Safety failure prediction method for alkaline electrolytic cell of hydrogen energy system

Abstract

The invention discloses a safety failure prediction method for an alkaline electrolytic cell of a hydrogen energy system, relates to the technical field of electrolytic cell safety prediction, and solves the problems that in the prior art, tiny dynamic fluctuation of electrolyte concentration cannot be captured through hydrogen evolution abnormal monitoring, and correlation between concentration change and bubble dynamics is not established. Specifically, the safety fault prediction method comprehensively covers three core safety risks of abnormal hydrogen evolution, thermal fatigue cracks and insulation performance degradation, and early recognition, accurate positioning and graded early warning of safety faults are realized through data fusion of a multi-source sensor, accurate extraction of characteristic parameters, construction of a coupling correlation model and full-life-cycle monitoring. An operable intervention strategy is provided for an administrator, the safety fault occurrence rate is effectively reduced, the service life of equipment is prolonged, and stable and reliable operation of the hydrogen energy system is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of electrolyzer safety prediction technology, specifically a method for predicting safety faults in alkaline electrolyzers for hydrogen energy systems. Background Technology

[0002] In the current hydrogen energy industry, alkaline electrolyzers have become the core equipment for green hydrogen production due to their cost advantages and technological maturity, and are widely used in new energy hydrogen production scenarios. However, under the conditions of long-term stable operation and frequent start-stop operations caused by grid peak shaving, their safety fault early warning faces multiple technical bottlenecks.

[0003] In the existing technology, the existing hydrogen evolution anomaly monitoring only focuses on the absolute value of the alkali concentration, which cannot capture small dynamic fluctuations, nor has it established a correlation between concentration changes and bubble dynamics, resulting in the lack of early warning of electrode process anomalies.

[0004] Secondly, in response to the frequent start-ups and shutdowns caused by power grid peak shaving, existing technologies lack targeted prediction of thermal fatigue cracks in metal components, making it difficult to detect the risk of crack initiation caused by cyclic thermal stress in advance.

[0005] In terms of insulation performance monitoring, existing solutions mostly focus solely on changes in insulation resistance without considering the special environment of high temperature, high humidity, and alkaline mist in electrolytic cells. They cannot distinguish the synergistic effects of environmental factors and material degradation itself, making it difficult to accurately predict creepage and short circuit risks. Furthermore, the stage division and early warning of insulation degradation are not precise enough.

[0006] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0007] The purpose of this invention is to solve the problems mentioned above by proposing a method for predicting safety faults in alkaline electrolyzers of hydrogen energy systems.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] A method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system, comprising the following steps:

[0010] S1: Hydrogen evolution anomaly prediction, predicting hydrogen evolution anomalies during the operation of the electrolyzer;

[0011] S2: Crack risk prediction, predicting cracks in metal components within the electrolytic cell based on the operating stage of the electrolytic cell;

[0012] S3: Performance degradation prediction, predicting the insulation performance of internal materials based on the operating environment of the electrolytic cell.

[0013] Furthermore, the process for predicting S1 hydrogen evolution anomalies is as follows:

[0014] The conductivity sensor records the measured values, a high-speed camera unit captures a frame of bubble image on the anode surface, and the current current density value of the current sensor is recorded; the continuously acquired data stream is packaged into a data packet with a unified timestamp.

[0015] Data processing is performed on the data packets at each time point to obtain the measured values ​​of the conductivity sensor in the data packets, i.e., the conductivity data, and the collected conductivity data is smoothed; the conductivity change rate is obtained by combining the conductivity data at each time point.

[0016] The bubble images collected in the data packet are analyzed, and the outline of each bubble is accurately identified using image processing algorithms. Based on the bubble generation process, the dynamic parameters of the bubbles are recorded, including the average diameter, distribution density, and detachment frequency. The dynamic parameters are statistically analyzed at the time of collection and arranged in chronological order to form a feature vector.

[0017] Furthermore, operational data is collected based on the historical operation process of the electrolyzer, and a model is constructed based on the historical operational data. That is, the current density value and conductivity data are set as input values, and the dynamic parameters of the bubbles at the corresponding time in the historical operation process are set as output values. Based on the historical operation process, from the generation of input values ​​to the generation of output values, all execution processes and the corresponding data generated by the execution processes are collected, and the fluctuation of the data processing and generation of the execution processes is set as the internal operation process of the model.

[0018] Record the range of input values ​​in the historical operating data, and record the output value corresponding to the electrolytic cell being in a healthy operating state to obtain the output value range;

[0019] Based on the historical input value range and the historical output value range, record the fluctuation trend of the corresponding input value and the fluctuation trend of the corresponding output value at the same time; and based on the cumulative running time, if the trends correspond, mark the numerical fluctuation trends as trend corresponding groups; if the trends do not correspond, monitor the fluctuation trends.

[0020] Furthermore, based on the current operating workload of the electrolytic cell, it is divided into completed tasks and tasks to be completed, thus obtaining the task completion stage and the task to be completed stage.

[0021] Analyze the task completion phase; record the rate of change of conductivity and the corresponding dynamic parameters of the current bubbles; if the rate of change of conductivity is continuously negative, that is, the concentration continues to decrease, and the actual average diameter of the bubbles in the dynamic parameters is significantly smaller than the lower limit of the average diameter range in the historical output values, and the detachment frequency is significantly higher than the upper limit of the detachment frequency range in the historical output values; in this scenario, it is marked as a coupling mode where the concentration decreases and the bubbles become smaller and faster, and the administrator adjusts and controls the current water quality of the electrolytic cell makeup water;

[0022] If the rate of change of conductivity remains positive, i.e. the concentration continues to increase, and the actual average diameter of bubbles in the bubble dynamics parameters is significantly smaller than the upper limit of the average diameter range in the historical output values, and the distribution density continues to decrease, this scenario is marked as a coupled mode of increasing concentration and decreasing bubble size and number, and the administrator controls the electrolyte viscosity.

[0023] Furthermore, the task completion phase is analyzed. At the end of the current task completion phase, the current density and conductivity data are obtained as real-time input values ​​and substituted into the model to obtain the predicted output value. The input values ​​are adjusted in real time according to the fluctuation trend of adjacent historical moments at the end time, and the predicted output values ​​of adjacent moments are obtained. The fluctuation trend is obtained based on the predicted output value and marked as the predicted fluctuation trend. The corresponding group of predicted trends is constructed.

[0024] Analyze the trend-corresponding group and the predicted trend-corresponding group:

[0025] If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are opposite, then the adjacent time of the prediction will be marked as the time of abnormal trend. If the adjacent time of the prediction are all time of abnormal trend, it is inferred that there is a risk of hydrogen evolution. In the task waiting to be completed stage, an execution preset segment is set. If the trend of the output value in the execution preset segment is consistent with the predicted fluctuation trend, then the operation of the electrolyzer will be intervened. Otherwise, monitoring will continue.

[0026] If the input values ​​in the trend corresponding group and the predicted trend corresponding group have opposite fluctuation trends, and the corresponding output values ​​have the same fluctuation trend, then the execution process and the data generated by the corresponding execution process are compared with the historical and current execution processes. The execution processes or execution data that are different are adjusted to complete the model update of the current running segment.

[0027] If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are consistent, then the predicted hydrogen evolution efficiency of the electrolyzer is normal. If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are opposite, and the fluctuation trends of the corresponding output values ​​are opposite, then it is inferred that the model is not suitable. During the task completion phase, the running time is selected as the data collection period for rebuilding the model, and prediction is made in a timely manner after the data collection and construction are completed.

[0028] Furthermore, the process for predicting the S2 crack risk is as follows:

[0029] Acoustic emission sensors are arranged at key locations such as the inner electrode plates of the electrolytic cell;

[0030] The temperature of the electrolytic cell is collected by the temperature sensor. The temperature change rate is used as the trigger condition. When the temperature change rate exceeds the set change rate threshold, it indicates that the electrolytic cell start-up and shutdown cycle has started. The start-up and shutdown cycle ends when the real-time temperature rises to the working temperature and then returns to the normal temperature, thus forming the entire start-up and shutdown cycle.

[0031] Record the original acoustic emission waveform, acquire the acoustic emission waveform in real time during the start-stop cycle, compare it with the original waveform, and mark the moment when the waveform changes as the stress change moment.

[0032] Record the temperature change span of continuous start-stop cycles, and under the current temperature change span trend, collect the duration corresponding to the stress change moment and the waveform change within the corresponding duration.

[0033] Furthermore, if the duration corresponding to the stress change exceeds the set duration threshold, or the waveform change within the corresponding duration exceeds the change threshold, it is inferred that there is a risk of stress change at the corresponding critical location. The administrator will repair the component at the corresponding critical location and record the temperature change span trend during the stress change phase. If the temperature change span trend is continuously increasing, the start-up and shutdown cycle of the electrolytic cell will be controlled after the component repair is completed, and the rate of increase of the temperature change span will be reduced.

[0034] If the duration corresponding to the stress change does not exceed the set duration threshold, and the waveform change within the corresponding duration does not exceed the change threshold, it is inferred that there is no risk of stress change at the corresponding key location. The start-up and shutdown cycle of the electrolytic cell is continuously recorded and data is monitored.

[0035] Furthermore, the S3 performance degradation prediction process is as follows:

[0036] Dew point temperature is collected using a dew point sensor;

[0037] The ambient temperature is measured using a temperature sensor;

[0038] The subtle fluctuations in internal air pressure are collected using a differential pressure sensor.

[0039] A high-frequency detection signal injection module is configured to superimpose a sine wave detection signal between the energized DC positive and negative buses during operation.

[0040] Data analysis was performed on minor fluctuations in dew point temperature, ambient temperature, and internal air pressure.

[0041] The condensation risk coefficient is obtained by calculating the difference between ambient temperature and dew point temperature and comparing the difference with the ambient temperature. The micro-pressure difference is obtained based on the small fluctuations in internal air pressure, and the standard deviation of the micro-pressure difference is calculated and marked as the micro-pressure standard deviation. This is used to characterize the intensity of internal airflow disturbance. Large disturbances will promote the migration and adhesion of alkaline mist.

[0042] Record the output voltage amplitude of the high-frequency detection signal and the original waveform of the measured leakage current signal; perform spectrum calculation on the acquired high-frequency leakage current waveform to obtain the leakage current amplitude and dielectric loss tangent; and uniformly label the leakage current amplitude and dielectric loss tangent as electrical signal characteristic parameters.

[0043] Furthermore, when the micro-pressure standard deviation is within the set range, and the condensation risk coefficient is continuously lower than the set coefficient threshold, but the electrical signal characteristic parameters do not exceed the corresponding red line threshold, it is inferred that there is a high risk of condensation in the current environment, but the insulation material has not been affected. The warning information is set to an abnormal environment but no impact is generated. When the electrolytic cell is running, the administrator controls the operating environment of the electrolytic cell in real time.

[0044] When the condensation risk coefficient is not lower than the set coefficient threshold, and any and not unique electrical signal characteristic parameter value exceeds the corresponding red line threshold, it is inferred that the insulation material has an impact, and the environment is not the only influencing factor. The warning signal is set to the initial deterioration of insulation performance. When the administrator controls the environment during the operation of the electrolytic cell, it is necessary to adjust the real-time insulation performance impact parameters synchronously, and at the same time maintain the insulation material itself.

[0045] When the condensation risk coefficient remains below the set threshold and any one and only electrical signal characteristic parameter exceeds the corresponding red line threshold, the harsh environment and the material's own deterioration are having a synergistic accelerating effect. The warning signal is set to accelerate the deterioration of insulation performance. When the electrolytic cell is running, the administrator needs to perform insulation performance maintenance and short circuit risk monitoring, and perform short circuit control in a timely manner to avoid short circuits causing electrolytic cell failure and affecting the internal components of the electrolytic cell.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. By fusing multi-source data from conductivity sensors, high-speed camera units, and current sensors, and packaging the dataset with a unified timestamp, the problem of data fragmentation in existing technologies is solved; the conductivity data is smoothed and the rate of change is calculated, breaking through the limitation of only monitoring the absolute value of concentration and realizing the accurate capture of minute dynamic fluctuations in electrolyte concentration; by using image processing algorithms to extract dynamic parameters such as average bubble diameter, distribution density, and detachment frequency, a coupled correlation between concentration change rate and bubble dynamics is established, and the risk of electrode reaction environment deterioration and local overheating can be identified at an early stage through a specific coupling mode;

[0048] By combining historical health data to build a predictive model, and comparing the trends of the completed and pending stages of the task, the model can achieve real-time monitoring of hydrogen evolution anomalies and prediction of future risks. The model can also be adaptively updated according to the corresponding trend, further improving the accuracy of prediction and the adaptability of the scenario.

[0049] 2. For the core operating condition of frequent start-up and shutdown of electrolytic cells, a wideband acoustic emission sensor is used, which is arranged in an array in combination with finite element stress analysis. This effectively captures high-frequency stress waves of plastic deformation and microcrack propagation of metal parts, while avoiding interference from low-frequency mechanical vibration and fluid noise, thus improving the signal-to-noise ratio of crack signal detection. The temperature change rate is used as the trigger condition for the start-up and shutdown cycle to accurately lock the period of thermal stress. By comparing the changes in acoustic emission waveform, the duration of stress change, and the amount of waveform change, early detection of precursors to thermal fatigue crack initiation is achieved.

[0050] The sensor employs a high-temperature coupling agent and waveguide rod design to ensure stable operation at the electrolytic cell operating temperature. The array arrangement supports precise positioning of the acoustic emission source, facilitating targeted repair by the administrator. Furthermore, by analyzing the temperature change span trend, the start-up and shutdown cycle can be adjusted accordingly to reduce the risk of crack propagation.

[0051] 3. By integrating environmental parameters such as dew point temperature, ambient temperature, and minor fluctuations in internal air pressure with electrical signal characteristic parameters such as high-frequency leakage current and dielectric loss tangent, a multi-dimensional monitoring system is constructed, overcoming the limitations of existing single-parameter monitoring. By calculating the condensation risk coefficient and micro-pressure standard deviation, the severity of the internal environment and the impact of airflow disturbance on the insulation material are accurately quantified. The introduction of high-frequency detection signals makes it more sensitive to microscopic defects in the insulation material compared to traditional insulation resistance monitoring. Combined with the calculation of dielectric loss tangent, insulation performance degradation can be detected earlier. Through the combined analysis of the condensation risk coefficient and electrical signal characteristic parameters, insulation degradation is divided into three stages, achieving precise graded early warning and providing administrators with targeted solutions to effectively avoid electrolytic cell malfunctions caused by creepage and short circuits. Attached Figure Description

[0052] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0053] Figure 1 This is a block diagram illustrating the principle of the method of the present invention;

[0054] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0057] Please see Figures 1-2 As shown, a method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system is described, and the specific steps of the safety fault prediction method are as follows:

[0058] S1: Hydrogen evolution anomaly prediction, predicting hydrogen evolution anomalies during the operation of the electrolyzer;

[0059] When an electrolyzer operates at a stable power for an extended period, fluctuations in the purity of the added deionized water or minor leaks can cause slow changes in the electrolyte concentration. These changes directly affect bubble generation, detachment behavior, and electrode processes, serving as early warning signs of decreased efficiency and potential safety hazards. Current technologies only monitor the absolute value of the alkali concentration, failing to detect its minute dynamic fluctuations and unable to correlate concentration fluctuations with changes in bubble dynamics, thus missing early warning opportunities for abnormal electrode processes.

[0060] S2: Crack risk prediction, predicting cracks in metal components within the electrolytic cell based on the operating stage of the electrolytic cell;

[0061] Hydrogen production stations that respond to grid peak shaving require frequent start-ups and shutdowns of electrolyzers; rapid heating during startup and natural cooling during shutdown cause metal components such as electrode plates and end plates to be subjected to cyclic thermal stress, making them prone to thermal fatigue cracks.

[0062] S3: Performance degradation prediction, predicting the insulation performance of internal materials based on the operating environment of the electrolytic cell;

[0063] When an electrolytic cell is in operation, there is a high temperature, high humidity and alkaline mist environment inside. Under long-term exposure, a conductive liquid film may form on the surface of electrical connectors and supporting insulators, or the insulation performance of the materials themselves may decrease, causing creepage or even short circuits.

[0064] The process for predicting S1 hydrogen evolution anomalies is as follows:

[0065] The conductivity sensor records the measured values, a high-speed camera unit captures a frame of bubble image on the anode surface, and the current current density value of the current sensor is recorded; the continuously acquired data stream is packaged into a data packet with a unified timestamp.

[0066] Data processing is performed on the data packets at each time point to obtain the measured values ​​of the conductivity sensor in the data packets, i.e., conductivity data. The collected conductivity data is then smoothed to remove high-frequency noise from the conductivity signal. The conductivity change rate is obtained by combining the conductivity data at each time point. The change rate is a key feature for sensing dynamic micro-changes in concentration. Its absolute value and positive / negative (increase / decrease) indicate the severity and direction of the change.

[0067] The bubble images collected within the data packet are analyzed, and image processing algorithms (such as the watershed algorithm or a deep learning-based instance segmentation model) are used to accurately identify the contour of each bubble. Based on the bubble generation process, the dynamic parameters of the bubbles are recorded, including the average diameter, distribution density, and detachment frequency. The average diameter is the average of the equivalent circle diameters of all bubbles in a single frame image. Distribution density refers to the number of bubbles per unit area of ​​the image. Detachment frequency is calculated by analyzing consecutive image frames and tracking the average time from bubble generation to detachment at a specific point.

[0068] The dynamic parameters are collected and statistically analyzed at different times and arranged in time order to form a feature vector.

[0069] Operational data is collected based on the historical operation process of the electrolyzer. The selected historical operation process corresponds to the electrolyzer's operating status being healthy. A model is built based on the historical operation data, that is, the current density value and conductivity data are set as input values, and the dynamic parameters of the bubbles at the corresponding time in the historical operation process are set as output values. From the generation of input values ​​to the generation of output values ​​in the historical operation process, all execution processes and the corresponding data generated by the execution processes are collected, and the fluctuation of the data processing and generation of the execution processes is set as the internal operation process of the model.

[0070] Record the range of input values ​​in the historical operating data, and record the output value corresponding to the electrolytic cell being in a healthy operating state to obtain the output value range;

[0071] Based on the historical input value range and the historical output value range, record the fluctuation trend of the corresponding input value and the fluctuation trend of the corresponding output value at the same time; and based on the cumulative time of operation, if the trends correspond, mark the numerical fluctuation trends as trend-corresponding groups; if the trends do not correspond, monitor the fluctuation trends.

[0072] Based on the current workload of the electrolyzer, it is divided into completed tasks and tasks to be completed, thus obtaining the task completion stage and the task to be completed stage.

[0073] Analyze the task completion phase; record the rate of change of conductivity and the corresponding dynamic parameters of the current bubbles; if the rate of change of conductivity is continuously negative, that is, the concentration continues to decrease, and the actual average diameter of the bubbles in the dynamic parameters is significantly smaller than the lower limit of the average diameter range in the historical output values, and the detachment frequency is significantly higher than the upper limit of the detachment frequency range in the historical output values; in this scenario, it is marked as a coupling mode of decreasing concentration and smaller and faster bubbles, which strongly indicates that the electrode reaction environment has deteriorated. It may be that impurities in the makeup water have affected the reaction overpotential and bubble surface tension. The administrator should adjust and control the current makeup water quality of the electrolytic cell.

[0074] It should be explained that "significantly less than or higher than" is expressed as a value being less than or higher than 1.3 times the threshold.

[0075] If the rate of change of conductivity remains positive, i.e. the concentration continues to increase, and the actual average diameter of bubbles in the bubble dynamics parameters is significantly smaller than the upper limit of the average diameter range in the historical output values, and the distribution density continues to decrease, this scenario is marked as a coupling mode of increasing concentration and increasing bubble size and number. This may mean that the electrolyte viscosity increases, affecting bubble merging and detachment, and there is a risk of local overheating. Administrators should control the electrolyte viscosity.

[0076] Apart from the two scenarios mentioned above, other scenarios would be considered to indicate that the hydrogen evolution detection in the electrolyzer was normal at the corresponding moment of the task completion stage.

[0077] The task is analyzed in the unfinished stage. At the end of the current task completion stage, the current density and conductivity data are obtained as real-time input values ​​and substituted into the model to obtain the predicted output value. The input values ​​are adjusted in real time according to the fluctuation trend of adjacent historical time points to obtain the predicted output value of adjacent time points. The fluctuation trend is obtained based on the predicted output value and marked as the predicted fluctuation trend. The corresponding group of predicted trends is constructed.

[0078] Analyze the trend-corresponding group and the predicted trend-corresponding group:

[0079] If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are opposite, then the adjacent time of the prediction will be marked as the time of abnormal trend. If the adjacent time of the prediction are all time of abnormal trend, it is inferred that there is a risk of hydrogen evolution. In the task waiting to be completed stage, an execution preset segment is set. If the trend of the output value in the execution preset segment is consistent with the predicted fluctuation trend, then the operation of the electrolyzer will be intervened. Otherwise, monitoring will continue.

[0080] If the input values ​​in the trend corresponding group and the predicted trend corresponding group have opposite fluctuation trends, and the corresponding output values ​​have the same fluctuation trend, then the execution process and the data generated by the corresponding execution process are compared with the historical and current execution processes. The execution processes or execution data that are different are adjusted to complete the model update of the current running segment and improve the prediction accuracy.

[0081] If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are consistent, then the predicted hydrogen evolution efficiency of the electrolyzer is normal; if the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are opposite, and the fluctuation trends of the corresponding output values ​​are opposite, then it is inferred that the model is not suitable. During the task completion phase, the running time is selected as the data collection period for rebuilding the model, and prediction is made in a timely manner after the data collection and construction are completed.

[0082] The process of S2 crack risk prediction is as follows:

[0083] Acoustic emission sensors are placed at key locations such as the electrode plates within the electrolytic cell. The sensor selection should be explained as follows: Broadband acoustic emission sensors are recommended, with a frequency response range of 100kHz-1MHz. This frequency band effectively captures high-frequency stress waves generated by plastic deformation and microcrack propagation in metallic materials, while avoiding most low-frequency interference from mechanical vibrations and fluid noise. Placement strategy: Based on finite element stress analysis results, sensors are placed in areas with the most significant stress concentration on the electrode plates and end plates (such as flow channels near inlets and outlets, and bolt connections). An array arrangement is used (e.g., at least 2-3 sensors per electrode plate) to facilitate subsequent acoustic emission source localization. Installation requirements: High-temperature coupling agent and waveguide rods (if required) are used to ensure stable operation and effective signal transmission of the sensors at the electrolytic cell's operating temperature (typically <90℃).

[0084] The temperature of the electrolytic cell is collected by the temperature sensor. The temperature change rate is used as the trigger condition. When the temperature change rate exceeds the set change rate threshold, it indicates that the electrolytic cell start-up and shutdown cycle has started. The start-up and shutdown cycle ends when the real-time temperature rises to the working temperature and then returns to the normal temperature, thus forming the entire start-up and shutdown cycle.

[0085] Record the original acoustic emission waveform, acquire the acoustic emission waveform in real time during the start-stop cycle, compare it with the original waveform, and mark the moment when the waveform changes as the stress change moment.

[0086] Record the temperature change span of continuous start-stop cycles, and under the current temperature change span trend, collect the duration corresponding to the stress change moment and the waveform change within the corresponding duration. The specific waveform change is based on the change distance of the same point on the waveform as the data standard.

[0087] If the duration of stress change exceeds the set duration threshold, or if the waveform change exceeds the threshold within the duration, it is inferred that there is a risk of stress change at the corresponding critical location. The administrator will repair the component at the corresponding critical location and record the temperature change span trend during the stress change phase. If the temperature change span trend is continuously increasing, the start-up and shutdown cycle of the electrolytic cell will be controlled after the component repair is completed, and the rate of increase of the temperature change span will be reduced.

[0088] If the duration corresponding to the stress change time does not exceed the set duration threshold, and the waveform change within the corresponding duration does not exceed the change threshold, it is inferred that there is no risk of stress change at the corresponding key location. The start-up and shutdown cycle of the electrolytic cell is continuously recorded and data is monitored.

[0089] The S3 performance degradation prediction process is as follows:

[0090] Dew point temperature is collected by a dew point sensor; a high-temperature mirror condenser dew point meter or a high-performance polymer capacitive dew point sensor is used, which is directly installed in the gas collection chamber at the top of the electrolytic cell to accurately measure the dew point temperature of the gas space.

[0091] Ambient temperature is measured using a temperature sensor; a platinum resistance thermometer is installed near the dew point sensor.

[0092] The micro differential pressure sensor collects minute fluctuations in internal air pressure; a high-precision sensor is connected between the inside of the electrolytic cell and the external environment to monitor minute fluctuations in internal air pressure. These fluctuations are related to gas release and the operating conditions of the circulation pump, and can indirectly reflect the intensity of internal airflow activity.

[0093] A high-frequency detection signal injection module is set up. When the equipment is running, this module superimposes a sinusoidal detection signal with a frequency of 1kHz-10kHz and an amplitude of extremely low (e.g., <5V) between the energized DC positive and negative buses (or to ground). This high-frequency signal is much higher than the power frequency / DC component, making it easy to separate and more sensitive to microscopic defects in the insulating material. The leakage current of this high-frequency signal flowing through the insulating material is detected by the leakage current detection module.

[0094] Data analysis was performed on minor fluctuations in dew point temperature, ambient temperature, and internal air pressure.

[0095] The condensation risk coefficient is obtained by calculating the difference between ambient temperature and dew point temperature and comparing the difference with the ambient temperature. The calculation is a relative value and dimensionless. The micro-pressure difference is obtained based on the small fluctuations in internal air pressure, and the standard deviation of the micro-pressure difference is calculated and marked as the micro-pressure standard deviation. This is used to characterize the intensity of internal airflow disturbance. Large disturbances will promote the migration and adhesion of alkaline mist.

[0096] Record the output voltage amplitude of the high-frequency detection signal and the original waveform of the measured leakage current signal; perform spectrum calculation on the acquired high-frequency leakage current waveform to obtain the leakage current amplitude and dielectric loss tangent; leakage current amplitude: the magnitude of the amplitude at the detection frequency; dielectric loss tangent (tanδ): calculated by comparing the phase difference (θ) between the leakage current waveform and the original detection voltage waveform, tanδ≈tan(90°-θ); tanδ is a very sensitive characteristic quantity reflecting the polarization loss and degradation degree of the insulating material, and its increase will precede the significant decrease in insulation resistance; the leakage current amplitude and dielectric loss tangent are uniformly labeled as electrical signal characteristic parameters;

[0097] When the micro-pressure standard deviation is within the set range, and the condensation risk coefficient is continuously lower than the set coefficient threshold, but the electrical signal characteristic parameters do not exceed the corresponding red line threshold, it is inferred that there is a high risk of condensation in the current environment, but the insulation material has not been affected. The warning information is set to an abnormal environment but no impact. When the electrolytic cell is running, the administrator controls the operating environment of the electrolytic cell in real time.

[0098] When the condensation risk coefficient is not lower than the set coefficient threshold, and any and not unique electrical signal characteristic parameter value exceeds the corresponding red line threshold, it is inferred that the insulation material has an impact, and the environment is not the only influencing factor. The warning signal is set to the initial deterioration of insulation performance. When the administrator controls the environment during the operation of the electrolytic cell, it is necessary to adjust the real-time insulation performance impact parameters synchronously, and at the same time maintain the insulation material itself.

[0099] When the condensation risk coefficient is consistently lower than the set threshold, and any and any non-unique electrical signal characteristic parameter exceeds the corresponding red line threshold, the harsh environment and the material's own deterioration are having a synergistic accelerating effect. The warning signal is set to accelerate the deterioration of insulation performance. When the electrolytic cell is running, the administrator needs to perform insulation performance maintenance and short circuit risk monitoring, and perform short circuit control in a timely manner to avoid short circuits causing electrolytic cell failure and affecting the internal components of the electrolytic cell.

[0100] When the micro-pressure standard deviation is not within the set range, the collected condensation risk coefficient and electrical signal characteristic parameters are recorded and the internal airflow disturbance is controlled. After stabilization, data analysis is performed and the recorded data is used as reference data.

[0101] Thresholds, preset values, or preset ranges are set for result comparison and analysis to determine whether they are good or bad. The value of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or rational factors.

[0102] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system, characterized in that, The steps of the safety fault prediction method are as follows: S1: Hydrogen evolution anomaly prediction, predicting hydrogen evolution anomalies during the operation of the electrolyzer; S2: Crack risk prediction, predicting cracks in metal components within the electrolytic cell based on the operating stage of the electrolytic cell; S3: Performance degradation prediction, predicting the insulation performance of internal materials based on the operating environment of the electrolytic cell.

2. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 1, characterized in that, The process for predicting S1 hydrogen evolution anomalies is as follows: The conductivity sensor records the measured values, a high-speed camera unit captures a frame of bubble image on the anode surface, and the current current density value of the current sensor is recorded; the continuously acquired data stream is packaged into a data packet with a unified timestamp. Data processing is performed on the data packets at each time point to obtain the measured values ​​of the conductivity sensor in the data packets, i.e., the conductivity data, and the collected conductivity data is smoothed; the conductivity change rate is obtained by combining the conductivity data at each time point. The bubble images collected in the data packet are analyzed, and the outline of each bubble is accurately identified using image processing algorithms. Based on the bubble generation process, the dynamic parameters of the bubbles are recorded, including the average diameter, distribution density, and detachment frequency. The dynamic parameters are statistically analyzed at the time of collection and arranged in chronological order to form a feature vector.

3. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 2, characterized in that, Operational data is collected based on the historical operation process of the electrolyzer. A model is built based on the historical operation data, that is, the current density value and conductivity data are set as input values, and the dynamic parameters of the bubbles at the corresponding time in the historical operation process are set as output values. From the generation of input values ​​to the generation of output values ​​in the historical operation process, all execution processes and the corresponding data generated by the execution processes are collected, and the fluctuation of the data processing and generation of the execution processes is set as the internal operation process of the model. Record the range of input values ​​in the historical operating data, and record the output value corresponding to the electrolytic cell being in a healthy operating state to obtain the output value range; Based on the historical input value range and the historical output value range, record the fluctuation trend of the corresponding input value and the fluctuation trend of the corresponding output value at the same moment; Based on the cumulative running time, if the trends correspond, the numerical fluctuation trends are marked as trend corresponding groups; If the trend does not correspond, the floating trend will be monitored.

4. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 3, characterized in that, Based on the current workload of the electrolyzer, it is divided into completed tasks and tasks to be completed, thus obtaining the task completion stage and the task to be completed stage. Analyze the task completion stage; record the rate of change of conductivity and the corresponding dynamic parameters of the current bubble; if the rate of change of conductivity is continuously negative, that is, the concentration continues to decrease, and the actual average diameter of the bubble in the bubble dynamic parameters is significantly smaller than the lower limit of the average diameter range in the historical output values, and the escape frequency is significantly higher than the upper limit of the escape frequency range in the historical output values. In this scenario, the coupling mode is marked as a decrease in concentration and a decrease in the size and speed of bubbles. The administrator then adjusts and controls the quality of the water added to the electrolytic cell. If the rate of change of conductivity remains positive, i.e. the concentration continues to increase, and the actual average diameter of bubbles in the bubble dynamics parameters is significantly smaller than the upper limit of the average diameter range in the historical output values, and the distribution density continues to decrease, this scenario is marked as a coupled mode of increasing concentration and decreasing bubble size and number, and the administrator controls the electrolyte viscosity.

5. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 4, characterized in that, The task is analyzed in the unfinished stage. At the end of the current task completion stage, the current density value and conductivity data are obtained as real-time input values ​​and substituted into the model to obtain the predicted output value. The real-time input value is adjusted according to the fluctuation trend of adjacent historical time points at the end time, and the predicted output value of adjacent time points is obtained. The fluctuation trend is obtained based on the predicted output value and marked as the predicted fluctuation trend. And construct corresponding groups for predicted trends; Analyze the trend-corresponding group and the predicted trend-corresponding group: If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are opposite, then the adjacent time of the prediction will be marked as the time of abnormal trend. If the adjacent time of the prediction are all time of abnormal trend, it is inferred that there is a risk of hydrogen evolution. In the task waiting to be completed stage, an execution preset segment is set. If the trend of the output value in the execution preset segment is consistent with the predicted fluctuation trend, then the operation of the electrolyzer will be intervened. Otherwise, monitoring will continue. If the input values ​​in the trend corresponding group and the predicted trend corresponding group have opposite fluctuation trends, and the corresponding output values ​​have the same fluctuation trend, then the execution process and the data generated by the corresponding execution process are compared with the historical and current execution processes. The execution processes or execution data that are different are adjusted to complete the model update of the current running segment. If the fluctuation trends of the input values ​​in the trend corresponding group and the predicted trend corresponding group are consistent, and the fluctuation trends of the corresponding output values ​​are consistent, then the predicted hydrogen evolution efficiency of the electrolyzer is normal. If the input values ​​in the trend corresponding group and the predicted trend corresponding group have opposite fluctuation trends, and the corresponding output values ​​have opposite fluctuation trends, it is inferred that the model is not suitable. During the task completion phase, the running time is selected as the data collection period for rebuilding the model, and prediction is made in a timely manner after the data collection and construction are completed.

6. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 1, characterized in that, The process of S2 crack risk prediction is as follows: Acoustic emission sensors are arranged at key locations within the electrolytic cell; The temperature of the electrolytic cell is collected by the temperature sensor. The temperature change rate is used as the trigger condition. When the temperature change rate exceeds the set change rate threshold, it indicates that the electrolytic cell start-up and shutdown cycle has started. The start-up and shutdown cycle ends when the real-time temperature rises to the working temperature and then returns to the normal temperature, thus forming the entire start-up and shutdown cycle. Record the original acoustic emission waveform, acquire the acoustic emission waveform in real time during the start-stop cycle, compare it with the original waveform, and mark the moment when the waveform changes as the stress change moment. Record the temperature change span of continuous start-stop cycles, and under the current temperature change span trend, collect the duration corresponding to the stress change moment and the waveform change within the corresponding duration.

7. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 6, characterized in that, If the duration of stress change exceeds the set duration threshold, or if the waveform change exceeds the threshold within the duration, it is inferred that there is a risk of stress change at the corresponding critical location. The administrator will repair the component at the corresponding critical location and record the temperature change span trend during the stress change phase. If the temperature change span trend is continuously increasing, the start-up and shutdown cycle of the electrolytic cell will be controlled after the component repair is completed, and the rate of increase of the temperature change span will be reduced. If the duration corresponding to the stress change does not exceed the set duration threshold, and the waveform change within the corresponding duration does not exceed the change threshold, it is inferred that there is no risk of stress change at the corresponding key location. The start-up and shutdown cycle of the electrolytic cell is continuously recorded and data is monitored.

8. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 1, characterized in that, The S3 performance degradation prediction process is as follows: Dew point temperature is collected using a dew point sensor; ambient temperature is measured using a temperature sensor; and minute fluctuations in internal air pressure are collected using a differential pressure sensor. A high-frequency detection signal injection module is set up to superimpose a sine wave detection signal between the energized DC positive and negative buses during operation; Data analysis was performed on minor fluctuations in dew point temperature, ambient temperature, and internal air pressure. The condensation risk coefficient is obtained by calculating the difference between ambient temperature and dew point temperature and comparing the difference with the ambient temperature. The micro-pressure difference is obtained based on the small fluctuations in internal air pressure, and the standard deviation of the micro-pressure difference is calculated and marked as the micro-pressure standard deviation. This is used to characterize the intensity of internal airflow disturbance. Large disturbances will promote the migration and adhesion of alkaline mist. Record the output voltage amplitude of the high-frequency detection signal and the original waveform of the measured leakage current signal; perform spectrum calculation on the acquired high-frequency leakage current waveform to obtain the leakage current amplitude and dielectric loss tangent; and uniformly label the leakage current amplitude and dielectric loss tangent as electrical signal characteristic parameters.

9. The method for predicting safety faults in an alkaline electrolyzer of a hydrogen energy system according to claim 8, characterized in that, When the micro-pressure standard deviation is within the set range, and the condensation risk coefficient is continuously lower than the set coefficient threshold, but the electrical signal characteristic parameters do not exceed the corresponding red line threshold, it is inferred that there is a high risk of condensation in the current environment, but the insulation material has not been affected. The warning information is set to an abnormal environment but no impact. When the electrolytic cell is running, the administrator controls the operating environment of the electrolytic cell in real time. When the condensation risk coefficient is not lower than the set coefficient threshold, and any and not unique electrical signal characteristic parameter value exceeds the corresponding red line threshold, it is inferred that the insulation material has an impact, and the environment is not the only influencing factor. The warning signal is set to the initial deterioration of insulation performance. When the administrator controls the environment during the operation of the electrolytic cell, it is necessary to adjust the real-time insulation performance impact parameters synchronously, and at the same time maintain the insulation material itself. When the condensation risk coefficient remains below the set threshold and any one and only electrical signal characteristic parameter exceeds the corresponding red line threshold, the harsh environment and the material's own deterioration are having a synergistic accelerating effect. The warning signal is set to accelerate the deterioration of insulation performance. When the electrolytic cell is running, the administrator needs to perform insulation performance maintenance and short circuit risk monitoring, and perform short circuit control in a timely manner to avoid short circuits causing electrolytic cell failure and affecting the internal components of the electrolytic cell.

Images