Abnormality monitoring method for cell of electrolytic cell, electronic equipment and storage medium

Abstract

The embodiment of the invention relates to the field of electrolytic cell monitoring, and discloses an abnormal monitoring method for a cell of an electrolytic cell, electronic equipment and a storage medium. The method comprises the following steps: acquiring voltage data of a plurality of small chambers in the electrolytic cell in real time; sequentially taking each of the plurality of small chambers as a target small chamber, and analyzing the target small chamber by using a plurality of voltage anomaly detection models to obtain an anomaly detection result corresponding to each voltage anomaly detection model; wherein the plurality of voltage anomaly detection models at least comprise a first type of models used for detecting the voltage fluctuation anomaly of a target cell, and a second type of models used for detecting the voltage fluctuation anomaly of the target cell relative to the plurality of cells; and based on the anomaly detection results of the plurality of voltage anomaly detection models of the target cell, judging the anomaly level of the target cell.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of electrolytic cell monitoring, and particularly to anomaly monitoring methods, electronic devices and storage media for electrolytic cell chambers. Background Technology

[0002] Electrolyzers, as core equipment in green chemistry, typically consist of numerous cells connected in series (for example, more than ten thousand in a GW-scale hydrogen production plant). Each cell contains positive and negative electrodes, an electrolyte or ion exchange membrane, and corresponding fluid channels. Cell voltage, as a key parameter reflecting the health of the equipment, can effectively characterize various fault states such as short circuits, electrolyte line blockages, and membrane perforation. Therefore, real-time monitoring of cell voltage is an important technical means to ensure the safe operation of electrolyzers.

[0003] In related technologies, monitoring the voltage of electrolyzer cells typically employs a single monitoring logic, such as setting fixed upper and lower voltage thresholds. However, in practical applications, especially in scenarios utilizing renewable energy sources like wind and solar power for hydrogen electrolysis, the operating load of the electrolyzer exhibits a wide range of dynamic changes with fluctuations in wind and solar power. These drastic load changes cause synchronized and significant voltage fluctuations in all cells. In such cases, existing monitoring methods based on a single logic (e.g., fixed thresholds) are prone to misinterpreting normal load fluctuations as faults, resulting in a very high false alarm rate. Furthermore, existing monitoring solutions often lack a tiered early warning mechanism; once a threshold is triggered, the same alarm or shutdown operation is executed, failing to differentiate the severity of the fault. This makes it difficult for maintenance personnel to take targeted maintenance measures based on the urgency of the alarm information, increasing the maintenance burden and potentially causing unnecessary unplanned downtime. Summary of the Invention

[0004] The purpose of this invention is to provide an abnormal monitoring method, electronic device and storage medium for electrolytic cell chambers, in order to solve the problems of high false alarm rate of voltage monitoring caused by electrolytic cell load fluctuation and lack of graded early warning mechanism in related technologies.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide an anomaly monitoring method for electrolytic cell chambers, comprising: real-time acquisition of voltage data from multiple chambers in an electrolytic cell; sequentially selecting each of the multiple chambers as a target chamber, analyzing the target chamber using multiple voltage anomaly detection models, and obtaining anomaly detection results corresponding to each voltage anomaly detection model; wherein the multiple voltage anomaly detection models include at least a first type of model for detecting voltage fluctuation anomalies in the target chamber itself, and a second type of model for detecting voltage fluctuation anomalies in the target chamber relative to the multiple chambers; and determining the anomaly level of the target chamber based on the anomaly detection results of the multiple voltage anomaly detection models of the target chamber.

[0006] Embodiments of the present invention also provide an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the abnormal monitoring method for an electrolytic cell as described above.

[0007] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the abnormal monitoring method for the electrolytic cell chamber as described above.

[0008] In this embodiment of the invention, a "multimodal" monitoring approach is introduced, utilizing multiple voltage anomaly detection models to simultaneously analyze the target cell. Specifically, the system combines a "first-type model" that focuses on the absolute change characteristics of the cell itself and a "second-type model" that focuses on the change characteristics of the cell relative to the group. This design effectively addresses the wide load fluctuation problem in wind and solar hydrogen production scenarios: when the overall load of the electrolyzer fluctuates, although the "first-type model" may trigger anomalies due to changes in the absolute voltage value, the "second-type model" can identify whether the trend of the cell's change is consistent with that of the group. Thus, the comprehensive judgment of multiple models effectively filters out false alarms caused by operating condition fluctuations, significantly improving the accuracy of fault identification. Furthermore, this invention no longer provides a simple "alarm / no alarm" binary output, but rather comprehensively judges the "anomaly level" based on the detection results of multiple models (e.g., the degree to which multiple models alarm simultaneously). This enables the system to achieve graded early warning (such as a red-yellow-green light mechanism), providing maintenance personnel with intuitive risk alerts (e.g., low level corresponds to attention and inspection, high level corresponds to immediate shutdown), thereby optimizing maintenance efficiency and equipment continuity while ensuring safety.

[0009] Furthermore, each of the multiple voltage anomaly detection models has a preset weight coefficient. Determining the anomaly level of the target compartment based on the anomaly detection results of the multiple voltage anomaly detection models includes: performing a weighted summation of the anomaly detection results of the multiple voltage anomaly detection models and the weight coefficients to obtain a weighted value; and determining the anomaly level of the target compartment based on the weighted value. Thus, by assigning weights to different models, differentiated configurations can be made according to the confidence level or sensitivity to specific faults of different models, making the final judgment result more consistent with actual working conditions and improving the scientific nature of the comprehensive judgment.

[0010] Furthermore, the weight coefficient of the second type of model is greater than that of the first type of model, and / or the number of models in the second type is greater than the number of models in the first type. Therefore, in scenarios with frequent load fluctuations, such as wind-solar coupled hydrogen production, increasing the weight or number of "second type models" (which focus on relative deviations and have strong resistance to overall load fluctuations) can further reduce the risk of false alarms caused by systemic load adjustments, ensuring the robustness of the system under dynamic operating conditions. Attached Figure Description

[0011] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0012] Figure 1 This is a flowchart of an abnormal monitoring method for an electrolytic cell chamber provided in this application; Figure 2 These are example diagrams of two types of models used in an abnormal monitoring method for an electrolytic cell chamber provided in this application; Figure 3 This is a first trend diagram of the target cell monitored by an abnormal monitoring method for an electrolytic cell cell provided in this application; Figure 4 This is a second trend diagram of the target cell monitored by an abnormal monitoring method for an electrolytic cell cell provided in this application; Figure 5 This is a structural diagram of an electronic device provided in this application. Detailed Implementation

[0013] As described in the background section, related technologies suffer from high false alarm rates in voltage monitoring due to fluctuations in electrolyzer load and a lack of tiered early warning mechanisms. The inventors discovered that the root cause of this problem lies in the fact that, in scenarios utilizing renewable energy sources such as wind and solar power for hydrogen electrolysis, the operating conditions (current and voltage) of the electrolyzer are primarily governed by the input power, exhibiting wide-range, high-frequency dynamic fluctuations. These macroscopic operating condition fluctuations introduce strong common-mode interference into the voltage signals of all cells. Related technologies employing a single monitoring logic (e.g., a fixed threshold model focusing only on the absolute value of voltage) often struggle to decouple the weak differential-mode signals characterizing individual cell faults from the drastically changing common-mode signals, failing to distinguish between "voltage increases caused by overall load increases" and "voltage increases caused by cell faults," thus easily triggering false alarms. Furthermore, due to the lack of comprehensive weighting and hierarchical processing of the output results of different monitoring models, the system can only output a binary "alarm / no alarm" result, and cannot provide feedback to maintenance personnel on the urgency of the fault (such as minor anomalies that only require attention and serious faults that require shutdown), resulting in low efficiency of maintenance decision-making.

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of the present invention to facilitate a better understanding of this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.

[0015] One embodiment of the present invention relates to an anomaly monitoring method for electrolytic cell chambers. This method can be applied to electronic devices in an electrolytic cell monitoring system, such as industrial control computers, monitoring servers, or management platforms deployed in the cloud. The electronic device communicates with each chamber of the electrolytic cell through a data acquisition system to obtain real-time voltage data. The embodiment includes: real-time acquisition of voltage data from multiple chambers in the electrolytic cell; sequentially treating each chamber as a target chamber, analyzing the target chamber using multiple voltage anomaly detection models, and obtaining anomaly detection results corresponding to each voltage anomaly detection model; wherein the multiple voltage anomaly detection models include at least a first-type model for detecting voltage fluctuation anomalies within the target chamber itself, and a second-type model for detecting voltage fluctuation anomalies of the target chamber relative to the multiple chambers; and determining the anomaly level of the target chamber based on the anomaly detection results of the multiple voltage anomaly detection models for the target chamber. In this embodiment, a "multimodal" monitoring approach is introduced, using multiple voltage anomaly detection models to simultaneously analyze the target chamber. Specifically, the system combines a "Type I model" that focuses on the absolute change characteristics of the individual cell and a "Type II model" that focuses on the change characteristics of the cell relative to the group. This design effectively addresses the wide load fluctuation problem in wind and solar hydrogen production scenarios: when the overall load of the electrolyzer fluctuates, although the "Type I model" may trigger anomalies due to changes in the absolute value of the voltage, the "Type II model" can identify whether the trend of the cell's change is consistent with that of the group. Thus, the comprehensive judgment of multiple models effectively filters out false alarms caused by operating condition fluctuations, significantly improving the accuracy of fault identification. Furthermore, this invention no longer provides a simple binary output of "alarm / no alarm," but rather comprehensively judges the "anomaly level" based on the detection results of multiple models (e.g., the degree to which multiple models alarm simultaneously). This enables the system to implement tiered early warning (such as a red-yellow-green light mechanism), providing maintenance personnel with intuitive risk alerts (e.g., low level corresponds to attention and inspection, high level corresponds to immediate shutdown), thereby optimizing maintenance efficiency and equipment continuity while ensuring safety.

[0016] The following describes in detail the implementation details of the abnormal monitoring method for the electrolytic cell chamber according to an embodiment of the present invention. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0017] like Figure 1 As shown, this embodiment of the invention provides an abnormal monitoring method for an electrolytic cell chamber, the process of which specifically includes the following steps 110 to 130.

[0018] Step 110: Real-time acquisition of voltage data from multiple chambers in the electrolytic cell.

[0019] Specifically, an electrolytic cell typically consists of dozens to hundreds of cells connected in series. The monitoring system synchronously acquires the operating voltage of each cell using high-precision voltage sensors at a preset sampling frequency (e.g., 1 Hz or higher). The acquired analog voltage signals are converted from analog to digital signals and then transmitted to a real-time database for storage.

[0020] In an alternative embodiment, in order to eliminate the interference of the on-site electromagnetic environment on weak voltage signals, the voltage data can be pre-processed by filtering before inputting it into the model. For example, a moving average filter or Kalman filter algorithm can be used to remove high-frequency noise and ensure the quality of the input data for subsequent model analysis.

[0021] Step 120: Sequentially take each of the multiple small chambers as the target small chamber, and use multiple voltage anomaly detection models to analyze the target small chamber to obtain the anomaly detection result corresponding to each voltage anomaly detection model.

[0022] In step 120, the system employs a "multimodal" analysis strategy, which involves running multiple mathematical models with different mechanisms in parallel to cross-validate the state of the target chamber. For example... Figure 2 As shown, these models are divided into two main categories: the first category is used to detect abnormal voltage fluctuations in the absolute value of the target cell itself; the second category is used to detect abnormal voltage fluctuations in the target cell relative to the behavior of the entire electrolyzer group.

[0023] In a specific example, the first type of model includes a fixed threshold abrupt change detection model. This model focuses on sudden and drastic voltage fluctuations. Specifically, it calculates the voltage difference (i.e., the first-order difference) between the target chamber at the current moment and the previous moment; if the absolute value of this voltage difference exceeds a preset fixed voltage threshold, the model's detection result is determined to be abnormal. This model can quickly respond to sudden hard faults such as short circuits or open circuits. Optionally, the preset fixed voltage threshold involved in the above fixed threshold abrupt change detection model can be set according to specific equipment parameters (such as the rated voltage of the electrolyzer, the maximum allowable fluctuation range, etc.).

[0024] In a specific example, the second type of model eliminates common-mode interference through statistical methods, specifically including at least one of the following three models: 1. Robust Z-score Outlier Detection Model: This model is used to capture significant anomalies relative to the overall change pattern. The system calculates the voltage difference between the current and previous time steps in the target cell. V i Simultaneously, the median (M) and absolute median difference (MAD) of the voltage difference across all cells are calculated. Then, the formula is used... (where 'a' is a preset coefficient) Calculate the corrected Z-value. If the corrected Z-value is greater than the preset Z-value threshold, the small cell is determined to be abnormal. This method uses the median instead of the mean, is insensitive to outliers, and has stronger robustness.

[0025] 2. Relative Deviation Abrupt Change Detection Model: This model focuses on abnormal patterns where the absolute value is normal but the relative position changes drastically. The system calculates the average voltage (V) of all cells at the current moment. avg ), and calculate the difference between the target cell voltage and the mean (D). i Further calculate the change in this difference over a period of time. D i If the change exceeds the preset deviation threshold, it indicates that although the small chamber follows the overall fluctuation, its relative position has shifted abnormally, and it is judged as abnormal.

[0026] 3. Dynamic Standard Deviation Outlier Detection Model: This model focuses on cells whose fluctuation patterns significantly deviate from the population characteristics. The system calculates the standard deviation (σ) of the voltage in the target cell over a preset time period (e.g., the past 60 seconds). i This characterizes the degree of fluctuation. Simultaneously, the median (M) of this standard deviation for all cells is calculated. σ ), and set a dynamic threshold (T) based on the median, for example (b is a preset coefficient). If the standard deviation of the target cell is σ i If the value exceeds the dynamic threshold T, it is considered abnormal.

[0027] Those skilled in the art will recognize that the three second-type models provided in the above embodiments each focus on specific characteristics of changes in the electrolysis chamber. The combination of these three models can minimize the blind spot in the monitoring of anomalies in the electrolysis chamber. Specifically, when the system simultaneously activates the relative deviation model focusing on "central trend shift," the dynamic standard deviation model focusing on "dispersion anomalies," and the robust Z-score model focusing on "distribution outliers," it can maximize the detection of relative fluctuations in the electrolysis chamber. These three models are mathematically complementary: the former can detect "quiet voltage drift," the middle model can detect "sharp voltage fluctuations," and the latter can detect "statistical outliers." Through this multi-dimensional cross-validation, even if a specific fault in the electrolysis chamber is not obvious in a single statistical dimension (e.g., the mean remains unchanged but the variance increases), it can still be keenly detected by models in other dimensions.

[0028] Step 130: Based on the anomaly detection results of multiple voltage anomaly detection models for the target cell, determine the anomaly level of the target cell. The multiple voltage anomaly detection models include at least a first-type model for detecting voltage fluctuation anomalies within the target cell itself, and a second-type model for detecting voltage fluctuation anomalies of the target cell relative to the multiple cells.

[0029] After obtaining the outputs of the above multiple models (e.g., 0 represents normal and 1 represents abnormal), the system does not adopt a "one-vote veto" system, but instead uses a weighted fusion mechanism to assess the risk level.

[0030] In a specific example, each voltage anomaly detection model has a preset weight coefficient. Step 130 above includes: determining the anomaly level of the target cell based on the anomaly detection results of multiple voltage anomaly detection models of the target cell, including: performing a weighted summation based on the anomaly detection results of multiple voltage anomaly detection models of the target cell and the weight coefficients to obtain a weighted value; and determining the anomaly level of the target cell based on the weighted value.

[0031] In a preferred example, considering the high-frequency, large-amplitude fluctuations in electrolyzer load during wind-solar coupled hydrogen production (e.g., instantaneous changes with wind and solar resources), the system is configured to employ an asymmetric weighting or model number configuration strategy. Therefore, the system can assign a greater weight coefficient to the second type of model than to the first type of model, and / or set the number of second-type models used to be greater than the number of first-type models. Specifically, the first type of model (e.g., a fixed threshold model) primarily relies on absolute voltage values ​​for judgment. When an overall increase in electrolyzer load causes a synchronous increase in the voltage of all cells (i.e., generating a common-mode signal), this type of model is prone to misjudging normal operating fluctuations as faults, resulting in a high false alarm rate. Conversely, the second type of model (e.g., median absolute deviation analysis, relative deviation mutation detection) focuses on analyzing the deviation of the target cell from the group trend (i.e., differential-mode signal), automatically filtering out common voltage changes in all cells caused by load fluctuations. Therefore, it has inherent anti-interference capabilities and higher fault detection confidence under dynamic operating conditions. Based on this, to improve the robustness of the system, the system is configured to satisfy at least one of the following conditions.

[0032] 1. Weight skew: The weight coefficients of the second type of model are set to be greater than those of the first type of model. For example, in a normalized weight system, the weight of the first type of model is set to 0.2, while the weight of the second type of model (or the sum of multiple second type models) is set to 0.8, so that the final anomaly level determination is more inclined to rely on the results of relative deviation analysis.

[0033] 2. Quantitative Advantage: In terms of model deployment, the number of second-type models exceeds the number of first-type models. For example, the system uses only one fixed-threshold model as a bottom-line protection, while simultaneously using three different statistical relative deviation models in parallel. Through this "majority rule" mechanism, even if the first-type model issues a false alarm signal during periods of drastic load fluctuations, it will be smoothed out by the comprehensive decision-making mechanism due to its low weight or fewer votes, thus effectively avoiding unnecessary unplanned downtime while ensuring safety.

[0034] In an optional example, although the first type of model has a lower weight in routine monitoring, the system sets a "fatal fault threshold" for it (e.g., voltage dropping to 0V or exceeding the 2.5V safety limit). Once the detection result of the first type of model reaches this fatal fault threshold, the system will bypass the aforementioned weighted summation logic, directly determine the target chamber as the highest anomaly level, and trigger a shutdown. This design ensures that in the event of instantaneous catastrophic faults such as "electrolyte breakdown" or "severe short circuit," the system will not delay millisecond-level protection opportunities due to waiting for the statistical calculation of the second type of model or insufficient weighted scores, reflecting the highest principle of "safety as a backup."

[0035] In a specific example, the system performs graded early warning based on the weighted value obtained from the above steps, including: when the weighted value is greater than or equal to a preset first early warning threshold and less than a preset second early warning threshold, the abnormality level is determined to be the first level; wherein, the second early warning threshold is greater than the first early warning threshold; when the weighted value is greater than or equal to the preset second early warning threshold, the abnormality level is determined to be the second level, and a shutdown control command is output.

[0036] Specifically, when the weighted value is greater than or equal to the preset first warning threshold and less than the second warning threshold (e.g., only some models alarm, or the weighted score is in the medium-risk range), the system determines that the target room is in a "Level 1 Anomaly". At this time, the system identifies this state as a "potential risk" rather than an "immediate danger," and therefore adopts a strategy of continuous operation and enhanced monitoring. The system outputs a yellow warning signal on the main control screen and highlights the target room's UI icon, prompting the operator to pay close attention to the room. Furthermore, the system automatically marks the data for this time period as "suspected anomaly samples" for subsequent model self-learning calibration. Finally, the system automatically increases the sampling frequency for the target room (e.g., from 1Hz to 10Hz) or triggers a momentary "active verification" process to further confirm the persistence of the anomaly and prevent misjudgments caused by transient interference.

[0037] When the weighted value is greater than or equal to the preset second warning threshold (e.g., all models alarm, or the second-class model with high weight has extremely high confidence), the system determines that the target cell is in a "second-level anomaly". At this time, the system identifies this state as a "serious fault" (such as diaphragm perforation or short circuit), and continued operation may lead to equipment damage or safety accidents. Therefore, the system adopts the highest priority blocking strategy. At this time, the system bypasses the operator's authority and directly triggers the interlock shutdown control command, quickly cutting off the DC power supply to the electrolyzer and simultaneously interlocking to close the feed valve or open the pressure relief valve. Then, the system outputs a red alarm signal, accompanied by a high-decibel buzzer alarm, to ensure that on-site and central control room personnel can perceive and intervene in the situation immediately.

[0038] like Figure 3 As shown, taking a monitoring window at a certain moment as an example, the voltage curve of the target cell exhibited an atypical downward trend with oscillations between 14:40 and 15:00, followed by a step jump around 15:07. Although the absolute value of the voltage at this time (approximately 1850mV-1950mV) did not reach the system's shutdown safety threshold, its fluctuation pattern deviated significantly from the stable trend of the group. In this situation, although the first type of model (fixed threshold) may not trigger an alarm, multiple second type models (such as the dynamic standard deviation model) identified the distribution anomaly. Therefore, the calculated comprehensive weighted value at this time is greater than or equal to the preset first warning threshold and less than the second warning threshold (e.g., only some models alarm). The system determines it as a first-level anomaly and outputs a yellow light signal, prompting the operator to pay attention to the cell, but does not trigger a shutdown to ensure production continuity.

[0039] like Figure 4 As shown, under another operating condition, the voltage of the target chamber remained stable (approximately 2000mV) before 13:36, but exhibited a monotonous, exponential, and dramatic upward trend from 13:36 onwards, exceeding 6000mV within a short period. This voltage behavior indicates an extremely high-risk physical fault (such as gas resistance, membrane drying, or severe polarization). In this scenario, both the first type of model and all second type models are triggered simultaneously. Therefore, the calculated comprehensive weighted value far exceeds the preset second warning threshold. The system determines it to be a second-level anomaly, indicating a severe fault with extremely high confidence. The system outputs a red alarm and directly bypasses operator permissions, triggering an interlock shutdown control command to prevent the electrolyzer from burning out or causing a safety accident.

[0040] In an optional example, to prevent frequent alarm state switching caused by slight fluctuations in the weighted value near the threshold, the system also introduces a hysteresis comparison strategy, or a time window confirmation mechanism, when determining the anomaly level. For example, the corresponding level change is only triggered when the weighted value exceeds the threshold for N consecutive sampling periods (e.g., 2 consecutive seconds). This design effectively filters out occasional glitches and further enhances the system's anti-interference capability.

[0041] In an optional embodiment, considering the different performance of the models under different load ranges, the aforementioned weighting coefficients are not static. The system can acquire the operating load of the electrolyzer (such as current density) in real time and dynamically adjust the weighting coefficients of each model according to the current load range. For example, under low load fluctuation conditions, the weight of the robust Z-score model can be further increased, thereby further reducing the false alarm rate.

[0042] In an optional example, to address the aging drift problem that occurs in electrolyzers as operating time increases, the system is also configured with a self-learning update mechanism, including: acquiring historical operating data and anomalies of the electrolyzers according to a preset update cycle; and recalculating at least one of the model parameters, weight coefficients, first warning threshold, and second warning threshold of multiple voltage anomaly detection models based on the historical operating data.

[0043] Specifically, during long-term operation, the physical characteristics of the electrolyzer, such as electrode activity and diaphragm impedance, undergo irreversible changes, causing an overall drift in the baseline value and fluctuation variance of the cell voltage. If the initially set model parameters are used, false alarms (too high threshold) or false alarms (too low threshold) may occur as the equipment ages. To address this, the system constructs a closed-loop feedback update process based on historical data. First, according to a preset update cycle (e.g., monthly, or after each major overhaul), the system automatically extracts historical operating data from the most recent cycle. To ensure calibration accuracy, the system combines "manual confirmation records" or "fault logs" from maintenance personnel, automatically removing time periods marked as actual faults from the historical data, retaining only data from when the equipment is in normal operating condition (including normal operating fluctuations) as the "calibration sample set." Then, based on the above calibration sample set, the system recalculates the statistical distribution characteristics of each cell voltage (e.g., a new median, a new standard deviation distribution baseline), and updates the internal parameters (e.g., Z-value threshold, dynamic threshold baseline) of each statistical model (e.g., robust Z-score model, dynamic standard deviation model). In this way, the system can "learn" the current aging state of the electrolyzer, incorporating normal aging voltage increases into the baseline range rather than misjudging them as faults. Finally, using manually verified anomaly records as truth values, the system backtests the judgment results of each voltage anomaly detection model over the past cycle, calculating the accuracy and false alarm rate of each model. Based on the backtest results, the system automatically adjusts the weight coefficients of each model: for models with high false alarm or false negative rates, its weight coefficient is automatically reduced; for models with high consistency with human judgment results, its weight coefficient is increased. Through this mechanism, the system achieves a self-evolving cycle of "monitoring-feedback-calibration," ensuring that the monitoring system remains at its optimal sensitivity throughout the entire lifecycle of the electrolyzer.

[0044] In this embodiment of the invention, a "multimodal" monitoring approach is introduced, utilizing multiple voltage anomaly detection models to simultaneously analyze the target cell. Specifically, the system combines a "first-type model" that focuses on the absolute change characteristics of the cell itself and a "second-type model" that focuses on the change characteristics of the cell relative to the group. This design effectively addresses the wide load fluctuation problem in wind and solar hydrogen production scenarios: when the overall load of the electrolyzer fluctuates, although the "first-type model" may trigger anomalies due to changes in the absolute voltage value, the "second-type model" can identify whether the trend of the cell's change is consistent with that of the group. Thus, the comprehensive judgment of multiple models effectively filters out false alarms caused by operating condition fluctuations, significantly improving the accuracy of fault identification. Furthermore, this invention no longer provides a simple "alarm / no alarm" binary output, but rather comprehensively judges the "anomaly level" based on the detection results of multiple models (e.g., the degree to which multiple models alarm simultaneously). This enables the system to achieve graded early warning (such as a red-yellow-green light mechanism), providing maintenance personnel with intuitive risk alerts (e.g., low level corresponds to attention and inspection, high level corresponds to immediate shutdown), thereby optimizing maintenance efficiency and equipment continuity while ensuring safety.

[0045] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0046] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A exists, A and B exist simultaneously, and B exists. In addition, the character " / " in this document generally indicates that the related objects before and after it have an "or" relationship.

[0047] The steps described above are for clarity only. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the protection scope of this invention. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, without changing the core design of the algorithm and process, are also within the protection scope of this invention.

[0048] Furthermore, the examples mentioned in the above embodiments can be freely combined, and any combination can be understood as an embodiment. The terms "embodiment" or "example" appearing in various locations in the specification do not necessarily refer to the same embodiment, nor are they independent or alternative embodiments mutually exclusive with other embodiments. Those skilled in the art will understand that the embodiments described herein can be combined with other embodiments.

[0049] Another embodiment of the present invention relates to an electronic device, such as Figure 5 As shown, it includes at least one processor 201; and a memory 202 communicatively connected to at least one processor 201; wherein the memory 202 stores instructions executable by at least one processor 201, the instructions being executed by at least one processor 201 to enable at least one processor 201 to perform the abnormal monitoring method for the electrolytic cell chamber as described above.

[0050] The memory 202 and processor 201 are connected via a bus, which may include any number of interconnecting buses and bridges, connecting various circuits of one or more processors 201 and memory 202 together. The bus may also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 201 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 201.

[0051] Processor 201 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 202 can be used to store data used by processor 201 during operation.

[0052] Another embodiment of the present invention relates to a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the method embodiments described above.

[0053] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0054] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing the present invention, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of the present invention.

Claims

1. A method for abnormal monitoring of an electrolytic cell chamber, characterized in that, include: Real-time acquisition of voltage data from multiple chambers in the electrolytic cell; Each of the plurality of small chambers is taken as a target chamber in turn, and the target chamber is analyzed using multiple voltage anomaly detection models to obtain the anomaly detection result corresponding to each voltage anomaly detection model; wherein, the plurality of voltage anomaly detection models include at least a first type of model for detecting voltage fluctuation anomalies of the target chamber itself, and a second type of model for detecting voltage fluctuation anomalies of the target chamber relative to the plurality of small chambers; Based on the anomaly detection results of multiple voltage anomaly detection models for the target cell, the anomaly level of the target cell is determined.

2. The abnormal monitoring method for the electrolytic cell chamber according to claim 1, characterized in that, Each of the multiple voltage anomaly detection models has a preset weighting coefficient. The anomaly detection results based on the multiple voltage anomaly detection models of the target cell are used to determine the anomaly level of the target cell, including: Based on the anomaly detection results of multiple voltage anomaly detection models of the target cell and the weighting coefficients, a weighted sum is obtained. Based on the weighted value, the abnormality level of the target chamber is determined.

3. The method according to claim 2, characterized in that, The weight coefficients of the second type of model are greater than those of the first type of model, and / or the number of models in the second type of model is greater than the number of models in the first type of model.

4. The abnormal monitoring method for the electrolytic cell chamber according to claim 2, characterized in that, The step of determining the abnormality level of the target chamber based on the weighted value includes: When the weighted value is greater than or equal to a preset first warning threshold and less than a preset second warning threshold, the abnormality level is determined to be the first level; wherein, the second warning threshold is greater than the first warning threshold; When the weighted value is greater than or equal to the preset second warning threshold, the abnormality level is determined to be the second level, and a shutdown control command is output.

5. The abnormal monitoring method for the electrolytic cell chamber according to claim 1, characterized in that, The second type of model includes at least an outlier detection model, and the analysis of the target chamber using multiple voltage anomaly detection models includes at least: Calculate the voltage difference between the target cell at the current time and the previous time. Calculate the median of the voltage differences among the plurality of cells, and the absolute median difference of the voltage differences among the plurality of cells; The corrected Z value is calculated based on the voltage difference of the target cell, the median of the voltage differences of the plurality of cells, and the absolute median difference of the voltage differences of the plurality of cells. If the corrected Z value is greater than the preset Z value threshold, the anomaly detection result is determined to be abnormal.

6. The method for abnormal monitoring of an electrolytic cell chamber according to claim 1, characterized in that, The first type of model includes at least a fixed threshold mutation detection model, and the analysis of the target chamber using multiple voltage anomaly detection models includes at least: Calculate the voltage difference between the target cell at the current time and the previous time. If the voltage difference in the target chamber is greater than a preset fixed voltage threshold, the abnormal detection result is determined to be abnormal.

7. The abnormal monitoring method for the electrolytic cell chamber according to claim 1, characterized in that, The second type of model includes at least a relative deviation mutation detection model, and the analysis of the target chamber using multiple voltage anomaly detection models includes at least: Calculate the average voltage of the multiple cells at the current moment; Calculate the difference between the voltage data of the target chamber and the average voltage to obtain the first difference; Calculate the change in the first difference between the current time and the previous time. If the change in the difference is greater than a preset deviation threshold, the abnormal detection result is determined to be abnormal.

8. The method for abnormal monitoring of an electrolytic cell chamber according to claim 1, characterized in that, The second type of model includes at least a dynamic standard deviation outlier detection model, and the analysis of the target cell using multiple voltage anomaly detection models includes at least: Calculate the standard deviation of the voltage in the target cell over a preset time period. Calculate the median of the standard deviations of the voltages in the multiple cells; The dynamic threshold is calculated based on the median of the voltage standard deviation; If the standard deviation of the voltage in the target chamber is greater than the dynamic threshold, the abnormal detection result is determined to be abnormal.

9. The abnormal monitoring method for the electrolytic cell chamber according to claim 4, characterized in that, The method further includes: According to the preset update cycle, acquire the historical operating data and anomalies of the electrolytic cell; Based on the historical operating data, at least one of the model parameters, weight coefficients, first warning threshold, and second warning threshold of the multiple voltage anomaly detection models is recalculated.

10. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the abnormal monitoring method for the electrolytic cell as described in any one of claims 1 to 9.

11. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the abnormal monitoring method for the electrolytic cell chamber as described in any one of claims 1 to 9.

Images