoiding fluctuations in coating thickness due to local viscosity differences in the slurry.

[0029] During the coating process, a pulsed alternating electromagnetic field is applied, and the conductive properties of platinum nanoparticles are used 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 be oriented along the direction of the magnetic field. The oriented platinum particles form a continuous electron conduction path, effectively reducing the bulk resistance of the catalytic layer; among them: the pulsed design (such as an intermittent magnetic field with a frequency of 1-10 kHz) can avoid excessive migration and accumulation of particles due to continuous force, and at the same time use the alternating characteristics to maintain the dynamic balance of the slurry fluidity to ensure the linear controllability of the coating thickness. Compared with the traditional static magnetic field, the alternating magnetic field can cover a wider particle distribution area by periodically changing direction, reducing the loss of catalytic active surfaces caused by a single orientation.

[0030] A dynamic rotating magnetic field is applied synchronously to make the direction of the magnetic field rotate periodically within the range of 0° to 180° (for example, the direction is switched every 5 seconds). This design breaks the excessive orientation of platinum particles in a single direction through magnetic field disturbance in the spatial dimension. When the direction of the magnetic field rotates, the migration path of the platinum particles is deflected to form a multi-dimensional interwoven mesh structure, forming a redundant conduction path in the catalyst layer. When the local area is structurally damaged due to heat pressure or activation, the multi-dimensional network can provide a bypass conduction channel to improve the stability of battery operation. This structure increases the tortuosity factor of the three-phase reaction interface at the microscopic level, expands the surface area of ​​the gas diffusion channel, and avoids the anisotropy of the mechanical strength of the catalyst layer caused by the unidirectional arrangement; in addition, the alternating switching of the rotation angle also alleviates the edge effect of the slurry in the coating direction, so that the thickness uniformity of the coating layer in the width direction is improved.

[0031] 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.

[0032] Specifically, the introduction of ultrasonic cavitation effect can further optimize the microscopic uniformity of the catalyst slurry and eliminate the internal defects of the slurry through the synergistic effect of the acoustic field and the electromagnetic field. During the coating process, the ultrasonic cavitation generator acts on the slurry layer at a specific frequency (for example, 20-40 kHz). When the ultrasonic wave propagates in the slurry, periodic compression and rarefaction areas are formed, generating cavitation bubbles. When the bubbles collapse, the local instantaneous pressure can reach hundreds of MPa. This impact force can break the over-thick coating of the catalyst particles by the ionomer and break the micron-sized bubbles remaining in the slurry.

[0033] 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 be used to directionally destroy the wrapping layer and bubbles, thereby improving the uniformity of ionomer distribution on the surface of platinum particles and reducing structural voids after the catalytic layer is cured.

[0034] The cavitation effect of ultrasound and the pulsed alternating electromagnetic field form a temporal and spatial complementarity: spatial dimension: the cavitation effect of ultrasound acts on the overall thickness direction of the slurry, while the electromagnetic field mainly affects the arrangement of platinum particles on the surface and near the surface area; time dimension: the continuous action of ultrasound and the pulse characteristics of the electromagnetic field form a dynamic superposition, continuously optimizing the microstructure during the slurry leveling stage. This synergistic mechanism can avoid the problem of local over-treatment under the action of a single physical field; for example, when the electromagnetic field drives the migration of platinum particles, ultrasound simultaneously removes the ionomer barrier on the migration path, thereby improving the efficiency of directional arrangement.

[0035] 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 characteristics 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 to prevent the arranged particles from secondary displacement due to the flow of the slurry.

[0036] 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: cooperate with the rapid switching of the rotating magnetic field (such as shortening the cycle from 0° to 180°), and use low viscosity conditions to achieve multi-directional rapid migration of platinum particles; the low temperature zone at the ending end: slow down the switching frequency of the magnetic field, so that the particles can gradually and stably orient in the high viscosity medium; this temperature-magnetic field linkage control avoids the problem of orientation attenuation caused by the mismatch between the slurry solidification rate and the electromagnetic response speed under a single temperature field.

[0037] Specifically, the present invention further discloses a process for real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly, including: Micro resistor 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 through 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 position distribution of high resistance areas (such as red patches) and low resistance areas (such as blue areas) through color gradients.

[0038] The standard deviation threshold of the contact resistance distribution is set to 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.

[0039] Specifically, according to actual usage, the mapping rules of several common resistance distribution characteristics and electromagnetic field parameter adjustment are refined: (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 judged that the edge climbing effect in the coating stage causes the catalyst layer thickness to decrease. At this time, the electromagnetic field frequency is increased by 10%-20% (for example, from 1kHz to 1.1-1.2kHz), and 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 conductor surface, which prompts the platinum particles in the edge area to redistribute to fill the thickness gap, while suppressing the waste of slurry in the center area due to excessive thickness.

[0040] (2) If the high resistance area is randomly scattered (such as discrete red spots appearing in the thermal map), it indicates that there is local agglomeration in the slurry or the ionomer is too thick. At this time, the magnetic field strength is increased by 0.1-0.2T (for example, 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 of the agglomerates and promote the separation of platinum particles from the agglomerates; while moderately increasing the viscosity (such as by adding thickeners in real time) can inhibit the secondary agglomeration of particles during migration and ensure the dispersion compensation effect.

[0041] (3) When low resistance areas and high resistance areas are distributed alternately (such as striped red and blue 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 a 200 ms pulse width alternating with a 100 ms interval), and the strong magnetic field at the pulse peak (such as 1.2 T) is used to break through the interface adsorption barrier. The intermittent period allows the slurry stress to relax, avoiding excessive migration and accumulation of particles caused by continuous high-intensity magnetic fields.

[0042] Specifically, the present invention further discloses an implementation process of the transfer learning algorithm, comprising the following steps: 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 characterizes 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 hidden layer of the network 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.

[0043] The training data set integrates multi-dimensional data from historical production, including: Coating parameters: platinum loading (affects charge transfer resistance), ionomer content (affects double layer capacitance); Hot pressing parameters: pressure (determines the tightness of the interface contact), temperature (affects the redistribution of the ionomer); Performance data: Measured values ​​of charge transfer resistance and double layer capacitance of the corresponding batch.

[0044] 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 associations between parameters.

[0045] The model output layer is set to the adjusted values ​​of the process parameters, including: Platinum loading adjustment: directly related to the charge transfer resistance value, because platinum particles are active site carriers of charge transfer reactions; Ionomer content adjustment: by adjusting the thickness of the ionomer, the proton transport path is affected and the double layer capacitance is indirectly modified; Hot pressing temperature / pressure compensation: Correct the balance between interface contact resistance and catalyst layer porosity.

[0046] Specifically, according to the actual usage, the corresponding rules of common electrochemical index combinations and process parameter adjustments are refined, and the targeted optimization of process parameters is achieved through classification threshold setting: (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 active site density and reduce the charge transfer activation energy; and the hot pressing temperature is increased by 5°C: this causes the ionomer to melt and redistribute 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.

[0047] (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. At this time, the following measures should be taken: ionomer content is reduced by 5%-8% to reduce the volume share of ionomer and expand the gas diffusion channel; and the hot pressing pressure is reduced 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.

[0048] (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 randomly disturbed (e.g., randomly fluctuates 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 that have been formed in the slurry, while avoiding the aggravation of resonant agglomeration caused by regular frequency switching.

[0049] Furthermore, on the basis of the above, the inventors also developed differentiated control strategies based on two typical abnormal modes commonly seen in production: (1) When the transfer learning model detects that the rate of decrease of the charge transfer resistance value exceeds the threshold value in the initial activation period (0-30 minutes), it indicates that there are micron-sized bubbles wrapped by ionomers in the slurry, and the following coordinated adjustments are triggered: 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 polymer wrapping layer on the bubble and releases the trapped gas. At the same time, the cavitation effect causes the polymer molecular chain to break and reorganize to form a denser covering layer.

[0050] Hot pressing gradient pressurization: A three-stage pressurization strategy is adopted in the hot pressing stage: 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; Gradient pressure increase stage (10-30 seconds): The pressure increases from 0.1 MPa / s to 1.2 MPa. The progressive compression causes the residual microbubbles to migrate laterally along the membrane surface to the edge and be discharged. Holding pressure stage (30-60 seconds): Maintain 1.2 MPa to allow the ionomer to flow fully and fill the interface gaps.

[0051] (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 difference of the slurry solvent leads to dynamic changes in the ionomer network structure. At this time, the following adjustments are performed: Optimization of solvent ratio: 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 the difference in solvent evaporation rate. At the same time, the rapid volatilization of ethanol forms micro-nano channels inside the catalyst layer, enhancing the gas diffusion capacity.

[0052] Extend the hot pressing holding time: Extend the hot pressing holding time from 60 seconds to 90-120 seconds to allow the ionomer to fully creep and flow at high temperature (130-140°C). Extending the holding time can eliminate the interfacial viscoelastic memory effect caused by solvent residue. Embodiment 2

[0053] On the basis of the first embodiment, in order to implement the above method, refer to Figure 2 As shown, the present application also discloses a preparation system of a hydrogen fuel cell based on multi-link feedback control, comprising: 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: Pulse alternating electromagnetic field generators are placed on both sides of the coating machine roller; The coating parameter dynamic adjustment unit receives external feedback instructions in real time and adjusts the electromagnetic field parameters; Hot Pressing-Resistance Monitoring Module, used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including: A distributed resistance sensor array is embedded in the upper and lower template surfaces of the hot press; A resistance data analysis unit, which sets trigger conditions according to the distributed resistance sensor array and provides coating parameter adjustment instructions; A short-term parameter control interface transmits adjustment instructions to a coating parameter dynamic adjustment unit of an electromagnetic field assisted coating module; Electrochemical activation-transfer learning optimization module, used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmark values, including: A high-frequency impedance spectrum acquisition unit is configured in the activation device; Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm; Long-term parameter update interface, 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.

[0054] 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 adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0055] 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 box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate 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.

[0056] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0057] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0058] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still 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; 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; Electrochemical impedance spectroscopy data is collected online during the electrochemical activation stage, and the long-term process benchmark values ​​in the catalyst slurry and hot pressing process in the electromagnetic field-assisted coating process are reversely optimized through the transfer learning algorithm.

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 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.

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: Real-time monitoring of the interface contact resistance distribution of the membrane electrode assembly includes: A distributed resistance sensor array is embedded on the upper and lower template surfaces of the hot press, and each sensor is arranged at equal intervals. The distributed resistance sensor array is used to generate a resistance thermal map of each area of ​​the membrane electrode assembly in real time; If the standard deviation of the resistance distribution exceeds 0.5 mΩ·cm², the electromagnetic field assisted coating process is triggered to adjust the electromagnetic field frequency, intensity or output mode.

6. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 5, 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 area is 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 the intermittent pulse mode.

7. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 1, characterized in that: The implementation of the transfer learning algorithm includes the following steps: Construct a convolutional neural network model, and set the input data to be the charge transfer resistance, double layer capacitance and real-time temperature in the activation stage in the electrochemical impedance spectroscopy; The training data set is 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 outputs of the neural network model are the adjusted values ​​of platinum loading and ionomer content in the catalyst slurry, and the compensated values ​​of hot pressing temperature and pressure during hot pressing.

8. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 7, characterized in that: When the charge transfer resistance value is greater than 0.25 Ω·cm² and the double layer capacitance value 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 of hot pressing molding is adjusted to increase by 5°C; When the charge transfer resistance value is less than 0.15 Ω·cm² and the double layer capacitance value is greater than 30 mF / cm², the ionomer content is controlled to be reduced by 5% to 8%, and the hot pressing pressure of hot pressing molding 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 be randomly disturbed to break up the slurry agglomeration.

9. The method for preparing a hydrogen fuel cell based on multi-link feedback control according to claim 7, characterized in that: The reverse 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 to destroy the ionomer-encapsulated bubbles in the slurry at a certain frequency, 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 the activation time, the ratio of water to ethanol in the catalyst slurry is adjusted, and the hot pressing holding time of the hot pressing process is extended.

10. A preparation system for a hydrogen fuel cell based on multi-link feedback control, characterized in that: 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: Pulse alternating electromagnetic field generators are placed on both sides of the coating machine roller; The coating parameter dynamic adjustment unit receives external feedback instructions in real time and adjusts the electromagnetic field parameters; Hot Pressing-Resistance Monitoring Module, used to monitor the interface contact resistance distribution of the membrane electrode assembly in real time and provide feedback to control coating parameters, including: A distributed resistance sensor array is embedded in the upper and lower template surfaces of the hot press; A resistance data analysis unit, which sets trigger conditions according to the distributed resistance sensor array and provides coating parameter adjustment instructions; A short-term parameter control interface transmits adjustment instructions to a coating parameter dynamic adjustment unit of an electromagnetic field assisted coating module; Electrochemical activation-transfer learning optimization module, used to collect electrochemical impedance spectroscopy data online and reversely optimize process benchmark values, including: A high-frequency impedance spectrum acquisition unit is configured in the activation device; Transfer learning algorithm engine, which provides reverse optimization correction through transfer learning algorithm; Long-term parameter update interface, 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