A hydrogen-oxygen mutual string early warning method and system based on intelligent membrane and sensing bipolar plate

Abstract

The application discloses a hydrogen-oxygen mutual string early warning method and system based on an intelligent membrane and a sensing bipolar plate, direct sensing of mechanical stress by the intelligent membrane realizes earliest early warning of membrane damage, early warning time is tens to hundreds of hours earlier than that of a traditional electrochemical monitoring method, and sufficient time window is provided for taking protective measures. Through distributed measurement of the sensing bipolar plate, fault positioning accuracy is improved from a stack level to a centimeter level, precise spatial positioning is realized, and maintenance cost is greatly reduced. Through multi-modal data fusion, limitations of single signal monitoring are overcome, normal working condition fluctuations and real fault precursors are effectively distinguished, based on precise positioning and early warning, maintenance of the fuel cell system is changed from regular maintenance or post-failure maintenance to state-based predictive maintenance, maintenance resource allocation is optimized, and system availability and full life cycle economy are improved.

Description

Technical Field

[0001] This invention relates to the field of fuel cell safety technology and condition monitoring, specifically to a method and system for early warning of hydrogen-oxygen crosstalk based on a smart membrane and a sensing bipolar plate. Background Technology

[0002] In the commercialization of proton exchange membrane fuel cells, hydrogen-oxygen intermingling has always been the most critical safety hazard. When the proton exchange membrane suffers microscopic damage due to factors such as chemical degradation, mechanical stress, or thermal aging, hydrogen and oxygen gases will mix, forming an explosive mixture inside the fuel cell stack. This mixture can lead to catastrophic consequences upon contact with the catalyst.

[0003] Existing technologies for monitoring and early warning of hydrogen-oxygen crosstalk have the following limitations:

[0004] The lag in diagnosis: While the most advanced online electrochemical impedance spectroscopy (EIS) technology can assess the health status of the membrane by monitoring the overall impedance of the fuel cell stack, it is essentially a "consequence-based" diagnosis. It detects the state after a measurable change in ionic conductivity has occurred in the membrane, at which point microscopic damage may have already formed, resulting in a limited warning window.

[0005] Lack of spatial resolution: Traditional EIS technology provides macroscopic average information for the entire fuel cell stack or large module. When anomalies are detected, it is impossible to quickly and accurately pinpoint which individual cell or even which localized area is experiencing a problem. This poses significant challenges to troubleshooting and predictive maintenance, necessitating a costly "holistic approach."

[0006] Limited sensing capabilities: Current monitoring methods primarily rely on electrochemical signals. They lack the ability to directly sense the physical state changes (such as microscopic strain and creep) that the membrane undergoes before failure. These physical changes are often precursors to the degradation of electrochemical performance; capturing such signals could significantly advance the warning time.

[0007] The "passive" role of core components: In current technology, proton exchange membranes and bipolar plates exist only as "passive" components to achieve power generation. They lack the ability to sense their own state and output signals; the assessment of their health status relies entirely on indirect measurements from external systems.

[0008] Therefore, there is an urgent need in this field for a technology that can sense risks from the source of materials and has high spatial resolution diagnostic capabilities to fill the monitoring gap between "microscopic damage occurrence" and "macroscopic performance degradation" and achieve truly early, localizable warnings. Summary of the Invention

[0009] To address this, the present invention provides a hydrogen-oxygen crosstalk early warning method and system based on a smart membrane and a sensing bipolar plate, in order to solve the technical problems of existing technologies that are difficult to perceive risks from the source of materials and lack high spatial resolution diagnostic capabilities.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] According to a first aspect of the present invention, a method for early warning of hydrogen-oxygen crosstalk based on a smart membrane and a sensing bipolar plate is provided, the method comprising:

[0012] Data acquisition steps: Physical sensing signals from the smart proton exchange membrane of the fuel cell and local electrochemical impedance spectroscopy data from the integrated sensing bipolar plate are acquired in parallel.

[0013] Feature extraction step: The physical sensing signal and local electrochemical impedance spectroscopy data are processed to extract feature parameters characterizing the mechanical strain state of the membrane and electrochemical feature parameters characterizing the local health state of the membrane, respectively.

[0014] Fusion diagnostic step: The extracted feature parameters are input into a pre-trained multimodal fusion diagnostic model. The model analyzes the correlation between physical sensing signals and electrochemical signals in the spatiotemporal dimension and outputs the probability and location information of the risk of hydrogen-oxygen crosstalk inside the fuel cell.

[0015] Early warning output steps: Based on the risk probability and location information output from the fusion diagnosis steps, generate and issue early warning information of different levels.

[0016] Furthermore, the physical sensing signal is an optical signal or an electrical signal; wherein,

[0017] The optical signal is generated by the change in fluorescence intensity caused by the strain-induced rupture of the fluorescent microcapsules 12 embedded in the smart proton exchange membrane.

[0018] The electrical signal is generated by the change in resistance caused by strain in the conductive nanonetwork constructed within the smart proton exchange membrane.

[0019] Furthermore, the local electrochemical impedance spectroscopy data is obtained by sequentially activating different microelectrode sensing units on the integrated sensing bipolar plate through a multiplexed scanning system, thereby constructing a spatial distribution map of the membrane health state.

[0020] Furthermore, the pre-trained multimodal fusion diagnostic model is a graph convolutional neural network model based on an attention mechanism; wherein,

[0021] The input to the model includes feature vectors with sensing units as nodes, and a graph structure constructed based on the bipolar plate flow field layout.

[0022] The model learns spatial dependencies through graph convolutional layers, learns temporal variation trends through recurrent neural network layers, and adaptively fuses feature parameters from different sources through an attention mechanism.

[0023] Furthermore, prior to the data acquisition step, an initialization and self-test step is included, specifically:

[0024] After starting the fuel cell and performing a self-test to confirm that all sensing channels are working properly, the initial parameters of the smart membrane and sensing bipolar plate are read to establish baseline reference values.

[0025] According to a second aspect of the present invention, a hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate is provided, the system comprising:

[0026] A smart proton exchange membrane, which has embedded sensing materials to convert the membrane’s mechanical strain into a measurable physical signal;

[0027] An integrated sensing bipolar plate integrates a microelectrode sensing array in its flow field region for independent electrochemical impedance spectroscopy measurements of its local area.

[0028] A data acquisition module is configured to communicate with the smart proton exchange membrane and the integrated sensing bipolar plate to acquire the physical signal and local electrochemical impedance spectroscopy data in parallel.

[0029] The feature extraction module is used to process the physical sensing signal and the local electrochemical impedance spectroscopy data to extract feature parameters characterizing the mechanical strain state of the membrane and electrochemical feature parameters characterizing the local health state of the membrane, respectively.

[0030] The fusion diagnostic module is used to input the extracted feature parameters into a pre-trained multimodal fusion diagnostic model. The model analyzes the correlation between physical sensing signals and electrochemical signals in the spatiotemporal dimension and outputs the probability and location information of hydrogen-oxygen crosstalk risk inside the fuel cell.

[0031] The early warning output module is used to generate and issue early warning information based on the output results of the data processing and diagnosis module.

[0032] Furthermore, the sensing material in the intelligent proton exchange membrane is a fluorescent microcapsule, which has urea-formaldehyde resin as the wall material and a core material containing a mixed solution of rhodamine B fluorescent dye and dimethylphenylpyrazolone quencher. The microcapsule has a particle size of 2±0.5μm and is uniformly dispersed in the proton exchange membrane resin at a ratio of 1.5-2.5wt%.

[0033] Furthermore, the sensing material in the intelligent proton exchange membrane is a sparse three-dimensional conductive network composed of silver nanowires, with a diameter of 50 nm, a length of 20-50 μm, and an areal density of 0.2-0.4 mg / cm².

[0034] Furthermore, the microelectrode sensing unit on the integrated sensing bipolar plate is a three-electrode system, including a circular platinum microelectrode working electrode with a diameter of 500μm, a ring platinum counter electrode surrounding it, and a micro reversible hydrogen reference electrode; the sensing unit is integrated on a metal or graphite bipolar plate using MEMS technology and protected by a silicon nitride insulating layer.

[0035] Furthermore, the data acquisition module includes a multiplexed scanning subsystem, which is built based on an analog switch matrix, supports scanning of at least 128 sensing channels, has a full-channel scanning time of less than 30 seconds, and has a crosstalk suppression capability of greater than 60dB.

[0036] The early warning output module communicates with the fuel cell main controller via a CAN bus to send early warning information and location maintenance suggestions to the user interface, and / or to link with the active explosion suppression system to activate the protection program when an emergency risk is detected.

[0037] The present invention has the following advantages:

[0038] This invention achieves early warning of membrane damage through direct sensing of mechanical stress via a smart membrane, providing an early warning time tens to hundreds of hours earlier than traditional electrochemical monitoring methods, thus offering ample time for protective measures. Distributed measurement via sensing bipolar plates improves fault location accuracy from the "stack level" to the "centimeter level," enabling precise spatial positioning and making predictive maintenance and targeted repair possible, significantly reducing maintenance costs. Multimodal data fusion overcomes the limitations of single-signal monitoring, effectively distinguishing between normal operating condition fluctuations and true fault precursors. Based on precise location and early warning, fuel cell system maintenance shifts from periodic or post-fault maintenance to condition-based predictive maintenance, optimizing maintenance resource allocation and improving system availability and overall lifecycle economics. Attached Figure Description

[0039] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0040] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0041] Figure 1 A schematic diagram illustrating the working process of a hydrogen-oxygen inter-contamination early warning system based on a smart membrane and a sensing bipolar plate provided by the present invention.

[0042] Figure 2 The present invention provides a logical architecture and data flow diagram of a hydrogen-oxygen inter-connection early warning system based on a smart membrane and a sensing bipolar plate.

[0043] Figure 3 A schematic diagram of the intelligent membrane structure in a hydrogen-oxygen inter-contamination early warning system based on an intelligent membrane and a sensing bipolar plate provided by the present invention.

[0044] Figure 4 This invention provides a schematic diagram of the sensing bipolar plate structure in a hydrogen-oxygen inter-contamination early warning system based on a smart membrane and a sensing bipolar plate.

[0045] Figure reference numerals: 11, proton exchange membrane substrate; 12, fluorescent microcapsule; 13, core material (mixed solution of fluorescent dye and quencher); 14, silver nanowire conductive network; 21, bipolar plate substrate; 22, inlet region; 23, central region; 24, outlet region; 25, sensing unit; 26, sensing signal circuit. Detailed Implementation

[0046] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] To address the technical problems mentioned above, such as the inability of existing technologies to detect risks at the source of materials and the lack of high spatial resolution diagnostic capabilities.

[0048] refer to Figure 1 and Figure 2This invention discloses a hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate. This system, by embedding sensing capabilities into the core components of a fuel cell, constructs a distributed intelligent monitoring network from the "material level" to the "system level." Working in conjunction with the main control unit, this system achieves ultra-early warning and precise location of hydrogen-oxygen crosstalk risks. The overall architecture of the system embodies a closed-loop logic of "sensing-transmission-decision." The system includes:

[0049] A smart proton exchange membrane, which has embedded sensing materials to convert the membrane’s mechanical strain into a measurable physical signal;

[0050] An integrated sensing bipolar plate integrates a microelectrode sensing array in its flow field region for independent electrochemical impedance spectroscopy measurements of its local area.

[0051] A data acquisition module is configured to communicate with the smart proton exchange membrane and the integrated sensing bipolar plate to acquire the physical signal and local electrochemical impedance spectroscopy data in parallel.

[0052] The feature extraction module is used to process the physical sensing signal and the local electrochemical impedance spectroscopy data to extract feature parameters characterizing the mechanical strain state of the membrane and electrochemical feature parameters characterizing the local health state of the membrane, respectively.

[0053] The fusion diagnostic module is used to input the extracted feature parameters into a pre-trained multimodal fusion diagnostic model. The model analyzes the correlation between physical sensing signals and electrochemical signals in the spatiotemporal dimension and outputs the probability and location information of hydrogen-oxygen crosstalk risk inside the fuel cell.

[0054] The early warning output module is used to generate and issue early warning information based on the output results of the data processing and diagnosis module.

[0055] This invention is based on the core idea of ​​multi-physics field coupled sensing and spatiotemporal data fusion diagnosis, and achieves a fundamental improvement in monitoring capabilities through material innovation and structural innovation.

[0056] The principle of strain sensing at the material level is as follows:

[0057] The smart proton exchange membrane directly converts mechanical strain into measurable physical signals through internally implanted sensitive materials. When the membrane material undergoes microscopic deformation due to hydrothermal cycling, assembly stress, or material aging, this deformation alters the physical properties of the built-in sensing element.

[0058] The microcapsule fluorescence pathway depends on the microcapsule rupture caused by stress concentration, which allows the fluorescent material to come into contact with the quencher, resulting in a quantitative decay of fluorescence intensity.

[0059] The conductive nanonetwork pathways are dependent on the disruption of conductive pathways caused by strain, resulting in a jump in the thin film resistance.

[0060] The principle of local resolution at the structural level is as follows:

[0061] The integrated sensing bipolar plate arranges microelectrode arrays at key locations in the flow field. Each electrode unit constitutes an independent three-electrode test system, which can perform independent EIS measurements on the limited electrochemical reaction area around it.

[0062] By using multiplexing scanning technology, the system can sequentially activate sensing units 25 at different locations to obtain electrochemical impedance information of different spaces inside the stack, thereby constructing a spatial distribution map of the membrane health status.

[0063] The principle of multimodal data fusion is as follows:

[0064] The system performs correlation analysis between physical signals (the strain state of the membrane) and electrochemical signals (global and local impedance) in the time and space dimensions.

[0065] By establishing a strain-impedance coupling model, the system can identify early failure modes that have not yet caused significant electrochemical performance degradation but already pose a risk of mechanical damage.

[0066] Furthermore, the sensing material in the intelligent proton exchange membrane is a fluorescent microcapsule, which has urea-formaldehyde resin as the wall material and a core material containing a mixed solution of rhodamine B fluorescent dye and dimethylphenylpyrazolone quencher. The microcapsule has a particle size of 2±0.5μm and is uniformly dispersed in the proton exchange membrane resin at a ratio of 1.5-2.5wt%.

[0067] Furthermore, the sensing material in the intelligent proton exchange membrane is a sparse three-dimensional conductive network composed of silver nanowires, with a diameter of 50 nm, a length of 20-50 μm, and an areal density of 0.2-0.4 mg / cm².

[0068] Furthermore, the microelectrode sensing unit 25 on the integrated sensing bipolar plate is a three-electrode system, including a circular platinum microelectrode working electrode with a diameter of 500μm, a ring platinum counter electrode surrounding it, and a micro reversible hydrogen reference electrode; the sensing unit 25 is integrated on a metal or graphite bipolar plate using MEMS technology and is protected by a silicon nitride insulating layer.

[0069] Furthermore, the data acquisition module includes a multiplexed scanning subsystem, which is built based on an analog switch matrix, supports scanning of at least 128 sensing channels, has a full-channel scanning time of less than 30 seconds, and has a crosstalk suppression capability of greater than 60dB.

[0070] The early warning output module communicates with the fuel cell main controller via a CAN bus to send early warning information and location maintenance suggestions to the user interface, and / or to link with the active explosion suppression system to activate the protection program when an emergency risk is detected.

[0071] This invention fundamentally improves the early warning level of hydrogen-oxygen cross-contamination risks by endowing core components with self-sensing capabilities:

[0072] Achieving early warning of damage: A smart proton exchange membrane is provided that can issue an early warning signal that is ahead of traditional EIS technology by using its built-in sensing material to respond to mechanical stress before physical damage such as microcracks occurs in the membrane.

[0073] Achieving precise fault location: A bipolar plate with an integrated sensor array is provided, which can measure the electrochemical impedance spectroscopy of specific local areas inside the fuel cell stack. This allows for rapid location of specific single cells or flow field regions when an anomaly is detected, providing a basis for precise maintenance.

[0074] Achieve multimodal information fusion: Construct a fusion diagnostic system that can simultaneously process physical signals (such as optical and electrical signals) and electrochemical signals. Through multi-dimensional and complementary information, improve the reliability and accuracy of diagnostic results and significantly extend the early warning time.

[0075] Realizing the functionality and intelligence of core components: Upgrading traditional "passive" functional components (membranes, bipolar plates) into "active" sensing components, providing new ideas for fuel cell design, and promoting its development towards a safer and more intelligent direction.

[0076] Improve system maintenance efficiency and economy: By accurately locating faults, avoid "one-size-fits-all" overall maintenance, realize predictive maintenance and targeted maintenance, significantly reduce operation and maintenance costs and time, and improve the economic efficiency of fuel cell systems throughout their entire life cycle.

[0077] This invention, through material functionalization modification, endows traditional proton exchange membranes with self-sensing capabilities, forming a smart proton exchange membrane subsystem, namely a smart membrane. The smart membrane consists of two parts: a microcapsule fluorescent pathway and a conductive nanonetwork pathway. (See reference...) Figure 3 :

[0078] a. Microencapsule fluorescence pathway

[0079] Microcapsule preparation: Microcapsules with urea-formaldehyde resin as the wall material were prepared by in-situ polymerization. The core material 13 was a mixed solution of rhodamine B fluorescent dye and dimethylphenylpyrazolone quencher.

[0080] Membrane preparation process: Microcapsules are uniformly dispersed in a perfluorosulfonic acid resin solution at a ratio of 1.5-2.5 wt%, and a composite membrane is prepared by precision casting film formation process.

[0081] Signal detection: A miniature fiber optic probe is integrated on the end plate of the fuel cell stack and connected to a multi-channel fluorescence spectrometer.

[0082] The microcapsules have a particle size of 2±0.5μm, and the fluorescence detection sensitivity is able to detect strain changes of 0.05%. The linearity standard is: linear correlation R²>0.98 in the strain range of 0.1%-1.5%. In terms of the impact on membrane performance, the proton conductivity decreases by <3%, and the hydrogen permeability increases by <8%.

[0083] b. Conductive nanonetwork pathway

[0084] Nanowire selection and treatment: Silver nanowires with a diameter of 50 nm and a length of 20-50 μm were selected and surface silanization treatment was performed to improve dispersibility.

[0085] Network construction: A sparse three-dimensional conductive network was constructed within the membrane using a vacuum-assisted filtration method, with a nanowire surface density of 0.2-0.4 mg / cm².

[0086] Resistance monitoring: The in-plane resistance change of the membrane is monitored in real time using the four-probe method.

[0087] The initial sheet resistance is 10-50Ω, the strain sensitivity coefficient should be >20(ΔR / R0) / ε, and the durability standard is: the resistance change is <15% after 10,000 strain cycles. Regarding the impact on membrane performance, the proton conductivity decreases by <5%.

[0088] By integrating a sensor array inside a bipolar plate using microfabrication technology, an integrated sensing bipolar plate subsystem is formed, as referenced. Figure 4 The integrated sensing bipolar plate subsystem consists of sensing unit 25 and multiplexed measurement system, enabling electrochemical diagnosis of local areas.

[0089] The electrode structure of the sensing unit 25 adopts a three-electrode system. The working electrode is a circular platinum microelectrode with a diameter of 500 μm, the counter electrode is a ring platinum electrode (inner diameter 600 μm, outer diameter 800 μm) surrounding the working electrode, and the reference electrode is a miniature reversible hydrogen electrode located on one side.

[0090] In terms of manufacturing process, MEMS technology is used to manufacture electrodes on 316L stainless steel bipolar plates through photolithography, sputtering, and electroplating processes; for graphite plates, micro-pits are processed by laser and then filled with platinum-carbon paste.

[0091] A 1μm silicon nitride layer is deposited by PECVD outside the electrode area to achieve insulation protection.

[0092] Spatial resolution: The monitoring area of ​​a single sensing unit is approximately 0.8 cm².

[0093] Electrode stability: Polarization potential drift < 5mV after 1000 hours of continuous operation under fuel cell conditions.

[0094] Compatibility with flow field: The height of the sensor unit 25 protrusion is < 10μm, and its impact on airflow distribution is negligible.

[0095] The hardware architecture of the multiplexed measurement system is based on an analog switch matrix (such as ADI's ADG1414) to build a 128-channel scanning system, which shares the excitation source and acquisition card with the main EIS hardware; the control strategy adopts a tree-like scanning strategy, which prioritizes scanning high-risk areas and optimizes scanning efficiency; for signal integrity, shielded twisted-pair cables are used to transmit weak signals, and an instrumentation amplifier is integrated at the front end.

[0096] The performance parameters of the multiplexed measurement system are as follows:

[0097] Number of channels: Supports up to 128 sensing units; Scanning speed: Full-channel scan time < 30 seconds; Crosstalk suppression: >60dB @ 1kHz; Measurement accuracy: Phase angle measurement error < 0.1°.

[0098] The core of this system is a diagnostic algorithm based on multi-source data fusion, which achieves accurate early warning through spatiotemporal feature extraction and intelligent modeling.

[0099] The first step is multi-source signal preprocessing and feature extraction, detailed in the following steps:

[0100] a) Fluorescence signal processing

[0101] Baseline correction and noise filtering were performed on the original fluorescence spectrum.

[0102] Extract the fluorescence intensity at the characteristic wavelength and calculate the attenuation rate relative to the initial value. .

[0103] Establish Quantitative relationship model with membrane strain ε: (k is the calibration coefficient).

[0104] b) Nanoscale network resistor signal processing

[0105] The four-probe method is used to eliminate the influence of contact resistance.

[0106] Calculate the relative rate of change of resistance High-frequency noise is eliminated by applying a moving average filter.

[0107] when When the threshold is exceeded (e.g., 5%), it is considered a microscopic damage warning.

[0108] c) Local EIS data processing

[0109] DRT analysis was performed on the EIS data of each sensing unit 25 to extract the characteristic relaxation time distribution.

[0110] Calculate the high-frequency impedance of each unit. This reflects the proton conduction resistance of the local membrane.

[0111] Construct a spatial distribution cloud map of membrane health status.

[0112] Secondly, there is an intelligent diagnostic model that integrates spatiotemporal features. It employs a graph convolutional neural network (GCN) model based on an attention mechanism, which is specifically designed to handle multi-source data with spatial correlation.

[0113] ① Input features

[0114] Node characteristics: [fluorescence intensity, rate of change of resistance, local high-frequency impedance, temperature, humidity] for each sensing location.

[0115] Graph structure: A graph network is constructed based on the bipolar plate flow field layout, and the edge weights reflect the physical and fluid connectivity between locations.

[0116] ② Model Architecture

[0117] Spatial feature extraction layer: 2-layer GCN, learning the spatial dependencies of sensor data at different locations.

[0118] Temporal feature extraction layer: Bi-LSTM layer, which captures the changing trends of each feature over time.

[0119] Attention fusion layer: Calculates the contribution weights of different data sources to the final diagnostic results to achieve adaptive fusion.

[0120] Output layer: Outputs the risk probability [0,1] for each location and the overall health score of the fuel cell stack.

[0121] ③ Training strategies

[0122] Supervised learning is performed using historical operational data and corresponding actual membrane failure labels.

[0123] We use transfer learning to pre-train on laboratory data and then fine-tune it on real-world operational data.

[0124] Corresponding to the aforementioned hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate, this invention also discloses a hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate. The following details the hydrogen-oxygen crosstalk early warning method disclosed in this invention, in conjunction with the aforementioned hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate.

[0125] This invention discloses a hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate, comprising:

[0126] S1: Initialization and Self-Test

[0127] When the fuel cell starts up, it performs a self-test to confirm that all sensing channels are working properly.

[0128] Read the initial parameters of the smart membrane and sensing bipolar plate to establish baseline reference values.

[0129] S2: Parallel Data Acquisition

[0130] The global EIS system performs full-pile impedance measurements at set intervals (e.g., every 10 minutes).

[0131] The fluorescence detection system continuously monitors fluorescence intensity at a sampling frequency of 1 Hz.

[0132] The nano-network resistance monitoring system continuously records resistance changes at a sampling frequency of 10 Hz.

[0133] The local EIS scanning system collects spatial distribution data according to a preset strategy (such as a full scan every hour).

[0134] S3: Multi-source signal processing and feature extraction

[0135] Preprocess each signal and extract characteristic parameters.

[0136] When any feature parameter exceeds the threshold, a deeper analysis process is triggered.

[0137] S4: Integrated Diagnostics and Risk Assessment

[0138] Multi-source features are input into the intelligent diagnostic model to calculate the risk probability at each location.

[0139] Generate a spatial distribution heatmap of the fuel cell stack's health status.

[0140] High-risk areas are identified and tracked.

[0141] S5: Early Warning and Decision Support

[0142] Based on the risk assessment results, different levels of early warning information are output.

[0143] For the identified faulty areas, provide targeted maintenance recommendations.

[0144] It works in conjunction with the active explosion suppression system to activate protective procedures when an emergency risk is detected.

[0145] This invention achieves original sensing innovation at the material level: it creates a "smart proton exchange membrane" with self-sensing capabilities. By implanting fluorescent microcapsules or conductive nanonetworks, it transforms invisible mechanical strain into measurable optical or electrical signals, enabling the earliest warning of membrane damage, with the warning time being several orders of magnitude earlier than traditional electrochemical methods.

[0146] At the structural level, distributed diagnostic innovation has been achieved: the invention of "bipolar plate with integrated sensor array" has been developed, which extends EIS measurement capabilities from the entire fuel cell stack to local areas, enabling centimeter-level precise location of faults and solving the industry problem that traditional monitoring technologies cannot spatially locate faults.

[0147] In terms of multimodal data fusion, a new diagnostic paradigm is proposed that deeply integrates physical signals (optical fluorescence, resistance changes) with electrochemical signals (global / local impedance). Through a graph neural network model based on attention mechanism, adaptive weighted fusion of different data sources is achieved, and a higher-dimensional and more reliable health status assessment system is constructed.

[0148] For non-invasive integrated engineering, both innovative solutions are designed to be compatible with existing fuel cell manufacturing processes and system architectures. The preparation of the smart membrane can be integrated into existing casting processes, and the manufacturing of the sensing bipolar plate adopts standard MEMS processes. With minimal modifications, maximum functional improvement is achieved, which has high engineering feasibility and industrialization prospects.

[0149] Achieving predictive maintenance: Through precise fault location and risk assessment, the maintenance strategy of fuel cell systems can be transformed from "periodic maintenance" or "post-failure maintenance" to "predictive targeted maintenance", which significantly improves system reliability and reduces the total life cycle operation and maintenance cost.

[0150] Example 1

[0151] This embodiment uses a commercial fuel cell city bus as an application scenario to illustrate the implementation of this system.

[0152] The intelligent proton exchange membrane employs a conductive nanonetwork path. The membrane thickness is 15 μm, with a silver nanowire areal density of 0.3 mg / cm². This membrane directly replaces the original vehicle's standard proton exchange membrane and is assembled into the fuel cell stack. A four-probe resistance monitoring module is installed at the endplate of the fuel cell stack, connected to the collectors on both sides of the membrane via extremely fine silver wires, for real-time monitoring of changes in the membrane's in-plane sheet resistance.

[0153] In the 5th, 10th, and 15th cells of the fuel cell stack, bipolar plates with integrated microelectrode sensor arrays are used. Each bipolar plate has two sensing units 25 arranged in its air inlet region 22, middle region 23, and air outlet region 24, for a total of six units per plate. The sensing units 25 are connected to a multiplexed acquisition box installed on the side of the fuel cell stack via wires led out from the microchannels within the plate.

[0154] A multi-channel fluorescence spectrometer and a local EIS scanning module are integrated into the expansion slot of the vehicle's fuel cell controller (FCU). The fusion diagnostic algorithm runs as embedded code on the FCU's real-time operating system.

[0155] The work process is as follows:

[0156] While the vehicle is in motion, the system performs monitoring in parallel:

[0157] The intelligent film resistance monitoring system samples continuously at a frequency of 10 Hz.

[0158] The local EIS scanning system performs a complete scan of all 18 sensing units 25 every 30 minutes.

[0159] The global EIS system performs a full-stack measurement every 5 minutes.

[0160] During a certain operation, the system detected a sudden change in the smart film resistance at position 24 of the venting region of the 10th single cell. Meanwhile, local EIS data at this location showed a 15% increase in high-frequency impedance. Based on these multi-source signals, the fusion diagnostic model determined the probability of membrane damage at this location to be 87%.

[0161] The system immediately sent a warning message to the driver via the CAN bus: "Risk of membrane damage detected in the 10th battery cell; reduced power operation and scheduled inspection recommended," and recorded precise fault location information. Maintenance personnel then conducted targeted inspections and adjustments to the 10th battery cell, avoiding potential membrane rupture and hydrogen-oxygen cross-contamination risks.

[0162] Example 2

[0163] This embodiment uses a 1MW containerized hydrogen energy storage power station as an application scenario to illustrate the application of this system in large-scale stationary installations.

[0164] This embodiment preferably employs a smart proton exchange membrane with a microcapsule fluorescence pathway to meet the power plant's requirement for extremely high sensitivity in early warning. Fiber optic probe arrays are integrated and installed at both end plates of each fuel cell module for exciting and collecting fluorescence signals.

[0165] The power plant comprises four 250kW fuel cell stack modules. All bipolar plates in each module feature an integrated sensing design, with eight sensing units per bipolar plate, resulting in approximately 1200 monitoring points across the entire system. An industrial-grade fluorescence spectrometer and a high-performance local EIS scanning system are deployed in the power plant's central control room, connected to each fuel cell stack via a fiber optic network and dedicated shielded cables.

[0166] A cloud-based integrated diagnostic platform is deployed on the power plant server, employing a deeper neural network model to process massive amounts of spatiotemporal monitoring data.

[0167] The work process is as follows:

[0168] During power plant operation, the system implements a tiered monitoring strategy to optimize resources:

[0169] Level 1 monitoring: Continuously monitor the fluorescence intensity of all smart membranes at a sampling frequency of 1 Hz.

[0170] Secondary monitoring: 10% of the sensor units 25 in the system are sampled and scanned every hour.

[0171] Level 3 monitoring: A complete scan of the entire system (1200 monitoring points) is conducted every 6 hours.

[0172] Through long-term data trend analysis, the system found that the fluorescence intensity at a certain location on the upper side of the air inlet area 22 of the #2 fuel cell stack module slowly decreased from 100% to 92% over 30 days. Although the local EIS parameter change at this location did not exceed 3% during the same period, the fusion diagnostic model predicted a high risk at this location within 400-600 operating hours based on the spatiotemporal trend of fluorescence decay.

[0173] Based on this, the system generates early warnings and schedules maintenance windows in advance, performing preventative replacements before visible physical damage occurs at that location, successfully avoiding an unplanned downtime, and accumulating valuable early damage data for model iteration and optimization.

[0174] Preferred parameter configuration:

[0175] The system's early warning thresholds can be configured as follows:

[0176] Fluorescence intensity decay: (Attention level), >8% (Warning level), >10% (Danger level).

[0177] Nanonetwork resistance variation: (Attention level), >5% (Warning level), >8% (Danger level).

[0178] Local high-frequency impedance changes: >10% (Caution Level), >20% (Warning Level), >30% (Danger Level).

[0179] The system saves all raw data and characteristic data, and establishes a digital health record for the entire life cycle of the fuel cell stack.

[0180] Compared with the prior art, the present invention has the following advantages:

[0181] 1) Early warning time: Through the direct sensing of mechanical stress by the smart membrane, the earliest warning of membrane damage is achieved. The warning time is tens to hundreds of hours earlier than that of traditional electrochemical monitoring methods, providing a sufficient time window for taking protective measures.

[0182] 2) Order-of-magnitude improvement in fault location accuracy: Through distributed measurement of sensing bipolar plates, the fault location accuracy is improved from "pile level" to "centimeter level", realizing precise spatial positioning, making predictive maintenance and targeted repair possible, and significantly reducing maintenance costs.

[0183] 3) Significantly enhanced diagnostic reliability: Through multimodal data fusion, the limitations of single signal monitoring are overcome, effectively distinguishing between normal operating condition fluctuations and real fault precursors, reducing the false alarm rate to below 5%, and significantly improving the detection rate of real faults.

[0184] 4) Intelligent upgrade of maintenance strategy: Based on accurate positioning and early warning, the maintenance of fuel cell system is transformed from "periodic maintenance" or "post-failure maintenance" to "condition-based predictive maintenance", which optimizes the allocation of maintenance resources and improves system availability and life cycle economy.

[0185] 5) Expansion of core component functions: Successfully transformed traditional "passive" functional components into "active" sensing components, providing a new technical path for the intelligent design of fuel cells and promoting technological progress throughout the industry.

[0186] 6) Improved safety protection system: It forms a technical synergy with the online EIS diagnostic system and active explosion suppression system, and builds a complete safety technology system from "early perception" to "precise diagnosis" and then to "active protection", providing a solid safety guarantee for the large-scale commercial application of fuel cells.

[0187] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for early warning of hydrogen-oxygen crosstalk based on a smart membrane and a sensing bipolar plate, characterized in that, The method includes: Data acquisition steps: Physical sensing signals from the smart proton exchange membrane of the fuel cell and local electrochemical impedance spectroscopy data from the integrated sensing bipolar plate are acquired in parallel. Feature extraction step: The physical sensing signal and local electrochemical impedance spectroscopy data are processed to extract feature parameters characterizing the mechanical strain state of the membrane and electrochemical feature parameters characterizing the local health state of the membrane, respectively. Fusion diagnostic step: The extracted feature parameters are input into a pre-trained multimodal fusion diagnostic model. The model analyzes the correlation between physical sensing signals and electrochemical signals in the spatiotemporal dimension and outputs the probability and location information of the risk of hydrogen-oxygen crosstalk inside the fuel cell. Early warning output steps: Based on the risk probability and location information output from the fusion diagnosis steps, generate and issue early warning information of different levels.

2. The hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate as described in claim 1, characterized in that, The physical sensing signal is an optical signal or an electrical signal; wherein... The optical signal is generated by the change in fluorescence intensity caused by the strain-induced rupture of fluorescent microcapsules embedded in the smart proton exchange membrane. The electrical signal is generated by the change in resistance caused by strain in the conductive nanonetwork constructed within the smart proton exchange membrane.

3. The hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate as described in claim 1, characterized in that, The local electrochemical impedance spectroscopy data are obtained by sequentially activating different microelectrode sensing units on the integrated sensing bipolar plate through a multiplexed scanning system, thereby constructing a spatial distribution map of the membrane health state.

4. The hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate as described in claim 1, characterized in that, The pre-trained multimodal fusion diagnostic model is a graph convolutional neural network model based on an attention mechanism; wherein, The input to the model includes feature vectors with sensing units as nodes, and a graph structure constructed based on the bipolar plate flow field layout. The model learns spatial dependencies through graph convolutional layers, learns temporal variation trends through recurrent neural network layers, and adaptively fuses feature parameters from different sources through an attention mechanism.

5. The hydrogen-oxygen crosstalk early warning method based on a smart membrane and a sensing bipolar plate as described in claim 1, characterized in that, Before the data acquisition step, there are also initialization and self-test steps, which specifically include: After starting the fuel cell and performing a self-test to confirm that all sensing channels are working properly, the initial parameters of the smart membrane and sensing bipolar plate are read to establish baseline reference values.

6. A hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate, characterized in that, The system includes: A smart proton exchange membrane, which has embedded sensing materials to convert the membrane’s mechanical strain into a measurable physical signal; An integrated sensing bipolar plate integrates a microelectrode sensing array in its flow field region for independent electrochemical impedance spectroscopy measurements of its local area. A data acquisition module is configured to communicate with the smart proton exchange membrane and the integrated sensing bipolar plate to acquire the physical signal and local electrochemical impedance spectroscopy data in parallel. The feature extraction module is used to process the physical sensing signal and the local electrochemical impedance spectroscopy data to extract feature parameters characterizing the mechanical strain state of the membrane and electrochemical feature parameters characterizing the local health state of the membrane, respectively. The fusion diagnostic module is used to input the extracted feature parameters into a pre-trained multimodal fusion diagnostic model. The model analyzes the correlation between physical sensing signals and electrochemical signals in the spatiotemporal dimension and outputs the probability and location information of hydrogen-oxygen crosstalk risk inside the fuel cell. The early warning output module is used to generate and issue early warning information based on the output results of the data processing and diagnosis module.

7. A hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate as described in claim 6, characterized in that, The sensing material in the intelligent proton exchange membrane is a fluorescent microcapsule with urea-formaldehyde resin as the wall material and a core material containing a mixed solution of rhodamine B fluorescent dye and dimethylphenylpyrazolone quencher. The microcapsule has a particle size of 2±0.5μm and is uniformly dispersed in the proton exchange membrane resin at a ratio of 1.5-2.5wt%.

8. A hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate as described in claim 6, characterized in that, The sensing material in the intelligent proton exchange membrane is a sparse three-dimensional conductive network composed of silver nanowires with a diameter of 50 nm, a length of 20-50 μm, and an areal density of 0.2-0.4 mg / cm².

9. A hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate as described in claim 6, characterized in that, The microelectrode sensing unit on the integrated sensing bipolar plate is a three-electrode system, including a circular platinum microelectrode working electrode with a diameter of 500μm, a ring platinum counter electrode surrounding it, and a micro reversible hydrogen reference electrode; the sensing unit is integrated on a metal or graphite bipolar plate using MEMS technology and protected by a silicon nitride insulating layer.

10. A hydrogen-oxygen crosstalk early warning system based on a smart membrane and a sensing bipolar plate as described in claim 6, characterized in that, The data acquisition module includes a multiplexed scanning subsystem, which is built on an analog switch matrix, supports scanning of at least 128 sensor channels, has a full-channel scanning time of less than 30 seconds, and has a crosstalk suppression capability of more than 60dB. The early warning output module communicates with the fuel cell main controller via a CAN bus to send early warning information and location maintenance suggestions to the user interface, and / or to link with the active explosion suppression system to activate the protection program when an emergency risk is detected.

Images