Hydrogen leakage monitoring system for hydrogen-oxygen fuel cell

Abstract

The invention discloses a hydrogen leakage monitoring system for a hydrogen-oxygen fuel cell, and particularly relates to the technical field of cell hydrogen leakage monitoring, and the hydrogen leakage monitoring system comprises an electrochemical impedance early warning module, a multi-source fusion anti-interference module and an active positioning and confirming module which are connected in sequence. The electrochemical impedance early warning module monitors the change of the impedance spectrum characteristic of the galvanic pile and starts a multi-stage confirmation process to realize early warning of the leakage risk; the multi-source fusion anti-interference module generates a high-reliability leakage alarm by fusing multi-sensor data and analyzing a leakage diffusion space-time mode; and after receiving the early warning, the active positioning and confirming module fuses acoustic imaging and airflow measurement information, and outputs position estimation and confidence evaluation of a leakage source based on Bayesian reasoning. According to the invention, the whole-process monitoring from early warning, reliable alarm and accurate positioning is realized, and the timeliness, accuracy and reliability of monitoring are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of battery hydrogen leakage monitoring technology, and more specifically, to a hydrogen leakage monitoring system for hydrogen-oxygen fuel cells. Background Technology

[0002] Hydrogen-oxygen fuel cells, as efficient and clean energy conversion devices, have shown broad application prospects in transportation, distributed power generation, and other fields. However, their anode side requires a continuous supply of high-pressure hydrogen. Hydrogen has extremely low ignition energy and a wide explosion limit; even a small leak accumulating in a closed or semi-closed space can lead to serious safety accidents such as combustion or even explosion. Therefore, developing a highly reliable and sensitive hydrogen leak monitoring system is a key prerequisite for ensuring the safe operation of fuel cell systems and promoting their large-scale commercial application.

[0003] Currently, the most commonly used hydrogen leak monitoring solutions in the industry mainly rely on discrete hydrogen concentration sensors placed at key points in the system (such as hydrogen storage tank valves, pipeline joints, and fuel cell stack cavities). This passive monitoring method, based on concentration threshold triggering, has significant limitations: First, its response is severely delayed; the sensor only alarms when the leaked hydrogen diffuses and accumulates near its sensitive element and the concentration exceeds a preset threshold, failing to provide early warning for slow leaks or intermittent minor leaks. Second, existing electrochemical hydrogen sensors are susceptible to cross-interference from volatile organic compounds such as alcohol and carbon monoxide in the environment, as well as changes in temperature and humidity, resulting in a high false alarm rate that severely impacts system reliability and user experience. Furthermore, even with multiple sensor alarms, it only indicates the presence of a leak, making it difficult to quickly and accurately locate the leak source, greatly hindering troubleshooting and repair. In addition, some research has attempted to use acoustic or pressure sensors for auxiliary detection, but these methods are mostly supplementary and lack intelligent fusion and collaborative verification mechanisms based on multi-dimensional information.

[0004] In summary, existing technologies struggle to detect risks in the very early stages of a leak and are insufficient in handling complex operating conditions and accurately locating the leak source. Therefore, there is an urgent need for an intelligent hydrogen leak monitoring system capable of early warning, strong anti-interference capabilities, and accurate location assistance to systematically improve the safety level of hydrogen-oxygen fuel cell operation. Summary of the Invention

[0005] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a hydrogen leakage monitoring system for hydrogen-oxygen fuel cells.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A hydrogen leakage monitoring system for hydrogen-oxygen fuel cells includes the following modules: An electrochemical impedance spectroscopy (EIS) early warning module is used to generate early leakage risk warning signals with confidence assessments based on EIS analysis. The multi-source fusion anti-interference module is used to fuse information from multiple sensors and analyze leakage propagation patterns to generate highly reliable leakage probability alarms. The active location and confirmation module is used to initiate active detection and fuse acoustic and airflow information after receiving early warnings and alarms, and output the location estimate and confidence assessment of the leak source.

[0007] Specifically, the electrochemical impedance early warning module performs the following steps: Periodically collect EIS data of the fuel cell stack and extract impedance characteristic values; Establish and update a dynamic baseline model for EIS features; The real-time feature values ​​are compared with the baseline model. When the feature values ​​continue to deviate and exceed the preset threshold, a multi-level confirmation process including rapid response, pattern verification and cross-validation is initiated. Based on the confirmation results at all levels, a Level 1 leakage risk warning signal with a confidence level is generated.

[0008] Specifically, the multi-source fusion anti-interference module performs the following steps: Data were collected from hydrogen concentration sensor arrays, pressure sensors, and VOC sensors located at different positions; The hydrogen concentration readings are weighted and fused, and the alarm threshold is dynamically adjusted based on ambient temperature, VOC readings, and pressure change rate. Leakage pattern identification is performed on the spatiotemporal reading distribution of the sensor array, and spatial gradient consistency, temporal evolution monotonicity, and correlation are calculated. A secondary leakage probability alarm is generated only when the fusion concentration exceeds the dynamic threshold, the pressure changes abnormally, and the leakage mode matches well.

[0009] Specifically, the active positioning and confirmation module performs the following steps: It is activated when both a Level 1 warning and a Level 2 alert are received simultaneously. The ultrasonic microphone array is controlled to scan and generate an acoustic imaging map to identify acoustic hotspots; at the same time, an airflow sensor is activated to measure the local flow velocity vector. Based on the Bayesian inference framework, the prior probability model of acoustic localization and the likelihood model of airflow observation are integrated to calculate the posterior probability distribution of the leak source location. Output the optimal estimate of the leak source location and the confidence region characterizing the location uncertainty.

[0010] Specifically, in the electrochemical impedance early warning module: During the cross-validation period in the multi-level confirmation process, the comprehensive deviation index is calculated by calling the data from the pressure and temperature auxiliary sensors, and combined with the consistency assessment score of the previous confirmation results, to jointly decide whether to generate a final warning.

[0011] Specifically, the leakage pattern identification includes: Construct the spatiotemporal matrix of the hydrogen concentration field; Calculation of concentration gradient field and average directional consistency index based on spatial distribution of sensor network; Analyze the monotonicity of concentration readings from each sensor and their temporal correlation with neighboring sensors; Calculate the leakage pattern matching degree by combining spatiotemporal characteristics.

[0012] Specifically, in the Bayesian inference framework: The prior probability model for acoustic localization is composed of a weighted mixture of Gaussian distribution models of multiple acoustic hotspots according to their signal strength; The likelihood model for airflow observation is based on the point source airflow field assumption and is constructed by comparing the degree of matching between the measured airflow vector and the predicted vector of the theoretical model.

[0013] Specifically, in the active positioning and confirmation module: After localization is completed, the information entropy of the posterior probability distribution is calculated to evaluate the quality of fused localization; if the information entropy is higher than the threshold, the localization reliability is marked as low in the report.

[0014] The system is also configured as follows: Upon confirmation of a leak, a safety response is triggered, including activating emergency ventilation and shutting off the main hydrogen valve, and a snapshot of the leak event, location results, and sensor data is stored. The positioning results and confidence regions output by the active positioning and confirmation module are highlighted in the system visualization interface.

[0015] The technical effects and advantages of this invention are as follows: This invention fundamentally changes the passive response mode of traditional hydrogen leak monitoring. Through a dual mechanism of electrochemical impedance spectroscopy (EIS) early warning and multi-source fusion interference suppression, it achieves early and reliable perception of leak risks. Utilizing EIS technology to indirectly monitor the internal state of the fuel cell stack, it can issue early warnings tens of seconds to minutes before a significant increase in hydrogen concentration, gaining valuable time for a safe response. Simultaneously, by fusing information from multiple sensors such as hydrogen, pressure, and VOCs and introducing an intelligent recognition algorithm based on spatiotemporal diffusion patterns, the system can effectively eliminate environmental interference and single-point sensor failures, reducing the false alarm rate by more than an order of magnitude. This solves the two core pain points of existing technologies: slow response and high false alarm rate.

[0016] The present invention constructs a complete safety decision-making closed loop from risk warning to precise positioning, significantly improving operation and maintenance safety and efficiency. After the system confirms a leak, it automatically activates the active detection module, which fuses acoustic imaging and airflow measurement data within the Bayesian probability framework. It can not only output the coordinates of the leak source with centimeter-level accuracy but also provide a quantitative evaluation of the positioning confidence area. This enables operation and maintenance personnel to quickly and accurately locate the leak point, greatly shortening the troubleshooting time. At the same time, the linked safety responses triggered by the system (such as emergency valve closing and ventilation) also achieve automation from risk perception to active protection, systematically enhancing the safety management level of fuel cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0019] As Figure 1 shown, the hydrogen leakage monitoring system module for a hydrogen-oxygen fuel cell is as follows: The electrochemical impedance warning module periodically collects the electrochemical impedance spectra of the fuel cell stack and extracts characteristic values such as the high-frequency resistance and the diameter of the low-frequency inductive reactance arc. By comparing the real-time characteristic values with the dynamic baseline model and introducing a three-level confirmation process of "quick response, pattern verification, and cross-verification", comprehensively evaluating the persistence of the abnormal pattern, and finally generating a first-level leakage risk warning signal with a confidence level to achieve early risk warnings much earlier than the concentration exceeds the standard; the process is as follows: When the fuel cell stack is operating normally and stably, control the EIS analysis unit integrated in the stack manager to apply a low-amplitude, multi-frequency alternating current perturbation signal at a fixed time interval (such as every minute) and synchronously collect the voltage and current responses of the stack.

[0020] Calculate the impedance characteristic values of the stack in specific high-frequency regions (for example, 1 kHz - 10 kHz) and low-frequency regions (for example, 0.1 Hz - 10 Hz) in real time, preferably the high-frequency resistance and the diameter of the low-frequency inductive reactance arc.

[0021] Establish and continuously update a dynamic baseline model of the EIS characteristics of the normal operating condition based on recent historical data. It includes the fluctuation range and short-term change trend of the above characteristic values in the normal leak-free state.

[0022] The real-time calculated EIS characteristic values ​​are compared with the dynamic baseline model. When an abnormal decrease in high-frequency resistance not caused by temperature is detected (suggesting an abnormal increase in membrane hydration, possibly due to wet hydrogen leakage), or a specific pattern of drift in the low-frequency inductive arc diameter (suggesting a change in anolyte gas mass transfer conditions), and this continues to exceed a preset statistical deviation threshold, a multi-cycle progressive confirmation mechanism is introduced to assess the warning confidence level; when the high-frequency resistance... Or low frequency inductance arc diameter This is the first time the confidence interval set by the dynamic baseline model has been exceeded. (in The mean of the eigenvalues, Standard deviation, When the preset confidence coefficient (between 2.5 and 3.0) is used, the three-level confirmation process is initiated: Level 1 Confirmation (Rapid Response Phase): Immediately increase the EIS monitoring frequency from the normal interval. (e.g., 60 seconds) Increase to a denser monitoring interval (e.g., 5 seconds), duration Rapid data acquisition (e.g., 30 seconds). During this period, the instantaneous gradient of the feature values ​​is calculated. :

[0023] in The feature value representing the current time ( or ), The sampling interval is... The previous sampling time The eigenvalues ​​of . If in Within, the proportion exceeding the threshold (e.g., 70%) of the sampling points satisfy And the gradient of change If the values ​​remain in the same direction (i.e., continuously positive or negative), then proceed to the second level of confirmation.

[0024] Level 2 Confirmation (Pattern Validation Period): Maintain a high monitoring frequency. (e.g., 10 seconds), duration (e.g., 90 seconds). During this period, not only is the absolute deviation of the eigenvalues ​​monitored, but their frequency domain characteristics are also analyzed. The periodic components of the eigenvalue sequence are analyzed using short-time Fourier transform to calculate the energy concentration index of the potential leakage signal. :

[0025] in For frequency variables, The differential element is the frequency. The power spectral density of the eigenvalue sequence, and This is a preset leakage characteristic frequency range (e.g., 0.01–0.1 Hz, corresponding to slow leakage changes). If If the threshold (e.g., 0.6) is reached and the feature value continues to deviate from the baseline, then the process proceeds to the third level of confirmation.

[0026] Level 3 Validation (Cross-validation Period): While maintaining EIS monitoring (intervals to...) Cross-validation is performed using auxiliary sensor data. The rate of change of pressure sensor readings over adjacent time periods is read. Calculate the overall deviation index based on the temperature sensor reading T. :

[0027]

[0028] Where X is the EIS feature value at the current time. The historical standard deviation of the rate of change of pressure. The historical standard deviation of temperature readings. and This is the baseline value under normal operating conditions. and For the corresponding standard deviation, , , Weighting coefficients (satisfying) ,and ).like (Critical value, such as 3.5) and the consistency assessment score for the three judgment periods. (Calculated based on decision fusion algorithm) Exceeds the threshold (e.g., 0.75), ultimately generating a high-confidence Level 1 leakage risk warning signal.

[0029] The signal includes a confidence level. The calculation formula is:

[0030] in This represents the confidence level of the Level 1 leakage risk warning signal, with a value range of [0,1]. 0.3, 0.4, and 0.3 are fixed weighting coefficients for the three confirmation stages, respectively. , , These are the pass indicators for the three confirmation periods (1 for pass, 0 otherwise). Confidence level. The warning signal will be output along with the alert signal for subsequent modules to make weighted decisions. If any level fails to confirm the error, only the abnormal event log will be recorded without issuing an alert, effectively avoiding false alarms caused by momentary interference.

[0031] The multi-source fusion anti-interference module first collects data from multiple sensors, including hydrogen concentration, pressure, and VOCs. It then performs weighted fusion of hydrogen concentration data and dynamically adjusts the alarm threshold based on environmental interference. Subsequently, it initiates spatiotemporal correlation analysis, constructing a leakage pattern matching degree by calculating the spatial gradient consistency, temporal evolution monotonicity, and correlation of the concentration field. Only when the fused concentration exceeds the threshold, the pressure is abnormal, and the pattern matching degree is high, is a high-confidence secondary leakage probability alarm generated, effectively suppressing false alarms. The process is as follows: Multiple high-precision electrochemical hydrogen sensors are arranged in a sensor array near the anode inlet manifold, circulation loop, exhaust valve, and key locations in the fuel cell stack compartment. Pressure sensors and VOC sensors are simultaneously arranged in close proximity.

[0032] The system continuously collects concentration readings from various hydrogen sensors, pressure change rates from pressure sensors, VOC sensor readings, and cabin temperature data.

[0033] Run a weighted fusion decision algorithm. First, perform a consistency check on the readings of each hydrogen sensor and remove abnormal sensor data that deviates significantly from the cluster mean. Then, calculate a weighted average of the valid hydrogen concentration data, with the weights dynamically adjusted based on the historical reliability of each sensor.

[0034] The fused hydrogen concentration value is compared with an adaptive alarm threshold. This adaptive threshold is not a fixed value, but is dynamically adjusted based on ambient temperature, VOC sensor readings (used to identify interference sources such as alcohol), and pressure change rate. For example, when the VOC reading increases, the system automatically and transiently raises the hydrogen alarm threshold to suppress false alarms caused by cross-interference; when a sudden drop in local pressure is detected, the alarm threshold is appropriately lowered to improve sensitivity.

[0035] Restart spatiotemporal correlation analysis and leakage pattern identification; perform the following steps: Constructing the concentration field spatiotemporal matrix: Read the current time t and the previous k sampling times (time windows) All N hydrogen sensors in three-dimensional space coordinates The readings at each location form a spatiotemporal data matrix. Its elements Indicates that the i-th sensor is in Concentration value at time ( , (Sampling interval).

[0036] Calculating spatial gradient and diffusion consistency: based on the current time ( The sensor readings distribution is used to estimate the spatial gradient vector field of the hydrogen concentration field. For each sensor location, the gradient is approximately:

[0037] in: Let be the spatial gradient vector of hydrogen concentration estimated at sensor i, where i is the index of the previous sensor and j is the set of neighboring sensors of sensor i. Sensor index in It is the set of neighboring sensors of sensor i. It is the position vector pointing from sensor j to i. The weight is inversely proportional to the distance between sensors i and j. Let j be the hydrogen concentration reading at the current moment. Let be the hydrogen concentration reading of sensor i at the current moment. Let be the position vector pointing from sensor j to sensor i. For vectors The modulus (length) is the Euclidean distance from sensor j to i. Then, the average directional consistency index of the gradient vectors of all sensors is calculated. :

[0038] Where: N is the total number of hydrogen sensors. Let i be the gradient vector at sensor i. To prevent division by zero in extremely small quantities. A gradient close to 1 indicates that the gradient directions of all points are highly consistent, pointing to a possible common source (leakage point); a gradient close to 0 indicates that the directions are scattered, more like global interference or independent noise.

[0039] Analysis of temporal evolution characteristics: time series for each sensor i Perform the analysis. Calculate its monotonicity fraction. and the time-domain correlation coefficient with neighboring sensors .

[0040] Monotonicity fractions It is obtained by calculating the statistical significance of the "upward" and "downward" trends in the sequence.

[0041] Time-domain correlation coefficient Let be the average Pearson correlation coefficient between sensor i and the time series of readings of its m nearest spatial neighbors.

[0042] True leakage typically manifests as: sensors located close to the source point. The value is high (continuously rising) and close to the neighboring sensor. High value (co-variation).

[0043] Integrated Pattern Matching and Decision Making: Defining a Leakage Pattern Matching Degree Based on the above spatiotemporal characteristics:

[0044] in, and These are all sensors and The weighted average (the weights can be set based on the distance between the sensor and the initial fusion concentration center). , , These are the weight coefficients for spatial consistency, monotonicity, and correlation, respectively. Set an empirical threshold. .

[0045] When the fusion concentration exceeds the dynamic threshold and the pressure change rate is abnormal, and the leakage mode matching is... A high-confidence level 2 leak probability alarm is generated when all three conditions are met. If the first two conditions are met but... If this occurs, a low-confidence abnormal event alert will be generated, and detailed spatiotemporal data will be recorded for operation and maintenance analysis, without triggering an emergency response.

[0046] The active localization and confirmation module activates upon receiving two levels of warnings. First, it controls an ultrasonic microphone array to scan and generate an acoustic image to identify acoustic hotspots. Simultaneously, it activates an airflow sensor to measure local velocity vectors. Then, based on a Bayesian inference framework, it fuses the prior probability model of acoustic localization with the likelihood model of airflow observation to calculate the posterior probability distribution of the leak source location. Finally, it outputs the optimal location estimate with a confidence ellipsoid, achieving precise localization and confirmation of the leak source. The process is as follows: The system is automatically activated when both a Level 1 leakage risk warning and a Level 2 leakage probability alarm are received simultaneously. If only a Level 2 leakage probability alarm is received without a Level 1 leakage risk warning, it is considered a suspicious false alarm, and only the log is logged without initiating the active troubleshooting process.

[0047] After automatic activation, the system first controls the ultrasonic microphone array mounted on a rotatable platform atop the fuel cell stack to scan key suspected areas (such as the locations of corresponding alarm sensors). High-pressure hydrogen leaks generate ultrasonic noise in a specific frequency band.

[0048] An ultrasonic microphone array collects acoustic signals and generates a real-time acoustic image using a beamforming algorithm. This image is then overlaid on the device structure diagram in the form of an acoustic cloud map on the system's visualization interface, visually displaying the location of the sound source hotspot at suspected leak points.

[0049] At the same time, multiple miniature high-response airflow sensors (such as hot-film anemometers) deployed near the suspected area were activated to measure the direction of extremely low-velocity airflow that may be caused by the leak.

[0050] By integrating acoustic hotspot location information and multi-point airflow vector data, and through triangulation and airflow tracing algorithms, the system calculates and outputs the three-dimensional coordinate estimation of the leakage source with centimeter-level accuracy, and highlights it on the visualization interface.

[0051] To improve the accuracy and reliability of leak source localization, a multi-source information fusion method based on a probabilistic framework is introduced on the basis of triangulation and airflow tracing algorithms. This method constructs the posterior probability distribution of the leak source location and outputs the localization result with confidence assessment. The process is as follows: Modeling the uncertainty of acoustic localization: Based on the beamforming results of the ultrasonic microphone array, the detection was performed. Each acoustic hotspot. Corresponding to a spatial location estimate and the covariance matrix of positioning uncertainty These uncertainties mainly stem from background noise, array geometric accuracy, and signal processing errors. Based on this, a probability density function for acoustic localization is constructed. Represents any point in space The probability that it is a sound source:

[0052] in, For a point in space Let M be the probability density of the sound source (leakage source), M be the number of detected acoustic hotspots, and m be the index of the acoustic hotspot. For weighting coefficients, and the first Signal strength of each acoustic hotspot Proportional, that is ,in The signal strength of the nth acoustic hotspot; This represents a multivariate Gaussian distribution.

[0053] Airflow localization probabilistic model construction: deployment The velocity vector is measured by an airflow sensor. To deduce the leak source from the airflow data, we assume the leak source is located at... A point source airflow field is generated at the location. Based on potential flow theory or a lookup table pre-established through computational fluid dynamics (CFD) simulation, the sensor's performance is predicted. Theoretical airflow velocity at the location The likelihood function for airflow positioning is defined as follows:

[0054] in, To locate the leak source Under these conditions, the velocity vector set of the airflow sensor was observed. The likelihood function, Let Q be the assumed location (vector) of the leak source, Q be the number of airflow sensors, and q be the index of the airflow sensor, where q = 1, 2, ..., Q; Let q be the velocity vector measured by the airflow sensor. Assuming the leak source is located at... At that time, the theoretical airflow velocity (vector) at the q-th sensor is predicted by the model. It is an airflow sensor The standard deviation of the measured noise. This function measures the difference between the observed airflow pattern and the assumed leak source. The degree of matching between the airflow patterns that should be generated at that time.

[0055] Bayesian fusion localization combines prior information from acoustic localization with airflow observation data, and calculates the posterior probability distribution of the leak source location through Bayesian inference.

[0056] in, In order to use known acoustic observation data and airflow observation data Under these conditions, the leak source is located The posterior probability, The likelihood function for airflow localization. Let the location of the leak source be a vector, and the variable to be determined be... Let be the integral variable, representing the locations (vectors) of all possible leak sources. Let be a vector element, representing the integral over a position in three-dimensional space. and These represent acoustic and airflow observation data, respectively. The prior probability density provided for the acoustic localization model.

[0057] Location Results Output and Uncertainty Quantification: Final Leak Source Location Estimation Let the mean be the expected value of the posterior probability distribution:

[0058] in: This is the optimal estimate (vector) of the leak source location, which is the expected value (mean) of the posterior probability distribution. For mathematical expectation, This is the posterior probability distribution; Simultaneously, calculate the posterior covariance matrix. Quantifying uncertainty:

[0059] in: This is the posterior covariance matrix, used to quantify the uncertainty of the positioning results. The transpose of the vector converts the column vector into a row vector; the eigenvalues ​​and eigenvectors of this covariance matrix define the confidence ellipsoid of the localization result. In the visualization interface, the estimated location is highlighted. It also displays its 95% confidence area in the form of a semi-transparent ellipsoid, providing maintenance personnel with an intuitive assessment of positioning accuracy.

[0060] Information entropy-based fusion quality assessment: To assess the reliability of this fusion localization, the information entropy of the posterior distribution is calculated. :

[0061] in: It is the natural logarithm (base e). Information entropy is the information entropy of the posterior probability distribution, used to measure the uncertainty (dispersion) of the distribution. The smaller the value, the more concentrated the posterior distribution, and the higher the certainty of the localization result. Values ​​and thresholds In comparison, if If the location reliability is low, the report will indicate that manual verification or a longer observation period is required.

[0062] The confirmed leak event, location results, and all sensor data snapshots are stored together, and the highest level of system safety response is triggered (such as starting emergency ventilation, closing the main hydrogen valve, and degrading the system to shutdown).

[0063] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0064] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0065] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0066] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0068] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0069] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0070] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A hydrogen leakage monitoring system for hydrogen-oxygen fuel cells, characterized in that, Includes the following modules: An electrochemical impedance spectroscopy (EIS) early warning module is used to generate early leakage risk warning signals with confidence assessments based on EIS analysis. The multi-source fusion anti-interference module is used to fuse information from multiple sensors and analyze leakage propagation patterns to generate highly reliable leakage probability alarms. The active location and confirmation module is used to initiate active detection and fuse acoustic and airflow information after receiving early warnings and alarms, and output the location estimate and confidence assessment of the leak source.

2. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 1, characterized in that, The electrochemical impedance early warning module performs the following steps: Periodically collect EIS data of the fuel cell stack and extract impedance characteristic values; Establish and update a dynamic baseline model for EIS features; The real-time feature values ​​are compared with the baseline model. When the feature values ​​continue to deviate and exceed the preset threshold, a multi-level confirmation process including rapid response, pattern verification and cross-validation is initiated. Based on the confirmation results at all levels, a Level 1 leakage risk warning signal with a confidence level is generated.

3. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 1, characterized in that, The multi-source fusion anti-interference module performs the following steps: Data were collected from hydrogen concentration sensor arrays, pressure sensors, and VOC sensors located at different positions; The hydrogen concentration readings are weighted and fused, and the alarm threshold is dynamically adjusted based on ambient temperature, VOC readings, and pressure change rate. Leakage pattern identification is performed on the spatiotemporal reading distribution of the sensor array, and spatial gradient consistency, temporal evolution monotonicity, and correlation are calculated. A secondary leakage probability alarm is generated only when the fusion concentration exceeds the dynamic threshold, the pressure changes abnormally, and the leakage mode matches well.

4. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 1, characterized in that, The active positioning and confirmation module performs the following steps: It is activated when both a Level 1 warning and a Level 2 alert are received simultaneously. The ultrasonic microphone array is controlled to scan and generate an acoustic imaging map to identify acoustic hotspots; at the same time, an airflow sensor is activated to measure the local flow velocity vector. Based on the Bayesian inference framework, the prior probability model of acoustic localization and the likelihood model of airflow observation are integrated to calculate the posterior probability distribution of the leak source location. Output the optimal estimate of the leak source location and the confidence region characterizing the location uncertainty.

5. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 2, characterized in that, In the electrochemical impedance early warning module: During the cross-validation period in the multi-level confirmation process, the comprehensive deviation index is calculated by calling the data from the pressure and temperature auxiliary sensors, and combined with the consistency assessment score of the previous confirmation results, to jointly decide whether to generate a final warning.

6. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 3, characterized in that, The leakage pattern identification includes: Construct the spatiotemporal matrix of the hydrogen concentration field; Calculation of concentration gradient field and average directional consistency index based on spatial distribution of sensor network; Analyze the monotonicity of concentration readings from each sensor and their temporal correlation with neighboring sensors; Calculate the leakage pattern matching degree by combining spatiotemporal characteristics.

7. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 4, characterized in that, In the Bayesian inference framework: The prior probability model for acoustic localization is composed of a weighted mixture of Gaussian distribution models of multiple acoustic hotspots according to their signal strength; The likelihood model for airflow observation is based on the point source airflow field assumption and is constructed by comparing the degree of matching between the measured airflow vector and the predicted vector of the theoretical model.

8. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to claim 7, characterized in that, In the active positioning and confirmation module: After localization is completed, the information entropy of the posterior probability distribution is calculated to evaluate the quality of fused localization; if the information entropy is higher than the threshold, the localization reliability is marked as low in the report.

9. The hydrogen leakage monitoring system for hydrogen-oxygen fuel cells according to any one of claims 1-8, characterized in that, The system is also configured as follows: Upon confirmation of a leak, a safety response is triggered, including activating emergency ventilation and shutting off the main hydrogen valve, and a snapshot of the leak event, location results, and sensor data is stored. The positioning results and confidence regions output by the active positioning and confirmation module are highlighted in the system visualization interface.

Images